Jędrzej Przeździęk, OpenGift: How to build retention in e-commerce?

Recorded before the premiere on 20 October 2026, so the guest's figures and plans reflect that time; we leave out the teaser at the start of the video and a short chat about AI video tools, and pick up with the small talk right after the introductions.

What happened with the goats and the farm podcast?

Tomek: Hi, this is Tomek Runowicz.

Maciek: And Maciek Czerniak.

Tomek: We run wecanfly, an agency that helps established brands with complex migrations to Shopify Plus.

Maciek: And this is Flying with Shopify, the channel where we talk to e-commerce experts and founders. Welcome to another episode.

Jędrzej: I bought myself a house in the countryside, bought goats and a hectare of land, and raised them there.

Tomek: So you're the guy from that saying about dropping everything and running off to the Bieszczady mountains.

Jędrzej: Not quite that far, it was near Malbork, and I kept working.

Tomek: And you're saying you made a podcast about it?

Jędrzej: I tried to make a podcast out of it. It's gone now, I hid it. It wasn't the best, I'll tell you. Goats don't walk where you want them to. It seems easy, and then there's no light in the barn, something's missing, and it all looked so tragic that I said: "No, okay, this has to go."

Tomek: Beautiful.

Jędrzej: But it was an attempt.

Why do young customers reject AI-generated content?

Jędrzej: There's research out now showing that 25 or 30% of young people don't trust AI content at all. When they see it's AI, they reject it completely, and another group is simply very skeptical. If they sense in a store that things are made by AI, that the reels are AI, they reject it faster. We're still delighted with it and think it's great, but young people, first, spot it quickly and, second, for some reason dislike it. For us it's great, because we can do things faster and better, so why not? But they seem to see it as fake. The survey only asked whether you trust it and whether you'd buy. They didn't trust it and didn't buy, and nobody asked why or dug deeper. But they don't want it. I think Szymon Negacz also showed that offers where you can tell they were made by AI are much less effective. So we're excited about it, and then it turns out that…

Tomek: You know, that doesn't surprise me. First, I have two fairly grown-up daughters from a very tech-savvy generation, and I often hear from them: "But Dad, that's AI, I'm not interested" or "I don't like it." There's a very negative attitude toward it. The second thing is from our own backyard. You said customers approach AI-generated offers with much less trust. As someone who receives briefs and RFPs from clients, I feel exactly the same. I understand someone using AI to enrich or fill something in, but there are absolutely extreme cases where I can see someone ran a single prompt and out came seven or 20 pages…

Jędrzej: And it makes more work for you.

Tomek: …and it makes a ton of work for me. Then all I'd want to do is reply the same way, with an offer written entirely by AI.

Jędrzej: If it's some small company, you'd really have to reply: sorry, this is our price range, we're not taking part. We recently had an inquiry like that at OpenGift, where we make promotional products. Someone wanted a pen with a pearl, leather, silver, with something else on top, and one unit came out at PLN 60–80. We wrote back asking whether they really wanted that, because it would be handmade work. They replied no, those were suggestions they'd glued into one. They'd made themselves a short brief saying it would be nice if the pen had leather, or this, or that. Somewhere along the way they probably combined it all and sent such an incredible brief that it would have been jeweler's work, not a promotional product. It's crazy.

On the other hand, I like to write down bullet points and say: "Turn this into an email for me, just keep it short and to the point." So AI can definitely make work easier if you give it good insight. It writes great articles for me. I'm writing a book, and I feel it's better than if I'd written it alone. But all the substance, the research and everything else is still mine. That's the difference.

Tomek: As long as you use AI as a tool and not as…

Jędrzej: When you fill a brief with technical requirements you don't understand and send it to a company that understands them well and sees strange, unnecessary or overblown things written in, that's crazy. Better to admit you don't know: "Okay, I don't know this point, let's talk about it," rather than fill in empty points with AI. Lots of people say AI doesn't really know their field, but when they need to write a brief for an agency, that's exactly when they use it. My girlfriend is a vet, and she says: "Gosh, I don't use it in veterinary work to make a diagnosis." And then she asks it where we should go, what to do, what to eat. I tell her: "If it doesn't know your field, why would you trust that it knows this particular one better?" There's some prompt engineering that can do it, grounding answers only in research or something else, but we should still have that knowledge ourselves.

Tomek: Agreed.

Maciek: Two questions. First: was the vet girlfriend part of the goat plan?

Jędrzej: No, that was after. Although when I was getting to know her and told her about it, I could have scored a point. A point lost, a point won, but that was after.

What is e-commerce actually using AI for today?

Maciek: Okay, and the second question. We're talking about AI, and it occurred to me that it would be great to hear what you're doing with AI in e-commerce.

Jędrzej: I recently had a panel discussion with four practitioners, and it's amazing, because contrary to appearances, AI hasn't changed that much on the storefront side. Nobody is yet set up so that AI does the buying itself. People talk about it, but from the user's point of view, not much really changes. From the point of view of the back office and customer acquisition, meaning acquisition sources, there are two huge changes. In the back office you can speed work up very easily. First, programming and testing. I don't know if anyone still uses UXPin or something like that for wireframes, because either we do it in Figma and Figma turns it into CSS, or we just say "code it for me" straight away, test whether it works the way we wanted, and then fix it. We have live HTML that shows a given feature, and then we speed that up. In design work it definitely makes things a lot easier. In writing copy it helps massively too. But the magic happens in the back office, because until now you had to click things over from one system to another, and now you can write a micro-connector that does it for you. You can even drop two or three SaaS apps, because we can write our own that does the same thing, handles tasks, and at the same time checks our graphics and pulls data from somewhere else.

But what I use AI for most is analysis. I upload data and say: "Check for me who bought most often, what was bought most often, what people came back for." Within a minute or two it can produce great reports and hypotheses that used to take two real analysts. On one hand it prepares a table, and on the other it gives five points on why it thinks that happened. Of course, I wouldn't trust all five. In retention, which I now work on regularly, it always concludes that new products have no retention. If you don't prompt it well, telling it these are new products and giving it the date they launched, it says nobody comes back for them. And nobody comes back because they've been on sale for two months, while the purchase cycle is six months. Nobody has had a chance to come back for them yet. But it says these are the top 10 to eliminate because they have no retention, while the oldest ones have the most because people have gotten used to them. So you need to manage that context a bit. Hypotheses are fine, but you still need a human in the loop. I wouldn't make a decision like "Okay, switch off the ones you think we should switch off," because it might switch off exactly the newest ones or ones that are super profitable for us.

Tomek: High-margin ones.

Jędrzej: High-margin, and we want those. Or the ones customers start with. There can be many levels of decision, so I wouldn't just let it run loose. But for analysis it speeds things up massively.

Who is Jędrzej and how did he get into retention?

Tomek: Okay, let's start with you telling our millions of radio listeners a bit about yourself. What do you do at OpenGift and beyond? Some people probably know you, but for those who don't, tell us in a few words.

Jędrzej: That's why I came to you, to find the few people who don't know me yet, because the others already do. At OpenGift I'm the CMO, the marketing director, but I've been asked to work on retention. That's one thing. Second: I run Fundamenty Retencji. It's a blog, training, courses, a whole system, a framework, and a book coming soon, focused solely on retention in e-commerce. I also have my own podcast, which is now a channel and the Buduj E-commerce portal.

Tomek: We're using that to get you on the other side of the table for once, because on your podcast you're usually the one hosting and asking questions. Now we'll be asking you. I told Maciek today that we're smarter: the knowledge you've gathered in all those conversations, we'll collect from a single person.

Jędrzej: That makes some sense, because those conversations are really valuable. Besides sharing knowledge, we get a lot out of them too. I now have 30, maybe 40 interviews on topics I mostly choose myself, so I know who I want to invite. That's great. Oskar Lipiński often hosts too, and you can see it with him: when he was interested in building a personal brand, he invited a bunch of influencers. Then he got interested in building communities, and suddenly Kuba Roskosz shows up, or someone else with a big community or newsletter. It's similar for me: when I'm interested in something, I invite specific people, even a bit subconsciously, because I probably listen to them and think: "It'd be great to talk to them." The topic interests me, so I ask. It works well.

