Every Monday morning starts the same way for Mehmet: a strong coffee and the glow of an e-commerce dashboard.
At first glance the numbers hypnotize. The campaigns ran all weekend, site traffic is up 20%. Every "new order" notification on his phone fires off a small victory in his brain. The mood in the office is good: marketing celebrates the lead count, the warehouse is buried in packing.
Then the month ends, he sits down with finance, and the sweet dream turns into a cold fact. Revenue is up but profitability hasn't moved. Ad costs have climbed so high that most of the profit has been paid straight back to Google and Meta.
What Mehmet can't see — or rather, forgot to look at — is this: his business has leaky bucket syndrome. He hauls fresh water into the bucket with enormous effort (new customers) while never noticing what seeps out of the holes underneath (the customers leaving).
I wrote this piece in December 2025 under the title "2026 trends", and I made exactly one claim: the winner of 2026 wouldn't be whoever bought the most traffic, but whoever squeezed the most revenue out of it. It's August now. Time to check the claim — first the bucket, then the report card.
Leaky bucket syndrome: why do we always want "more new customers"?
It's no accident that e-commerce managers fall into this trap; it's a trick of human psychology. In behavioural science we call it hedonic adaptation. Our brains release a burst of dopamine for novel stimuli and rewards. A first-time visitor's first order thrills the business owner. The quiet, regular purchase of a loyal customer who already knows you produces nothing of the sort.
So businesses slip instinctively into hunter mode: forever chasing the next kill while leaving the farm they already own — the existing customer base — to the drought. Yet the brutal math of e-commerce says this:
The cost of bringing a customer through the front door rises every day; the price of not holding the one walking out the back is bankruptcy.
If you feel your ad budget isn't paying for itself, the problem may not be your traffic. Traffic carries water to the bucket; it doesn't plug the hole.
Mid-year reality check: what we said in December, what happened by August
Most trend pieces get written in January and never opened again. I opened this one eight months later, because a prediction is worth what it's worth on the day it's tested, not the day it's written. Here's what we saw in the field:
- Held — acquisition didn't get cheaper. Across the accounts we ran this year, the cost of a new customer never came down; the same budget reached the same person for more money. That's where the claim was confirmed: growth is a retention job, not a spending job.
- Accelerated — budget allocation moved to the algorithm. Performance Max and Advantage+ style automations now split the budget themselves. The trap is subtle: teach the algorithm "number of conversions" instead of "revenue" and it will hand you a mountain of cheap, unprofitable orders. It runs toward the wrong target at flawless speed.
- Arrived faster than expected — discovery now starts in AI engines. Customers hear about your brand inside an answer rather than a search result, and land on your site with the decision already made. That made the first order easier; it did nothing for the second. Which is exactly why the bucket metaphor got more relevant in 2026, not less.
- Lagged — first-party data. Everyone talked about it, few built it. Yet this is retention's fuel: if you're not collecting your own data, you can't know who you're keeping and who you're losing. That gap will widen in the second half of the year.
All four come down to measurement. We told the story of getting the order right in the SOYLU AVM case: we rebuilt pixel and conversion tracking from scratch first, and only then opened the campaign. Day six recorded $1.5M in revenue. Reverse the order and the same budget wouldn't have produced the same result — nobody would have known where the money went.
The math: why retention beats acquisition
If the metaphor lands, let's go into the kitchen. In marketing, emotion triggers the decision and data confirms it. Most e-commerce businesses spend roughly 80% of the budget finding new customers and 20% keeping the ones they have. The statistics of commerce whisper the opposite.
Have you met the goose that lays the golden eggs?
The 80/20 rule Vilfredo Pareto proposed in the 19th century applies mercilessly in e-commerce: most of your future revenue comes from a small slice of your existing customer base. If you spend all your energy chasing that vast new audience, you're fighting for the smaller piece of the revenue.
The real treasure sits quietly in your database. That slice knows your brand, has already saved its card details and is ready to tell a friend about your product. Reaching them costs one email; reaching a new customer costs an auction.
Segmentation engineering: why isn't every customer equal?
We've got the math. But who are these loyal customers? Among the thousands in your database, how do you tell the champions from the sleeping beauties? Most businesses treat the list as one audience and send everyone the same message: "10% off everything." Sending a summer sandal promotion to a man who just bought size-45 boots tells him "I don't know you and I don't care." The result: angry fingers on the unsubscribe button.
RFM analysis: taking the customer's X-ray
Categorize your customers with the RFM model, not with your feelings. Three columns turn the mess in your database into a clear strategy:
- Recency: when did they last buy? Yesterday's customer is worth more than last year's.
- Frequency: how often do they buy?
- Monetary: how much have they earned you?
