AI conversion optimisation
AI conversion optimisation: the machine builds, I decide.
I create changes, test variants and analyse with AI. That gets your changes live fast, and your own team does not have to build a thing. What AI does not know is which change makes your shop sell more. I decide that, from 20 years of conversion optimisation. You approve what goes live.
I'll show you where your shop loses buyers.
In short
AI conversion optimisation means AI takes over the work that used to cost time and developer budget: building variants, analysing data, monitoring tests. What gets tested, why it should bring more buyers and which result counts is decided by a person. At JDKRUEGER&CO that person is Jörg Dennis Krüger, Der Conversion Hacker®. Wherever traffic allows, an A/B test proves the effect, with a Bayesian analysis.
What AI has changed
Three things are different today. All three make conversion optimisation faster. None of them makes it better on its own.
Building is no longer the bottleneck
A test variant used to sit in a developer ticket and wait. Today I create it with AI: a new element on the product page, a rebuilt cart, the arguments in a different order. That gets your changes live fast, and your own team does not have to build. For you this means more of your budget goes into deciding what gets built, and less into the building itself.
Iterate instead of waiting
When a variant is quick to build, an idea does not have to be perfect before it goes into a test. Once the first version delivers data, I create the second. Weak ideas drop out sooner, good ones get better faster. Conversion optimisation becomes what it always should have been: a cycle. After the test is before the test.
Analysis straight on your data
AI queries your analytics, your testing tool and your shop system directly. Questions that once meant someone pulling exports and building spreadsheets, it answers on the data itself: where do first-time mobile visitors drop off? Which category loses buyers in the cart? I check every answer before it turns into a decision. A plausible explanation is not yet a correct one.
What AI cannot do
Ask AI and it hands you a hundred ideas for your shop. That is exactly the problem. Ideas were never scarce. What is scarce is the judgement of which of them makes you money, and in what order.
Which idea comes first
I rank every idea by the revenue at stake in your shop, divided by the effort. Your own ideas sit in the same list, including the ones you generated with AI yourself. They are judged by the same yardstick as mine.
Why a change sells
A change sells when it takes away a doubt or gives a buyer a reason, at exactly this point. Seeing that takes experience. AI knows patterns from the average of the internet. Your shop is not average.
The fine detail
Whether the shipping note sits above or below the button, whether a sentence is three words too long: details like these decide whether a test wins. I work out every change down to these details before I create it. That is most of the work, and it stays with me.
Steering the AI
AI does what it is told. Including the wrong thing. I set what it should build, what it is measured against and what is off limits. Product title, price and buy button stay as they are unless there is a clear reason. Your buyers should always know what they are buying.
Judging the result
An A/B test measures whether a change works. It never measures why. That interpretation stays my job, and I call it interpretation, not measurement.
My AI tools
I use AI where it saves time: building variants, analysing data, monitoring tests. What gets built and what goes live is always decided by a person.
Dropoff Analysis
Before our call I check your shop and mark where buyers drop off and what that costs you.
Direct data access
AI queries analytics, testing tool and shop system directly. Metrics, not customer lists: your buyers' personal data stays closed.
Idea list ranked in euros
Every idea, from you, your team or me, lands in one list. Ranked by the revenue at stake, per unit of effort.
Test ideas from your numbers
AI keeps suggesting new test ideas from your data. None reaches your list without my approval.
Variant building
I create changes and test variants with AI. This is where AI has sped the work up the most.
Approval before go-live
I check every variant before it goes live. You see every change beforehand and approve it.
Controlled test operation
Tests start, run and stop under control. If a variant gets out of hand, it is stopped.
Results database
Every change and every test lands in a database with its result. A change only counts as done once it is measured.
What you see of all this: a client area with your idea list, the status of every change and what it measured.
How a sprint with AI works
I work in sprints. Each sprint takes the top-rated changes from your list. Who does what is clearly split.
- 1
Dropoff Analysis · Person, AI prepares
I show you on your own shop where buyers drop off. The most expensive point first.