Tomek: As hosts, we've noticed the same thing: if the topic we're discussing genuinely interests us, the conversation is more interesting later, because we're not just playing the role of people asking questions and hosting, but people who are really interested in something and dig into it.

Jędrzej: You can hear it in the voice afterward. When someone is nodding off, you can hear it in the voice.

Tomek: Yes. So we'll try to stay alert today.

Jędrzej: Despite the late hour.

Tomek: Despite the late evening hour. So, Jędrzej, tell us how it happened in your career that you went from being an e-commerce director and manager working in online stores to focusing strictly on marketing. When did that shift happen for you, and the full focus on retention that comes through in your channels today? As you mentioned, you've built a framework around it and a separate portal where you mainly talk about it. Which observation, at which stage of your career, was the turning point when retention, and making sure the customer stays with the store, became the most important thing for you?

Jędrzej: Interestingly, you always work that way subconsciously. As an e-commerce manager you were responsible for every area, so you wanted the customer to come back to you. But back then I didn't know it was so interesting. At that time I had a phase where I said I was a direct-to-consumer person and wanted to work only with manufacturers. Interestingly, the shift also happened because direct-to-consumer companies don't quite get it. They want to sell both B2B and B2C, and they don't realize that as a manufacturer you face completely different challenges. Part of it came from me looking for my place.

But the biggest shift came at a company where we sold sinks, where there's really no retention. My boss came to me and said: "We're doing great, but if we switched off Google Ads, we might not sell anything," because Meta doesn't sell sinks. There were no LLMs yet; it was probably four or five years ago, roughly. He said: "We need another leg to stand on. Maybe retention?" I replied: "But people buy a sink once every seven years." The company was maybe 10 years old then, so we said: "Maybe the first customers would come back if they remembered us." And we started thinking about it. Surely someone's there. We check: 1–2% of customers come back. Why? We check. Often they had a Polish tax ID, so we check their business activity codes and other things. It turned out there was a group that buys more than one sink a year. They're designers, installers, interior decorators, or people with small businesses who install those sinks for others. And we said: "Oh! Then we can work on retention." I didn't know yet that a framework would come out of it, but I knew you could find those customers. In every company there's a group of customers with higher retention than everyone else.

Later, when I thought it over, it turned out you could run special ads for that group. We had ads aimed only at designers. It also turned out that social media was worth running. Before, the sink accounts existed to rack up a few likes, but it was a pretty dead channel, because who would want to like social media about sinks? And suddenly we said: "Okay, they want it, because they want to see what's new and what interiors those sinks were used in." These people like that, they want to know what's new, see the trends and fresh inspiration. Suddenly the channel started to make sense. A whole team worked on it, but only for a selected group. Because it didn't pay to work on the retention of an ordinary John Smith who bought a sink from us for his kitchen, since we'd have to keep up a relationship with him for seven years. I'd look through the whole market again anyway, even if I'd been happy, because after that many years everything could have changed.

Tomek: Of course.

Why does retention start with acquisition?

Jędrzej: That's actually very interesting, because it occurs to me now that in that situation retention really starts with acquiring the right customer. That's very important, because there are campaigns, for example Black Friday or promotional ones, after which customers, as a cohort, have much lower retention than the rest. If an ad is aimed at "buy now," at FOMO, if we regularly run campaigns like "time's running out, someone's about to buy everything," then the customer either doesn't remember, because they bought on impulse, or didn't need the product, or bought it as a gift. When we advertise face creams for Mother's Day, that campaign usually has low retention simply because the creams are bought as gifts. As a son, I'll buy a nice cream because the campaign reached me, and maybe I'll remember when my mom says again, "I want the same one," but for other purchases I won't be the one buying. So the right acquisition is really important here. You have to choose your customers well.

It's the same with products. We think we should advertise the product that gets the most clicks. We look at ROAS, and now we'll be talking about POAS, so that we're actually building profit, meaning we look not at return on investment alone but at how much profit we make on each złoty. But it may turn out that some products aren't that attractive and we don't make that much on them, yet customers come back for them regularly, month after month. In the long run, it turns out that advertising the product that brought in less on the first purchase was the better investment.

Tomek: There are even brands that know full well they make their first sale at a loss.

Jędrzej: Many companies work that way, in services too. The entry product is often one with a very low or zero margin. Winning a project, especially a cheap one, often costs more than we'll make on it. Or in SEO, for the first three months we put in more work than it pays for, but we know we have a contract with the client for the next six months and we'll make it back. That's common. But understanding that we have retention can make us super competitive precisely in acquisition. If we understand that we have a product customers come back for three times, and that we'll make three times as much on that customer within a year as on the first purchase, we can afford to overpay. A classic e-commerce business asks: "Why is your ROAS at a level where my margin doesn't add up?" and thinks you're crazy. And suddenly someone concludes: "This business doesn't pay, because the cost of acquiring a customer is higher than what I make on the product." And they either shut the business down or try to keep it going, come up with new tricks, complain: "Oh God, the world is so unfair." They raise prices, sales drop, and the business stops being profitable altogether. If they'd thought that customers come back for the product three or four times, they'd have said: "Okay, we acquire a customer, on average we make this much from them, so that's what our acquisition cost can be. We can wait." Of course you need a somewhat bigger financial cushion, but that's exactly what lets us pay more.

Why does retention get only 12% of the budget?

Tomek: While preparing for this episode, I read the retention report you have on your website. You present some really good statistics there, including percentages, on how much of their budgets companies spend on acquiring a customer and how much on keeping them afterward. To quote the numbers: 55%, right?

Jędrzej: At least 55% of the budget goes to acquisition.

Tomek: And only 12% to retention.

Jędrzej: Yes. There's often branding and other activities in there too, but retention gets only 12%.

Tomek: For many people in the market it's no longer news that acquisition costs are rising, and for some companies they're reaching levels where, as we said, the first sale has to be below any profit at all. From what you've seen: what has to happen, or why hasn't it happened yet, for companies to change their thinking and significantly increase the share of the budget that goes to retention?

Jędrzej: Better dashboards. I mean better data dashboards. If you look at any ad data dashboard, it shows ROAS, POAS, and if we connect it up, the number of new customers and revenue. But it doesn't calculate LTV, meaning how much we earn on average per customer. If we don't have that data point, we optimize, we grow, and we show: "Look, we have more and more new customers." Neither Google Analytics nor the Google Ads dashboard shows how much we got over the long term from a customer acquired through a specific campaign. And if we don't have a tool that calculates it, what do we focus on? The marketing team has to show some numbers, so it shows the numbers from its dashboards, and since those are what it shows, they become the most important numbers.

The second part: lots of people think retention just happens. We acquire customers and they'll simply stay. They'll like our cream, so they'll stick around. So why should we do anything more about it? And what would we even do? It's somewhat unmeasurable and hard to grasp. That 12% is often remarketing, but if the purchase window is six months, remarketing no longer reaches that far, because it often works within a 90-day window.

Tomek: Most brands run some kind of email marketing.

Jędrzej: Yes, but that's exactly the 12%: someone bought something and they send them emails. If the customer has a six-month purchase window and we've sent them six "buy, buy, buy" emails in that time, they probably won't make it to month six before they hit "Unsubscribe." If we design it badly and send them new offers every two weeks, while they buy a cream they use up once every six months, why would they need an offer every two weeks? So that remarketing or email marketing is done badly: we bombard people with offers that have no value.

Maciek: You can certainly do anything badly, but I also have the impression that many brands still do it well.

Jędrzej: When I ask, very few brands have a rule that after a purchase and sign-up they leave the customer alone for a month. They don't. Once someone is in the database, they go into the cycle. If there happens to be a promotion that week, they get it. There's no such window.

Maciek: On the one hand, it's about building awareness that there are other products.

Jędrzej: If you sell in some kind of bundle, you should know, for example, that if someone bought a handbag, we don't bombard them with handbag discounts for at least a month, because that will only annoy them.

Maciek: Agreed.