Combine the three and your customers separate into champions (frequent, high-spending), loyalists and the at-risk group (once frequent, now gone). Your approach must differ for each: you thank a champion, you tell the at-risk group you've missed them, and you don't send both the same coupon. We've covered how to build RFM step by step in a separate piece; you'll find it in the articles archive.
The personalisation paradox and the RAS effect
Our brains carry a filter called the reticular activating system (RAS). It makes us notice only what concerns us — our name, a subject we care about. For an e-commerce customer, the sweetest sight is their own name and their own past choices.
- Wrong: "Dear customer, take a look at our new products." The brain files this under advertising and skips it.
- Right: "Hi Ayşe, how's the coffee machine you bought last month? We picked the three filter coffees that suit it best."
The second approach creates cognitive ease. The customer doesn't have to think, because you thought for them and narrowed the options. Personalisation isn't a courtesy; it's a technique for removing load.
Behavioural triggers: not a forgotten cart, a missed opportunity
Use psychological triggers when you build your automations. A standard "you left something in your cart" message is boring. Try the scarcity principle instead: "Ayşe, the items in your cart are waiting, but stock is running low." The message creates a mild fear of loss and raises the odds of action. One caveat: don't write that sentence if stock isn't actually running low — a bluff caught once kills loyalty itself.
Case in point: Amazon's "customers who bought this also bought"
By a widely cited figure, Amazon owes around 35% of its revenue to its recommendation engine. It never shows you a random product: it reads the pages you browsed and the items you clicked, then fires the social proof principle with "customers who viewed this also bought". Even Netflix's cover art is personal — the same film gets a romantic frame for one viewer and an explosion for another. How something is presented matters as much as what is sold.
The peak-end rule: the experience ends when the box opens
The human brain remembers an experience not by its length or its average, but by its most intense moment and how it ended.
In e-commerce, the "end" is the moment the customer opens the door, takes the parcel and opens the box. If that moment is ordinary, the whole purchase gets filed as ordinary. If it's captivating, the customer bonds with your brand. The box is the cheapest and most neglected line in the marketing budget.
The plain brown box mistake
The great shortcoming of buying online is touch. The customer can't hold the product or smell it. The first moment it reaches their hands is the only chance to feed that hunger. A plain taped-up brown box says "we're done here, I got the money". A branded box, the rustle of tissue paper inside it, a faint scent — those say "you matter". The stronger the tactile experience, the higher the perceived value.
Reciprocity: the power of a small sweet
Robert Cialdini's famous reciprocity principle is our sense of owing something to whoever did us a favour. An unordered little gift falling out of the box — a handful of jelly beans, a sticker, a tester, a handwritten thank-you note — creates a positive shock. That gesture leaves a sense of debt, and the customer settles it in two ways: by choosing the brand again, and by photographing the box and sharing it. The second one is free advertising.
Expectation management: making the process transparent
The wait before the box arrives is full of anxiety. "Where's my parcel?", "Have I been scammed?" run in the background. Instead of a bare "shipped", a human notification — "Ayşe, we packed your order with care and it's on its way" — turns waiting from torture into an exciting countdown. Same information, different emotion.
Apple is the master of this. Opening an iPhone box, did you ever notice the lid doesn't just drop — it resists for a couple of seconds against the vacuum, then slides down slowly? That's not a coincidence, it's engineering. Those seconds of delay push anticipation to its peak. Brands like Sephora and L'Occitane do the same with testers dropped into every order, however small, so the customer always feels they came out ahead.
Fixing the bucket: what order do you follow in 90 days?
The mid-year report card in one line: the trends changed, the claim didn't. Getting the order right is still the cheapest growth lever there is. The order we recommend has four steps:
- Verify measurement. If you can't see which channel produced which order, the other three steps are guesswork.
- Tie the ad objective to revenue. Teach the automation income and margin, not conversion counts; the algorithm produces whatever you reward.
- Split customers with RFM and write a different sentence for each segment. One campaign message produces one result.
- Design the second order: the box, the thank-you note, the delivery notification and the post-purchase flow. The cheapest growth line hides here.
When the bucket holds, acquisition pays back many times over. In the OdorGo case we reached ₺10M in revenue in eight months in a category that effectively didn't exist in Türkiye; the commercials, the CRO-led e-commerce site, email, organic search and marketplace storefronts all ran inside one measurement frame. The target was revenue, not traffic, and the measurement was built accordingly.
I'm not telling you to stop acquiring customers. In the FYR launch we reached $100K in three months with return on ad spend holding above 20× — acquisition works. But where it works, the bucket holds. Hauling water without plugging the hole is donating your budget to Google's and Meta's auctions.
When your product reaches the customer's hands, is the feeling "finally, that's done" or "I feel special, I should share this"? People forget the product they bought; they don't forget how they felt buying it and opening it. If you'd like to find the holes in your bucket together, look at our CRO service and the performance marketing side — where those two meet is exactly this.