Why doesn't the shopping cart method work on bol anymore, and what are the consequences for Rylee?
Lars HurkmansCo-founder1 July 2026Temps de lecture 8 minutesThe shopping cart method was the standard for sales estimates on bol for years, and the basis on which Rylee presented their data. In November 2024, bol blocked this method with measures against bot traffic. Since then, Rylee's historical weekly data disappeared, a model arrived that by their own admission is less accurate, and the display shifted to ranges. A reconstruction.
Why does the method behind sales estimates matter?
The method behind sales estimates determines how accurate they are, how far back the historical data goes, and how stable the whole thing is over time. If you use a product research tool for bol, you base your decisions on those estimates: how much does this product sell per month, and is there enough volume to be profitable? Those numbers form the basis on which you decide whether to source a product, how much you order and what you expect to earn. But those estimates come from somewhere.
How does the shopping cart method work on bol?
The shopping cart method was the most commonly used technique for estimating sales on bol for years. The principle: a tool adds a product to a shopping cart on bol, reads out the available stock, waits a period, and measures again. The difference is counted as sales. A product with a stock of 200 on Monday and 185 on Tuesday supposedly sold 15 units.
The method was popular because it's directly linked to what happens on the platform. Practically every product research tool on the market used a variant of it.
But the method has structural limitations. It measures stock changes, not actual sales. If a seller manually adjusts their stock, that looks like a sale. bol showed stock up to a maximum of 500 units, so products with higher stock fell outside the picture. And the biggest bottleneck is scalability: for every product, the stock needs to be measured multiple times a day. That forces tools to choose between tracking more products with a lower measurement frequency, or fewer products with higher accuracy. That's also why product trackers typically work with a limited set of products you select yourself as a user.
We described those technical limitations in detail earlier in MarktMentor vs product trackers.
These limitations had been known for years. Yet the shopping cart method remained the standard in the market, simply because it worked. Until November 2024.
What changed in November 2024?
In November 2024, bol took measures against bot traffic on the platform. bol was troubled by the volume of automated requests and wanted to limit it. The result was that the shopping cart method was shut down on a large scale.
The shopping cart method is a form of scraping: automatically reading out information from a website. To measure stock information at scale, tools need to query bol's platform with automated requests.
For tools running on a different method, nothing changed. But for tools that depended on the shopping cart method, this had direct consequences. If the technique with which you collect sales estimates disappears, that hits the core of your product.
And that dependency was no side note for Rylee. It was their selling point.
How did Rylee position the shopping cart method?
Rylee was one of the best-known tools using the shopping cart method, and they were open about it. On their comparison page about MarktMentor, archived via the Wayback Machine (July 2025), they explicitly presented the shopping cart method as a core advantage.
They wrote that their data was based on "the proven shopping cart method with actual sales". Alternative methods, including MarktMentor's, were labelled "experimental" with "substantial margins of error". The message was clear: the shopping cart method was the reliable standard, and anything that deviated from it was inferior.
That positioning wasn't subtle. It was the core of their comparison argument. It was the reason they gave for why a seller should choose Rylee instead of an alternative running on a different method.
That makes what happened next all the more relevant.
What has changed at Rylee since then?
After November 2024, a series of visible changes followed at Rylee. We've listed the publicly observable developments, based on what Rylee itself communicated via community posts and their own platform.
November 2024: bol blocks the shopping cart method. bol takes measures against automated requests. The method Rylee had run on for years is hit on a large scale.
December 2024: historical weekly data disappears. Historical sales estimates per product are no longer available on a weekly basis at Rylee. Data sellers previously used to track sales performance over longer periods is no longer shown the same way.
August 2025: announcement of new model. Rylee's owner announces on a community platform that a new machine learning model has been developed for calculating sales data in the product database. He states that the model has been run for over 3 million products, but that there are still outliers he wants to work through before everything goes live.
January 2026: confirmation of lower accuracy. In another community post, Rylee's owner states that the new model is "less accurate than the shopping cart method", but that for 80% of products the calculations fall within 20% of reality, "comparable to MarktMentor". He mentions that major updates are being rolled out to improve the model.