Jędrzej: We know, for example, and again AI pulled this out for us, that if someone bought a handbag, their next purchase will be in six months. If we want to shorten that, maybe we should bombard them with offers after four months, but not in the same week. Right now, instead, we can send them a nice email: "Thanks for joining our list. Great choice," and confirm why it was a good purchase. We reassure them so they don't want to return it. We smile, we show things, we give value. I think 8a.pl once had a sequence where they sent material on how to clean shoes or how to care for a down jacket. I thought: "Thanks, I just bought shoes from you." Nice, you could tell they knew what I'd bought. That's a nice benefit. That's worth thinking through.

Second thing: if we know what customers most often buy next, let's not show that item right away. The first email shouldn't feature the product they add on. It's worth making sure that after signing up for the newsletter the customer doesn't go straight into our usual, standard biweekly mailing cycle, and instead, for example, designing the first four emails to be the most valuable, with lots of links and the whole story of the company. Only after those four emails should they enter the cycle. A kind of customer onboarding. Tell them we also have a great Facebook group, encourage them here, show them something there.

Can AI tell you who shouldn't get a discount?

Maciek: I definitely agree you have to do it smartly, so you don't throw the baby out with the bathwater. But let me give you a great example I heard recently. Klaviyo just had a big conference of its own in Boston, where they showed a case from a large brand, a cosmetics one I think, I don't remember the name. They showed what their AI tool, a kind of composer they built, can do. A marketer had a newsletter ready for a segment of a million profiles and, as usual, wanted to send a 20% discount coupon. But before sending, the composer checked and found that out of that million, 300,000 people had never bought with a discount. It suggested, and immediately set up itself, a split: the coupon only for those who respond to discounts, and for those 300,000 the same message, slightly changed, but without the discount code.

Jędrzej: That's great. On the other hand, the question is whether they really don't buy with a code, or whether they just don't enter it and then forget. We have lots of tests on websites: you show a coupon, even if it's glowing everywhere, 20% of people see there's a promotion, often buy because there's a promotion, and then don't enter the code. They didn't get around to it, didn't enter it. I agree that some people definitely don't like promotions. Especially with somewhat more premium cosmetics, showing that they're on sale can lower the brand's perceived quality for them. In a case like that, showing a coupon could even backfire. It really is great to separate out that group and then maybe test with and without a coupon. It's great that the tool already notices that and can make those kinds of…

Maciek: For me that's probably the most important insight: AI can already genuinely help us make better decisions, even without being asked.

Jędrzej: Yes. That's exactly why we talked about AI at the very beginning, it was a kind of intro, because you can generate a lot of hypotheses. And then you say: "Oh! So either I won't send it, or I'll send it and see whether they really don't buy." Or maybe they need a much bigger discount, because they only respond to a bigger one, and we really want to push them to buy. We recently had exactly that discussion at OpenGift, whether to offer a discount to grow a bit. We said no, because we calculated how big a discount we'd have to give and how much sales would have to grow. We're in a fairly big peak right now, and when looking for a bit of extra traffic, we can't say: "Let's just lower prices for everyone," because that starts to be unprofitable. You'd have to think about who hasn't bought but is in the database and opens emails. Pick out a group that potentially left because the price was too high, and try to win over only them. For now that's still hard.

Maciek: That kind of multidimensional analysis with AI is already entirely realistic.

Jędrzej: Yes, we can already tailor communication better. AI can search through all of a customer's purchases, check how many times they bought and in what cycle, and really tailor things much better. We used to think in terms of statistics, and now you could do it per customer: how often they should get emails, because we can see they don't open them or only open every few, and with what content, because we could check what their open rate is on which topics. We used to run an A/B test and find that half like a subject line with a question and the other half like icons. Now we could run a few tests and say: this one responds to icons, this one to questions, this one to numbers in subject lines, and split the list into groups that respond better. We used to decide that numbers in subject lines were best and then send everyone subject lines with numbers.

Who in the company should own retention?

Tomek: That's true. Let's go back to the current reality of stores and e-commerce managers. If looking at budgets clearly shows that retention, partly because of incomplete data, gets so little attention: among the clients we've talked to, and we work with hundreds of brands, I haven't met a single company with someone whose specialty is retention manager. Do you think it's a role that should actually be separated out? Or is it a competence within the e-commerce manager's job, or the team that handles e-commerce, acquisition and customer service? Obviously the bigger the scale, the more it matters, but by default: is it a big enough piece of the business to need its own owner?

Jędrzej: Sometimes they already exist. Big brands have such people, but they usually look after loyalty programs, so they're loyalty program product managers, and retention is somewhat in the background. If a brand has an app, it knows the app has retention, so it takes care of it. Should the role be separated out? I don't know, but it should definitely be named. We should say: you're in charge of retention. Whether it's the marketing director or someone on their team, somebody should be accountable for it.

I ran a survey of more than 200 stores. The results will probably come out in January, we're preparing a report, but I can already say that I think only 10% of stores have a designated person responsible for this. If you don't have such a person, who is realistically going to check it every month? So to start with, let's point at someone and keep checking whether it's going up or down. Let's keep a finger on the pulse of four simple numbers: retention, frequency, time to second purchase and LTV. We see whether they're rising or falling. If they're falling, why? If they're rising, maybe there's nothing to do, or maybe it's worth looking into, and then we pick that person.

What's worrying is something my framework shows. If we take someone from marketing, and a specialist rather than a director at that, they have very little leverage. I say: you need to pick a better product, change your customers, improve customer service, build a community, give something for free, and only at the very end code a loyalty program. And they answer: in my area I can send a newsletter, and maybe I'll get the loyalty program if I talk the boss into it. If we anchor retention in the wrong place, or don't say it's our strategic goal that the organization should adapt to, it ends with here-and-now activities. Marketing automation is a nice activity, because we already have a customer base, we grow it and try to activate it. It could live entirely inside marketing, which is why so many people start there.

Retention is often linked with loyalty. We fight for customer loyalty, or at least that's what people say, and right away we think: loyalty program. If we went to most store owners and said we're working on loyalty, they'd probably jump straight to a loyalty program in their heads. So there should be such a person and they should measure it, but it also has to be company strategy, otherwise it won't work.

You see this most in the third point of my framework, which is experience. A customer has a lot of touchpoints. One is advertising, where we show what's needed to pick the right customer, as we've already discussed. The second is pre-sales service: was it good enough? Then post-sales service, and even logistics. Did it arrive on time? Was it nicely packed? If you get a parcel that's rolled up, dirty and three days late on top of that, you no longer feel like buying from that store again. Even if the product is valuable, the first impression is ruined. You've botched my first impression, I take the product out and I already don't like it.

I have an anecdote. I once bought running clothes from a big store: ultralight running tops, terribly expensive. Adidas packs them in plain plastic film, without even any print, because in stores they get hung on hangers anyway. On top of that, the store threw it all into a poly mailer. Shortly after, I bought a top for my son at Decathlon, and they packed it neatly in a thick little box with a tear strip, with a leaflet of recommendations and information on how to return it inside. A better product, at least in how it felt. Of course once you've run in it you have your own impressions, but if at the back of your mind you think you probably overpaid because of how they shipped it, then while running it's hard to tell whether the top really is 8 grams lighter or breathes better. That feeling of having overpaid stays with you.

That's what experience is: did they ship on time, was it clean, was it packed well. But also, for example, did accounting handle the refund properly? Did they issue the invoice on time, and if there was a problem with a correction or a return, how did we handle it? Krzysztof Kukla always says that after the first complaint or the first return you can judge a company. If it handles it well, next time your threshold for coming back is much lower, because you know that even if something happens, and at the back of your mind you're always worried the size won't fit, they'll accept the return or complaint without a fuss. And then you say: "Oh, I'm coming back here."

If you say, "Right, we're building retention," maybe nobody will understand. But you can go to the accountant and ask: "What can you do to make customers want to come back more often?" I, for one, would come back more often if refunds took three days. Same in logistics: on time, and ideally showing when it will arrive, because sometimes I can wait three days, while for furniture people wait even two weeks. If it's two weeks, it's good to show that it's exactly 14 days, not 21, so the customer can plan. And if it usually takes seven days, let's not write 14 just to be safe, because someone may need it in 14 days and the parcel arrives after seven, when they happen to be on vacation. Let's design it and show it. Logistics will say too: "OK, we can do that." If we set ourselves the strategic goal of wanting customers to come back, everyone can work on it.