- 2
Idea list · Person
Your ideas and mine go into one list. I rate each one: what it is, why it should sell, which metric it has to move and what is at stake.
- 3
Working it out · Person
For every change in the sprint I work out exactly what it looks like and what it leaves out. This is where the value comes from.
- 4
Build and approval · AI and person
I work with AI. Before go-live I check every change, and you approve it.
- 5
Live or test · You
You decide per change: switch it live directly or prove it with an A/B test. A test runs at least two full weeks and one full business cycle, with a Bayesian analysis.
- 6
Measure and move on · Person, AI calculates
Every result goes into the results database. In the monthly review you see what went live and what it measured. Then the next sprint begins.
What does a Bayesian analysis mean?
A Bayesian analysis means the result tells you how likely it is that the new variant is better than the old one, and by how much. A classic analysis mainly checks whether a difference is significant. For a shop the first question is the more useful one, because you see how certain a win is and how big it is likely to be. The minimum runtime of at least two full weeks and one full business cycle makes sure the result covers every weekday and every payday, instead of measuring one week that happened to be strong or weak.
What comes out of it
EUR 245.3M¹
additional revenue for my clients
+35.3%
average conversion uplift
20 years
of conversion optimisation
German Stevie Awards
reifen.com Automotive & Mobility
+12.5% conversion rate on the product page. Around +11% in the cart. Both measured by A/B test against the control.
Cotton Ballon Kids & Family, Shopify
+120% conversion rate in two weeks. A single result.
„Schon beim ersten Entwurf sieht man, dass hier ein Profi am Werk ist. Ich bin begeistert.“
Translation: "From the very first draft you can see a professional at work. I am thrilled."
„Ich glaube, er ist der richtige Mann, wenn es darum geht, Online-Umsätze zu pushen.“
Translation: "I believe he is the right man when it comes to pushing online sales."
Frequently asked questions about AI conversion optimisation
What is AI conversion optimisation?
AI conversion optimisation is conversion optimisation in which AI does the execution: analysing data, building test variants, monitoring tests. The goal stays the same: more buyers from the traffic your shop already pays for. At JDKRUEGER&CO, Jörg Dennis Krüger decides which changes get tested and why.
Can AI replace a conversion expert?
No. AI generates ideas and builds variants. It knows neither your buyers nor your margin, and it suggests what sounds plausible on average. Which idea makes money in your shop first, why it should work and how to read a result remains a matter of judgement. My judgement comes from 20 years of conversion optimisation.
Can't I build A/B tests myself with ChatGPT?
Ideas and drafts, yes. With AI you can build a draft in an afternoon today. The bottleneck comes after that: testing the right idea first, getting it live cleanly, measuring correctly and reading the result correctly. I am happy to add your own AI ideas to your list. They are ranked by the same yardstick as mine.
How is AI changing A/B testing?
AI builds test variants faster, suggests test ideas from your data and monitors running tests. It does not change the statistics: a test needs enough buyers and enough time. At JDKRUEGER&CO every test runs at least two full weeks and one full business cycle, with a Bayesian analysis. The result tells you how likely it is that the new variant is better than the old one, and by how much.
What happens to my shop's data?
I work with metrics from analytics, testing tool and shop system. It does not open your buyers' customer lists, exports or customer accounts. The analysis does not need them.
Does it fit my shop?
No shop system is ruled out: I work with all of them. Whether an A/B test makes sense depends on your traffic. Where there is enough, a test proves the effect. Where there is not, you switch a change live directly and measure afterwards. You decide that per change. It is not a fit if all you want is ideas. Your AI gives you those itself.
What is a Dropoff Analysis?
In the Dropoff Analysis I show you on your own shop where buyers drop off and what that costs you. The most expensive point first. I prepare the analysis beforehand.
Tool or team? Comparisons with A/B testing tools and agencies
Let's start with your shop.
In the Dropoff Analysis I show you where your shop loses buyers. The most expensive point first.
¹ Cumulative additional revenue across all client projects since 2012, as of 23 Oct 2026.