March 2026: from specific estimates to ranges. Rylee's Chrome extension and product database no longer show specific sales numbers, but ranges. Where a concrete number used to appear, sellers now see a range within which the actual sales supposedly fall. On their comparison page, Rylee writes that they deliberately choose ranges because a model "never has a 100% grip on the truth".
What does this timeline show?
If you lay these developments side by side, a pattern emerges.
The method Rylee presented as the most reliable in the market for years has fallen away. The model that replaced it has been described by Rylee itself as less accurate. The display of sales data changed multiple times in the intervening period. And the switch from specific estimates to ranges was made in the same period the shopping cart method was no longer available.
There's more. The terms Rylee now uses to describe their new approach, machine learning, supervised learning, regression models, are the same terms with which they previously labelled MarktMentor's method as "experimental". The approach that was portrayed as less reliable back then is now the direction Rylee itself is taking. The difference is that MarktMentor has been using and developing that method for years, while Rylee only started with it since late 2024.
That in itself isn't unreasonable. Every tool has to adapt when circumstances change. But it stands in sharp contrast with how firmly that same method was rejected earlier on their own comparison page.
What does this mean concretely for you as a seller?
This isn't an abstract technical story. The method behind sales estimates directly affects the decisions you make as a seller.
Specific estimates versus ranges
With a specific estimate of 87 sales a month, you can calculate. At a purchase price of 12 euros, a sale price of 29 euros and a bol commission of 15%, you keep around 8 euros per unit. At 87 sales that's 696 euros gross profit a month, before advertising and shipping costs. You can assess whether it's worth sourcing this product.
With a range of 50 to 150 sales, that same calculation looks different. At 50 sales you keep 400 euros, at 150 sales 1,200 euros. The lower and upper bounds lead to opposite conclusions about the same product choice. At 50 sales it's probably not profitable after advertising costs. At 150 sales it's one of your best products. That uncertainty makes it hard to make concrete decisions about sourcing, stock and advertising budget.
Comparability over time
You recognise seasonal patterns by comparing the same data across multiple years. You see growth trends because a product structurally sells more than a year earlier. That only works if the data has been collected consistently.
If the method changes partway through, a difference in the numbers doesn't tell you whether it's the market changing or the measurement method. A product that showed 80 sales a month in 2024 and now shows a range of 40 to 120 hasn't necessarily started selling worse. It could also be that the old figure was calculated with a different method than the current one. As a seller you can't make that distinction, but it does influence your decision.
How you factor historical data into your product choices is worked out further in historical data and product choice.
How does MarktMentor handle this?
MarktMentor moved away from the shopping cart method at an early stage. Not because we couldn't do it, but because the structural limitations in scalability and accuracy weren't solvable within that model. We wrote about that at the time in the switch from shopping cart method to algorithm.
We've developed our own data collection method that works independently of the stock information bol shows. That method makes it possible to track millions of products without the trade-off between coverage and accuracy the shopping cart method forces.
bol's measures in November 2024 had no impact on our data. There have been no interruptions, no method changes, no changes in the display. Today's data is built on the same basis as data from years ago.
MarktMentor shows specific sales estimates per product, supplemented with trend indicators. The database contains over 5 million relevant products, growing daily, with multiple years of uninterrupted historical data. That consistency isn't a coincidence. It's the result of a method choice made years ago, precisely because we foresaw the vulnerability of the shopping cart method.
Want to see the differences side by side, point by point? The full comparison between MarktMentor and Rylee is worked out separately.
What can you watch for when choosing a product research tool?
When assessing a product research tool, there are two points relevant within the scope of this article.
Specific estimates or ranges. Specific estimates give you the basis to make concrete calculations around sourcing, pricing and advertising budget. Ranges give an indication, but with a wide range, the lower and upper bounds can lead to opposite conclusions about the same product choice.
Method continuity. Historical data is valuable, but only if that data is based on a consistent method. If the method changed recently, comparing with the past is less reliable. It's worth asking how long a tool has used its current method, and whether there have been interruptions or changes since.
Want all the criteria laid out? We've worked out the step-by-step plan for choosing a product research tool separately.