What are the Fundamenty Retencji and why is product at the very bottom?

Tomek: Yes. You've already started talking about your framework, and that's exactly where I wanted to go. I get the impression that many people start with this equation: retention equals a loyalty program equals a mobile app. I have a program, I have an app, so I take care of the customer.

Jędrzej: And I've got retention.

Tomek: And I've got retention. You think about it differently. You've created your own framework for building retention deliberately. Let's talk about it: what it's made of and where you start, and we'll ask follow-up questions.

Jędrzej: It's called "Fundamenty Retencji," the fundamentals of retention, and not by accident, because you should start from the foundations. I arranged it into a pyramid and I say: if you don't do what's at the bottom, what's at the top may even work a bit, but much less, or not at all. At the very bottom is the product. We've already talked about it a bit. You can choose a retention product, say razors. There used to be Dollar Shave Club, where people bought regularly. I think they were the first product subscription, or at least the first really well-known one. So you can take a product people simply come back for. Great.

You can also take a product people want to build on. Here you have PlayStation: you buy a games subscription, extra games, a controller here, some other piece there. You don't always need a consumable product. You can build a product with an ecosystem big enough that you keep buying add-ons and building them around that first product. There used to be a myth that PlayStation doesn't make money on the console itself. Apparently they make decent money, but they spend a lot on advertising precisely to sell the console. And now the console has apparently even gone up in price, because GTA VI is coming, and it's more expensive than a year ago.

Tomek: We're getting ready to book time off.

Jędrzej: Yes, I'd play too. There's also a third type of product: fashion. Can I have three of the same product without it really bothering me? I have a third pair of jeans, but now a different color is in fashion, so I'll buy another. You can plan for that. But the most important thing isn't that the product is repeatable, it's that it keeps its promise. If the razor were poor, I wouldn't want it, even if it wears out. I'll take another, because I have alternatives. If the console didn't meet expectations, nobody would buy more games. And when I want beige trousers, I'll buy them from someone I like and know has the same quality. So right from the start the product has to keep its promise, so that people want to come back for it.

Why is it so important and why is it at the bottom? Imagine a great loyalty program or, closer to the base, a great customer experience. We serve the customer brilliantly, we call, we check in, we ship on time. Everything is great, except the sink arrives cracked. The customer says: "Damn, that's poor." Soon the second one arrives, on time again, everything great, I called, and they say: "We know, they often crack." The second one is cracked too, the third arrives scratched, and in the end it turns out it's just poor quality. I say: "You know what? Give me my money back." Even though the service was great. Even if I'd picked a group of customers who wanted to buy regularly, I could do a million things, and the customer could get another 100 points for every sink in a loyalty program, to exchange for something. But the product doesn't keep its promise. It wasn't what I wanted. Without a good product there's no retention. Whatever community you have, you'll sell maybe two things.

Right now there's this whole buzz around Książulo's pizza. If it isn't tasty, it'll have a big spike, because everyone will buy it. Everyone likes him, but nobody will buy it a third time just because Książulo says in another video that it's good. If it doesn't keep the promise of being a good enough pizza, people won't buy it again. The first time, everybody buys. He has a reach in the millions and he'll make money anyway, but you won't keep going back for it.

How do you build retention when a product is bought once every many years?

Tomek: OK, but what about products people buy once every 10 or 20 years? Furniture, for example.

Jędrzej: You'll have it much harder, but people buy different furniture: for different rooms, for different things. You can try going in that direction. Then you have to widen your range, and the retention isn't for the same product: we see that someone will also buy a bed. IKEA has retention. When I need to buy furniture, I go back. Is it a retention product? No, and yet it turns out it has retention. On top of that it has accessories people buy downstairs, they come back for plates. So it's not true that such companies can't have retention. Nor is it true that they can give up advertising and acquisition, because they really do have to bring customers in more often and more regularly.

An easier business, with retention built in, is beauty or pet food. Pet food probably has the highest retention by every benchmark, because once your dog likes the food, you don't want to change it. You'll change something for yourself just for variety, to shake things up in life, but not for the dog, because you know it then has a better diet, better digestion and no problems at all. That's why dog food subscriptions were probably among the faster-growing ones. Once the dog likes it, you can buy it for a long time.

Tomek: Yes. I have a digression about animals. I've had a rabbit for eight years. When we picked it up from the breeder, the lady wrote down on a slip of paper which food to buy. Eight years, and to this day only that one. The rabbit is so sensitive to a change of food that when we buy a new bag of the same food, we have to mix it with the old food for a month, because it won't eat the new one.

Jędrzej: Because the old one has lost its smell.

Tomek: Because it's already something different. Exactly.

Jędrzej: That's exactly why pet foods come straight away in a junior version. Breeders are given starter packs for free to pass on to their customers. Breeders get the food for free, and big breeders get older and younger dogs used to one food from one brand. Then the brand grows up together with the dog.

How do you choose customers who will keep coming back?

Maciek: Back to the topic: what's next?

Jędrzej: First was the product, second is the customer. As we said, we can pick customers who typically buy on promotion, and you see it in beauty. There's a shelf with creams that are always discounted. Lots of people at Rossmann buy the one on promotion. They come in and say: "I have PLN 50." The brand doesn't matter that much to them. They look for the one that was the most expensive and happens to be marked down to PLN 50, because it fits their budget. If the cream used to cost PLN 150, then next time, with PLN 50 in their pocket, they won't buy it for PLN 150, even if it worked well for them, because to them that's often overpaying. These are promotional customers who hunt for deals or bought by accident.

If we pick the wrong customers, then, just like with the sinks, they won't buy a second time. We can find a particular group, and if we work in that direction we can assemble a really good retention group. If we shoot blindly, left and right, at every occasion, say we're a beauty brand but run our biggest campaigns around gift occasions, those customers usually won't be retention customers. We talked a bit about that cream bought as a gift for mom.

Maciek: So how do you identify that kind of customer?

Jędrzej: Let's try to connect ad campaign cohorts with where customers came from and what LTV they have afterwards, meaning how much we earn over 12 months on a customer who came from a given campaign. That's how we try to identify the campaign and the customer group. If we can't do that, let's try to single out, within our customer base, the people who buy most often. We see they're women. Now, when we set up a campaign, even if it's a bit more expensive, let's set it up so that more women see the ad. Or, if we can check age, then by age. If it's meant to be companies with a tax ID, we set up a lookalike based on companies with a tax ID, or use that information.

If we know how it should look, we can do it. We can change all our messaging. Say we want to reach companies: maybe we stop showing smiling children and start showing people in suits, so that viewers feel it's a business case. Even with campaign creative we can signal whom we want to attract. And once the campaign learns by itself that these people click best, it will redirect the message on its own. The message alone lets us choose whom we want to find.

I had an example like that in a company where we sold cardboard boxes. We sold them to everyone and it was going great. In Q4 we were very rich and happy, then the rest of the year dragged a bit, and again Q4 was great. But it turned out there was no retention in it. Q4 customers buy gift boxes, much more expensive and more profitable, but they don't come back, and if they do, it's the following year and they look at price again. You have to remind them often, because someone who bought a box for a corporate gift doesn't remember where, since they bought so many things in the run-up to Christmas. Besides, Q4 is generally a period of lower retention, or we have to plan for it right away.

With the boxes we said to ourselves: "Look, there's a group of e-commerce businesses and they buy regularly." Maybe they buy at a lower margin and care more about price, but if we give them that price and the boxes are good, they'll get used to it and keep coming back regularly. We got to the point where almost the whole company's messaging was boxes for e-commerce, while the gift boxes still went into advertising in the Christmas season and sold anyway. We started going to trade fairs and showing all our branding around boxes for e-commerce. We looked after them in our messaging and in SEO, while the gift boxes kept selling, were very profitable and were a nice extra profit. We didn't close that leg of the business, but we no longer built whole campaigns around gift boxes, because for the next nine months they simply wouldn't sell.

Maciek: I think in Shopify that would actually be easy to pull out thanks to RFM segmentation.

Jędrzej: That segmentation is partly exactly this: we look for who those customers are. It's usually set up for marketing automation, but if we pulled out those customers a bit earlier, saw what the groups are and ran ads or lookalikes for them, it would work without a problem.

Maciek: In the sense that I can check what a specific segment, say the champions, buys and how often.

Jędrzej: We skipped this a bit under product, but we mentioned that there are products customers come back for. Like breakfast cereal. I have an example with cereal. There's Nesquik, which says: "There's a TV ad on now," and everyone's buying it. Soon Chocapic has the ad and then everyone wants that one, because kids see the ad, go to the store and say they want that cereal. There are also the cheapest ones, Carrefour's or Biedronka's own brands, which often give the chains their margin: tasteless black balls, similar, but they don't dissolve and don't have that taste. And somewhere in the middle there's, in quotation marks, that Lubella with Kubuś in the drawing. It isn't advertised, but its taste is already very close to Nesquik. Once you get used to it, you say: "PLN 4 cheaper than that one, tastier than that one," and you keep going back for it. You're used to it, you don't need to see an ad, you want that cereal and you don't change it for a long time. Unless someone gets bored and mixes something in, but they more or less know this is the segment, the shelf, they like and buy.

So you also need to look among your own products. As you said: see what they buy. Retention products often aren't the obvious ones. If we sell fashion, for example, let's see after which product customers go on to buy something else, meaning which one is good enough to open our brand up to them. Say I sell handbags. One model makes customers want to come back for another handbag, because it's comfortable, high quality or well thought out. If someone buys it first, they'll like our brand. Another one looks great in photos but doesn't meet expectations, has a very poor result, and we should, for example, improve it.

What kind of customer experience makes people want to come back?

Maciek: We have the product, we have the customer. Next step.

Jędrzej: Experience. We've already said a bit about how the whole company has to be involved. The idea is that the customer meets us all along the journey, and at every stage we should do everything we can to make them want to come back.

Maciek: Sure, so take care of a great customer experience, good shipping and nice packaging. I think the back-end software matters too, so that nothing breaks along the way.

Jędrzej: So that the customer gets what they wanted. So that they get emails at the moments they're needed: "We've received your order," "We're shipping your order," "Here's your tracking code." That's concrete information. Not too much, because when I buy on some big marketplace, I get eight emails almost one after another: that you've paid, that we've accepted it, that we've confirmed something. And I think: "Gosh, I don't know where my invoice is." That can be overdoing it too. So experience is also about the number of emails, but also whether, when I open the order email, I can see when I'll get it, whom to contact if I don't, what I need to do, and when I'll get the invoice. If all that information is nicely presented, say there are three little dots showing I'm at the first stage, two are left and then the parcel will be with me, I feel looked after. That's experience too.

Is a newsletter already a community?

Maciek: Right. We have experience. What's fourth in your framework?

Jędrzej: Community. We build a community, we build even a newsletter channel, because that's also a kind of community. On the one hand we should build a community where someone is happy to belong to our, in quotation marks, family, or to buy from us. Like with Subaru: people wave at each other.

Maciek: I know, I wave.

Jędrzej: You have a Subaru? Well, that's how it is: Subaru drivers wave at each other when they pass. Someone wears a particular T-shirt because they feel they belong to a certain group, a certain status. If we could build that with our brand, and that's partly branding, it's a really great thing. I also keep that community going through social channels and newsletter channels. We build contact with the brand and give people value so they want to stay with us. I've put this somewhat under the word "community," because I can see you're making a face, but it can't be fully described. You're not buying it?

Maciek: No, because it seems to me that a real community is still something more than a newsletter.

Jędrzej: It is more than a newsletter, but how do you build it? On the one hand, we can set up a Facebook group and have 30,000 engaged people there. If your customers are women, it's much easier. Among men, I don't know of a big community that forms around a brand.

Maciek: I think there are some male topics, bushcraft for example.

Jędrzej: Maybe, but it's much harder to build that around a brand. Maybe they have their own club, their own community. But a community can be, for example, your podcast, which is partly a company one, the fact that you shoot videos, that you do a particular thing. It doesn't always have to be a 100% community where people talk to each other and build something together, know each other and start conversations. It can be a group you know is there, that wants to be with you, and for whom it's also a kind of branding. It's hard to name.

Maciek: So it's more about…

Jędrzej: Belonging, or some other word.

Maciek: Maybe not brand ambassadors, but some kind of brand advocates.

Jędrzej: That would help too. For me it covers a bit of the newsletter space, a bit of the social space, a bit of brand ambassadors or people who'd like to show off. On top of that there's the feeling, because apparently there are three levels in buying: what I feel when I buy, what others think of me once I've bought the product, and I think how I'd like to feel. I don't remember exactly anymore, but two of those points are external. Take BMW. Some people like it because it feels nice inside. Some feel good because they know they have a BMW, because they always wanted one. And some like getting out of it, because people see them getting out of an expensive car, showing off. That's probably not great, but lots of people buy many things for show. Building a brand that makes people want to feel good on all three levels would be brilliant.

Maciek: It seems to me that building that kind of real community is achievable for a very small percentage of brands. For me an example of a company that did have a community is the cycling brand Rapha, which organized races, events and community meetups, where people actually interacted, rather than being just an imaginary, abstract online community.

Jędrzej: Probably, but closer to home in Poland: there's a nail brand that has a community around nails and everything else. People watch the live streams, write to each other, know each other. They may have around 30,000 people. I had a creator on my podcast who has 40,000 people interested in plants. That's kind of enough too. Why do people buy from him straight away when he posts a reel? Because he posts it right there, people swap ideas, and he talks a bit about where to plant what. You have 40,000 people interested in flowers and outdoor spaces, because he has his own outdoors, and they discuss things a bit without him. And when it's time to buy, they buy from him, because they already like him and know the product will work.

It doesn't have to be a community you'd give your life for, because sometimes that's how it sounds: that it already lives and breathes, and the brand is somewhere on the side. It can be a community focused on sales. There's some dress company with a community where you can also swap, because customers then sell the brand's dresses to each other.

Maciek: We have one like that.

Tomek: We know them. Hello to The Odder Side.

Jędrzej: So that's a community too. People know it exists, that they belong to it, they feel it a bit, and it works.

Maciek: Fair enough. That's a good example too.

Tomek: Yes. At that level, absolutely, because the definition itself is quite broad.

Jędrzej: I wanted to name each level with a single word, and that isn't easy.

Maciek: I think in English the word "community" is broader and more precise than the Polish translation, "społeczność."

Jędrzej: I think so, but it's more about the people and that feeling connected with the brand. About how I feel with it, so it's broad and hard to describe.

Maciek: My problem with it is that if I had to tell someone right now what to do, step by step, to build that community, maybe it's a lack of knowledge, but I'd just…

Tomek: You wouldn't quite know how to approach it.

Maciek: When would I know what the checklist is, and that I can say I've got it?

Jędrzej: You know what, if you have a list of 20,000 people who read your newsletter, then you have a community, for example. I have a community that reads me and wants to read me. I have a group of people who know me, come back to me and want to read me. That's why the newsletter is listed here. If I have a weak product, they'll unsubscribe quickly, and if I write nice, funny texts, maybe they'll stay, but they still won't buy. I'll have a community that doesn't sell. That's why product is at the bottom and community higher up. But even a newsletter list of 10,000 people or a Facebook group where I can get my message across to people who already know me and who, with some encouragement, may come back to me, is a kind of community.

Maciek: OK, I think we can leave it at that.

What is a gesture and why is it so hard to scale?

Jędrzej: Yes, that may have been philosophy and debate already. Next is the gesture. A lot of people disagree with it, they say it's difficult and doesn't scale well. But a gesture is when I add something extra to an order. I do something the customer…

Maciek: A return form.

Jędrzej: Maybe like this: I have a return form, I didn't expect it, and someone put in a slip explaining how to return. There used to be a problem that people were afraid to buy clothes online, and I think it was Zalando that first offered 100 days for returns, plus an easy box to send it back in. That broke down a certain barrier. But at some point it stopped being a gesture, because everyone knew that's how it works. It became an action that builds experience, so it moved down the pyramid. A gesture is something I didn't expect.

Why did I include it at all, even though it scales poorly and is very hard? Because recently at my son's birthday party I said I work in e-commerce. He was running around, and we parents were sitting on the sofas. If you have kids, you know that's unfortunately how it goes. One of the dads says: "Oh, e-commerce. I only buy on marketplaces, but Rubik's cubes I buy in one cube shop," I don't remember the name anymore, "because not only do they always throw in a coupon for my next purchase, they also add some little thing, a cube worth PLN 3, or some gummies." And he's telling me all this. When did anyone last tell you about an online store? And he named everything.

Then I talked to Sławek Mokrzycki, who runs Acus Med, and he said: "We'd put in funny notes and throw in some needles to try, because we knew someone was a physiotherapist." And those people said: "OK, nice store." And they recommended it. Because at a barbecue you won't say someone handled your complaint brilliantly, or that they simply did something well. Poland today is so used to fast payments that people won't buy without them, rather than treating them as a benefit. Fast payments, 24-hour shipping, parcel lockers: nobody sees those as differentiators anymore. If someone thinks that's their USP…

Maciek: OK, but look, it also works the other way: a gesture can very quickly become an expected feature, right?

Jędrzej: Yes, because a gesture is supposed to be unexpected, nice, surprising. It shouldn't then become part of the standard service. If we always add a small sample and someone gets really used to it, the first and second time can be a gesture, but the third is already a bit of an element of experience, because I have a good experience and I know what to expect. That's why the gesture is much harder to pin down within the whole framework, but I wanted to include it, because we can, for example, call and ask how it went. We can send socks on the first anniversary of a purchase. I always joke that Sklep Biegacza owes me money if they do this: if someone buys shoes for more than PLN 500 or 1,000, then every year, on the anniversary of when we met, we can send them socks, because we know their size. Just so they get something. Or a T-shirt, if it's expensive stuff, so they run in it, because they'll be advertising us on top of that. These things can be a bit randomized. I think Coffeedesk used to draw things on its boxes.

Maciek: And now at checkout you can even pay an extra PLN 5, 10 or 15 for a better version of that drawing.

Jędrzej: Yes. But back then the drawing was the kind of thing someone would photograph and post online. It was like free delivery, nice packaging or anything else. That's exactly the kind of gesture I didn't expect. It isn't part of standard delivery, and it can make me remember and come back. It's a differentiator, and that's why I say it's hard to scale. Imagine Zalando having to draw for everyone. It would be hard, but there was something. I remember that at one point they used to sign who packed the parcel. There was a time when lots of companies did that: the packer had a marker. How long could it take? Three seconds? And the parcel always felt a bit personally packed for you by Jacek. It's a small thing that can change a lot. And that's the difference in retention: you'll even brag about it and remember it. You'll subconsciously value good delivery, but you won't talk about it, and often you won't even remember it yourself.

Maciek: You'll remember it if it's bad.

Jędrzej: Yes.

When does a loyalty program make sense?

Jędrzej: And the last one left is the loyalty program. Only once I have at least four of the pillars in place (we can let the gesture go for now, although I can already do a bit of the gesture too) do I say: right, I'm designing this so that customers come back to me. I'm reinforcing what I built underneath, because a loyalty program won't achieve anything if all those points aren't done. Once I already have some habit, at least a gentle habit of wanting to buy everything there, the loyalty program tells me: "Oh, I'm still collecting points, I'll still drive over there."

It was probably like that with fuel back in the day. BP had its card, you collected some rewards, 5 million points to get a cap. But once someone had started and had that card, and those were the early days of loyalty programs, they always pulled in at BP. If it had turned out to be PLN 0.15 more expensive, though, they wouldn't have pulled in anymore. Everything has to be well matched, and ideally it should be a customer who drives and uses fuel, because giving a card to a lady who doesn't have a car and came into town for a hot dog misses the point. "I'll have a hot dog, and then a loyalty card, please, I'd like to collect points." How much can you rack up on that card with poor hot dogs or a beer at 11 p.m.?

That's why a loyalty program is great for reinforcing and it works well, but in many companies it shouldn't exist. If I sell sinks, why would I need a loyalty program? Maybe a referral program then? That's a variation of loyalty: I want to recommend something to someone. It isn't quite retention, but it's another way of acquiring customers. If we do everything well, then even if the customer doesn't buy again, they might say: "Damn, that was a good product, get one too." Anyone who has done a renovation knows that you always ask other people who have renovated about their crew and where they got this or that. When you're in that mode, you go around different houses, look at nice solutions and often ask where they got them.

That element of loyalty doesn't always have to be points. And if it is points, do it wisely. Don't just give points for a purchase and then a discount, because that's crazy. But maybe for new products bought in the first two weeks you get double points, so that the customer wants to look for new products. Or like in computer games, where you get something for logging in every day. There was also that language app. What's it called?

Maciek: Duolingo.

Jędrzej: Duolingo. You got lots of points for streaks, to build the habit of coming back in. So we look for what we want to reward the customer for. Not just for a purchase.

Maciek: So non-financial ways of rewarding as well. We recorded two episodes on this with Marton from Love Loyalty and looked at a few different trends.

Jędrzej: Yes, non-financial, but even if it's points, then more points for buying new products, more points for buying the specific products we care about.

Maciek: Your own brand, for example.

Jędrzej: Or for buying bundles. Say we decide we'd like customers to increase their average order value. So we try to structure the loyalty program to reward more those who spend more with us in a single purchase. Packing is expensive, we have a hidden delivery cost, and we'd rather the order went over PLN 150. So we say, for example: "An extra 100 points for an order over PLN 150." And then someone says: "Oh, I'll add something." We didn't give them yet another thing, because let's say they already got free delivery earlier; we reward them for that instead. We have to think about what we'd like the customer to do and reward them for it.

If we have an app and someone can buy regularly, then even just opening the app every day could earn streaks: another 10, 12 points, and let the amount they get for each day keep growing, as an extra. Because above all we want them to get used to coming in, browsing, doing something. We have to think about which behaviors we want to reinforce and do that with the program, not with loyalty on its own. If I get 1,000 points and most people don't know what they can exchange them for or what to do with them, that's crazy.

Companies that do loyalty programs have statistics saying that 80% of customers use loyalty programs. Except that all it took was a click, or you got the points automatically. So, sure, you got them. But in the report we talked about, it came out that 67% of stores want a loyalty program or already have one. Then we asked the same people what they think is most valuable for retention, and only 12% chose the loyalty program. Meanwhile 57% say the product is the key to retention, but I think only twenty-something percent checked which products bring them retention. That's crazy.

Who took part in the retention survey of Polish stores?

Maciek: To close the framework thread and go back to the survey you mentioned: are these only brands from Poland, and how were they recruited?

Jędrzej: Yes, it's a Polish-only survey. I had 20 partners, from bigger to smaller ones, and we sent it out to our own customers, in newsletters, everywhere. The opening question was whether you run a store. If not, you got a shorter survey; it will probably show what those people ticked, but that interested us less. If you ran an online store, you went through 25 questions. We wanted to check not what people say about retention, but how stores actually operate, what they pay attention to and what they check. We wanted to see the state of Polish e-commerce, not what people declare. Because if you ask people whether a loyalty program convinces them, some will probably click yes. It's the same when you ask why we lost a tender, especially in services: it's easiest for everyone to tick "price."

Maciek: Or that the bids were rigged.

Jędrzej: Yes, but they tick "price." And when you dig a little deeper later, it turns out they didn't know something, or didn't understand why this price was higher than that one. If you ask: "Did the loyalty program convince you?" someone will say something. But do we as store owners even think about how to work on retention and what it gives us? Because, as it turns out, looking at LTV, it gives us a huge amount: in the future it's the only profitable sales we have.

Maciek: I'm asking whether it's only Poland because I have a feeling that in Western Europe or the US, where e-commerce has been ground down more by acquisition costs, the balance of forces looks completely different.

Jędrzej: There it's already natural. People say that for many companies it's natural for the first transaction to be at a loss, but in Polish e-commerce that isn't true. Apart from people who understand retention well, I don't know anyone who knows that we have to make the first sale at a loss. Most stores keep hammering ROAS and go under if they haven't set things up so that they make money on the very first transaction. Over there it really is natural that we acquire customers at a bit of a loss. That's the first thing.

On the other hand, I was afraid you'd say e-commerce there is developed, rather than ground down. Because when I sometimes visit those stores and see what they look like and how much their UX resembles the nineties, I can't believe it. In Poland we have really well-developed solutions in probably every category, from reviews to payments. We just don't quite understand the long wave. In my opinion that's our biggest problem.

Maciek: Understanding is one thing, but we simply haven't been hit that hard by those costs yet.

Jędrzej: Maybe. For many years e-commerce simply grew. Every report showed 30% more customers year on year, 30% more baskets, that's how much e-commerce was growing. Whatever you sold and to whomever, it always grew, and the pandemic boosted it even further. A new player would come in, but the whole market was growing, so the others just didn't grow, and the newcomer found a slice of the pie. That's gone now. If you want to sell anything online today, someone else probably won't sell it. It's a real knife fight, and improving your average order value by even 10%, which means your ROAS gets a bit better, lets you pay more. So retention is one way, and average order value is another good one.

How do you measure retention?

Maciek: Let's talk a bit about the metrics. How do you measure retention? You said it's worth looking at those four parameters once a month, once a week.

Jędrzej: Once a week, no. Let's start with the first number. The first number is the time to second purchase. We check it for our cohorts or for a given product, but to begin with we can also pull an average or a median. The average, so that nobody says we're using difficult words, although I'd prefer the median. How long after the first purchase do customers buy a second time? If it's 90 days, then no change we make now will work faster than 90 days. Even if we've greatly improved service quality, the customer will only come back in 90 days on average. Cream or pet food gets used up at its own pace, and even if someone served you well, you won't say: "Oh, I'll buy more for that rabbit and let it sit there." When it runs out, I'll buy more. Retention works a bit in slots: our changes aren't visible here and now. In acquisition I tweak an ad a little and immediately see it's cheaper; I change something and you see it faster. Here it's a process, so time to second purchase is a very good metric.

First, it's worth shortening it and thinking about how to do that. Second, it's worth doing your strongest activities toward returning customers at around 80–90% of that time, depending on how long it is, to remind the customer about our brand just before they start looking. Say we have a pet food store. The customer has chosen a food, but 10 stores sell it. If you send them a message: "Your food is probably running out, because for eight years you've been buying every two months and you have two weeks left," they say: "Oh, I think that's right." And if you also give them an easy "Add to cart" or "Reorder" button or whatever, then partly out of habit and partly because they can't be bothered to search, they'll buy sooner. Otherwise they'd type "show me" into an LLM or Google, see the prices and could buy the cheapest, if you haven't built a community, a relationship with the store or anything. If we get slightly ahead of that, we hit the spot perfectly.

It's a good checkpoint. First, so you can shorten it, because then the business makes more money, since the customer buys more often. Second, so you know when it's worth communicating with them in a strongly transactional way. We can communicate the whole time: show the offer, teach, educate, show why our company is great, give some value. For a running store, it's nice to show running events, something the customer would want to read and watch. But in that last stretch, a stronger transactional push would be the way to go: a discount code, or news that certain shoes have come in, new arrivals.

Once we have that, we look at the second metric, most often called retention. It tells us what percentage of customers come back to us a second time. This repeat rate is simply the share of customers who come back a second time. The third, fourth, fifth purchase doesn't count; we only check how many customers come back to us two or more times. Why does this matter? If we have 100 purchases and 90 customers in a store, we could say that 10% are customers who bought a second time or more. But it may turn out that one customer made 10 purchases and everyone else bought only once. So something about our product isn't working for retention. Or with 50 customers for 100 purchases we say: "Great, as many as 50% of customers come back," and then it turns out we have one wholesaler placing those 50 orders while the rest never come back at all. That's why we simply have to check how many customers come back to us a second time.

This metric also shows us whether we have a great product, because great products, if they have any built-in retention (let's not talk about furniture), will show that we have returning customers. If the metric is growing, we can see the great products. It also works well broken down by product. AI can show us our top 10 best products and the worst ones, the ones customers don't come back for, and we wonder what's wrong with them. Why don't they come back for these, but do come back for those? You can also think about categories, because products in one category should usually have a similar rate. And it may turn out that a few work so well you won't believe it, and we wonder why. If we're a manufacturer, we think about what to do so the rest have that too. And why don't they come back for the others? Maybe switch off their ads or stop selling them altogether, because something is wrong with them.

The next metric is frequency, meaning how many times a year customers buy from us. Once they're buying, how often? At the company where we sold cardboard boxes, online stores bought four times a year on average. So we can plan that if we acquire a customer, they'll buy from us four times on average, and then we think: "Okay, how much can we pay? How do we win them over on the first transaction? What can we do for them?" Because we know that once we've acquired them, they'll visit us, say, four times a year. And the last, most important metric is LTV, lifetime value, meaning how much we'll earn from the customer over their whole lifecycle. How much we make on them once we've acquired them.

How do you calculate LTV and how much can you spend to acquire a customer?

Maciek: Calculated on margin?

Jędrzej: You can calculate LTV on profit only, or on margin. Everyone can calculate it their own way, because for some, if they have a very stable margin, revenue is enough. They know in their heads: right, we have 25%, this is what we paid. Then there's CAC, the customer acquisition cost, and you can divide these by each other in many ways. But if I know my LTV, meaning how much profit I have per customer, most people can roughly work out that value in their heads. If I see that on average I get PLN 1,000 from a customer and a margin of, I don't know, 20% that I've calculated somewhere, then I know I have PLN 200, and anything under PLN 200 to acquire them more or less makes sense. I think about it over 12 months: how much do I earn over 12 months? Okay, I can spend, say, PLN 150. And anyone who only looks at first transactions says: no way would I pay that much. Because we think about how much the customer earns for us over a period that matters to us, we can think about how much to spend on acquiring them.

These four metrics are best measured in cohorts: which ad customers came from, which product, which categories, so you can separate things a bit. Averaging across the whole store can be quite dangerous, especially in a big store.

Maciek: I also think segmentation is key here, because averaging…

Jędrzej: You know, in a category like our promotional products we try to split it in many ways and it doesn't work all that well. We have segments of bigger and smaller customers and we see their LTV is different, but the time after which they come back is often similar.

Maciek: What I mean is that averaging across all customers in the store is a bit like measuring conversion for the whole store instead of broken down by channel.

Jędrzej: Yes. But in the first phase, if you haven't measured anything and, as we established, you don't have anyone in charge of retention, check the overall picture first. Then tell everyone: do something nice with delivery, let's improve the look, let's drop the worst products nobody comes back for. Is it improving? Good, I've done the most obvious things, now I have to go deeper. In the same way, if you knew the conversion rate for the whole store and were improving its look, you'd be improving it across the whole store and on every channel. Only then can you tighten the screws on the details.

Where should a company start working on retention?

Maciek: Agreed.

Tomek: That's right. We're slowly wrapping up. I think we've gone through this very systematically. As you said, your framework is also there to approach this systematically, and in analytics from the general to the specific. You come to a meeting at a company that's hiring you for a retention task because they have a problem with it. Where do you usually start? Or what would you recommend such a company do first, so that it opens…

Maciek: A diagnosis.

Tomek: Yes. So that it opens up the topic and really tackles the task ahead of it.

Jędrzej: Partly with those metrics, just more developed. But my favorite and fastest advice is: take your data, strip out the customers' personal data, put it into an LLM and ask how my retention is doing. Today Claude or ChatGPT can really spit out the first hypotheses. Even if you don't tell it much, it suddenly shows you a few charts. It does it really well: okay, these are the products people come back for, and these they don't. Here are the customers, time to second purchase is 187 days, but there's a point at around 100 days: if customers haven't come back by then, they come back spread over two years, so that's your point. It does this really nicely.

I usually try to establish at the start where we are. If we come to a furniture company that has a 20% repeat rate and says: we're going to work on retention, I say: you know, that might not be the best direction, because you're already well above the industry benchmark. Let's look at why and on what. So at the start we check where the company is and where it leaks the most. Do we see that we have a great product and customers still don't come back? Or has nobody ever checked the product? We advertise the bestsellers, but nobody wants to come back for them, while somewhere in our range we have products people do come back for. Or maybe there's nothing to come back for at all, and we go to the product or purchasing team and say: listen, we have to build an ecosystem around this product. Find add-ons that could be sold around it.

It can also be that the product is seasonal or part of an ecosystem, but we don't have those add-ons; we haven't figured out what should be there. Or it's a fashion product, which often has seasonality, but none of our buyers buy the trending colors. We always have blue jeans. I don't know much about fashion, but I know there's always some color of the year, and it would be good for the buyers to have a goal of adding trending products to the range every year too. We have a certain standard, and this is an action that gives someone a reason to come back at all. Or we see that products sell, customers come back, but they drop off somewhere at the experience stage.

So at the start, simply a self-diagnosis: where am I. The survey we ran is still up on the website; you can take it and it shows where you drop off. 60% already drop off at the product: either they don't know which products drive retention, or they never check, or they know but do nothing about it. It would be good to go through that, and only then say: "Okay, we want to be a retention company." And if we want to be a retention company, then all departments, as we said: purchasing, customer service, accounting, logistics, have to have it in their goals. It can't be that we go to marketing and say: "Now build us retention marketing." Marketing will send out newsletters, and will even try to build that community (we don't agree on the word, but we do on the logic itself, that people buy because they like a brand or are attached to it).

How will retention change when an AI agent does the shopping?

Maciek: I have one last thought, because we started a bit with AI and we've come back to it again. Looking ahead a little: you mentioned at the start that agentic shopping isn't happening yet. With our clients we already see agentic sales in the statistics, and we can expect there to be more of it over time. What do you think retention, loyalty and newsletters will look like when it's not a person buying, but an agent?

Jędrzej: Then manufacturers have a chance. If we decide we like the Adidas brand, your agent will buy you Adidas shoes, and then the manufacturer won't much care which store it's from. So from a D2C point of view, it's the manufacturer who will look after retention. And if you're a multi-brand store, you have to carry retention brands, because they're the ones building community. You can build something like Sklep Biegacza around your store. A store like that can create content, and people like buying from it because they feel sure. But if it doesn't have those brands, there won't be retention.

The second thing. Many of us will get used to it and say: "Don't compete on price for me, I want to buy from the store that gives the most certainty." Because if we're buying within some price range, we won't say: "Buy it for me from the cheapest store." If people bought like that, there'd be a war like the one on Allegro back in the day, where people always undercut by one grosz. Interestingly, there's an anecdote that it was usually the second seller who sold best, because people were afraid to buy from the first one. If one had 87 and the second had 88, the one at 88 sold better, because everyone was afraid the first was fake and selling a brick, since they'd cut the price even further. Often it was the second-cheapest who had the most sales, because people said: "No, the cheapest one might be a scammer, the second one probably isn't." That's how it was on Allegro for a while, because people were afraid, but they still killed each other on price. So if we don't tell the agent "buy the cheapest," it will take into account the number of reviews, what people say about you, how your store is built, that you have 100,000 customers and 20 years of experience.

If I had to advise stores now on how to build their communication, most people say: "Right, talk a lot about the product, give the context." Not true. Because what difference does it make if I'm talking to an LLM? I say: "Okay, I want to buy running shoes. I'm 1.90 meters tall, I weigh 90 kilos, I mostly run in the forest." Do you care who told you that these Adizero are the best? No. It could have taken that from the manufacturer's site or from a foreign site. Nice that it found the match. Contextually, we chose together with it that this brand and this shoe are the best.

Maciek: Exactly.

Jędrzej: Then comes the question: where should I buy Adizero? Pick the best store for me. And here the agent makes a second decision. It's not that this store talks nicely about shoes and knows a lot, so I'll buy from it. What will count then? The number of reviews or rankings, but not rankings of the shoes, rankings of stores compared with others. A review saying they give the most time for returns. The agent will say: "This store is good for you because you'll have 200 days to return." Really, the question isn't what to buy, but where to buy. Here I have to convince the LLM to pick our store. Manufacturers should take care of content quality, so that the LLM wants to pick their shoes as the best running shoes for me. Let's take care of that and strengthen the brand, but for us the second question will matter: "Right, I know I want these. So where do I buy them?" And suddenly a list pops up.

Try it with any store: start asking about cosmetics, then ask where to buy them, and read why it answers the way it does. This one has lots of reviews, that one is praised for something. It builds the ranking itself and shows: this one is the biggest, that one has something else. But it doesn't say this one has the widest choice, because you want to buy one pair and you've already picked a specific one. It shows specific points. And as multi-brand stores we should definitely already be competing for that query: where to buy this specific shoe.

Maciek: Community may turn out to be an important element.

Jędrzej: Yes. Or some people will say: "Okay, buy it for me at Sklep Biegacza." Because I know it, I like it, I trust it. And the agent will ask: "Should I buy at Sklep Biegacza again, or should I do some research?" "Buy there again." When the customer buys again, the agent will ask: "Should I buy the same thing again? Shall we buy where we always do, or do you want to do research?" And the customer will answer: "Where we always do, because I know it's a safe bet." There's a chance the agent won't buy from a random store, but will ask you where you'd like to buy. So if the customer remembers your store because you have branding, the agent will remember it too. That's where branding will matter.

Maciek: That's our theory too, that the brand will keep getting stronger. And I think we've found the word we were looking for: it's more about attachment to the brand.

Jędrzej: Could be. Brand sits a bit at every level, because if you ask the people who work on brand, they'll say it's the experience. The box is brand too, the product is brand, because you brand a bit of everything. But understanding and remembering the brand could indeed be that word.

Maciek: In the end, it's about building associations.

Where to find Jędrzej, and why is retention getting more attention?

Tomek: Great. Listen, to finish, Jędrzej, a standard question, but a very technical one. Where's the best place to find you if someone is interested in your services or your knowledge? We know you have your own site where the framework is also available, and you invite people to get in touch there. But apart from that, do you attend events in person, are you available on LinkedIn?

Jędrzej: I try to attend e-commerce events, and now retention events too. If you want to meet in person, the soonest option, and you still have a chance, because I hope this episode comes out before it: see you at the Fight Club Retencja event. It's on October 22 in Warsaw, so you can meet me there in person and, interestingly, talk about retention with lots of people, not just me, because the whole event is dedicated to retention. If someone wants to contact me directly, LinkedIn is best, where I have a fairly large community. And Fundamenty Retencji. The framework is available there to download for free, and without giving your email. I set it up so you click and you see it. If you're interested, you can sign up for the newsletter, but it isn't mandatory. I want to share it and give value, not build a mailing list. I have a full-time job, so I don't have that many clients and I'm not fighting for them; I'd rather spread the word about how this works.

It isn't any magic knowledge either, because whoever you ask, everyone knows the product matters. It's more a kind of structure that helps someone know a bit better how to build it. I gathered it from around the world, and many people were already working this way without knowing it was a framework, because it was logical to them. We at least make it a little easier to build.

Maciek: We definitely agree on the direction, that this topic will…

Jędrzej: It already is. Take this event: you can see there's an event about retention, and other events have retention stages. First, you can no longer grow attribution, because, as we said, the pie has stopped growing. The number of Poles shopping online has pretty much hit the ceiling. On top of that, we're a shrinking society, so there will be even fewer of them. We won't get more out of attribution, because the pie is what it is. Now we have to think about how to get a bit more out of the customer, and how not to fight for them, but to make them come back to us. You can see it. I've been working on this for six months and I already see so many different ideas: from sensible loyalty programs to tools, analytics and the rest.

So I think it's a hot topic, but to make it interesting: you can't live without attribution. There's always churn, and you have to remember that, so I still have to acquire customers and build the brand. I even have to win back the same customers, because if someone uses Google search, I still have to fight for them. It's just that if they see our brand, they may be more willing to buy from us. It's not a golden method after which we switch off Ads or any kind of attribution. The departments should simply work with the aim of making customers the ones who come back most often.

Tomek: Exactly, I fully agree. Jędrzej, thank you very much for what I think was a great episode. Really a lot of valuable knowledge, and it was a real pleasure talking with you.

Jędrzej: I didn't let you get a word in, dang, and you said…

Tomek: We're not here to take up the airtime, so we handed it to you. Thanks again so much for coming to our studio, and see you soon in person in Warsaw.

Jędrzej: See you. Bye.

Maciek: Thanks, lovely.