Proof-of-value: How Corelex proves its impact on your website
Every AI vendor shows you a case study. None of them show you your numbers.
The problem with “trust us, it works”
Most AI chat tools for webshops come with case studies from other stores, in other markets, with other traffic. That’s not proof – that’s marketing. And it puts you in an uncomfortable position: commit budget to a tool on the strength of somebody else’s results, then hope your store behaves the same way.
That skepticism is backed by research, not just gut feeling: only 35% of software buyers say they actually trust vendor-produced case studies – most assume the story was cherry-picked to sell the tool, not to inform the decision.
It usually doesn’t. Conversion rates vary wildly by niche, price point, and audience – the e-commerce average sits somewhere around 1.65–3%, but that range hides enormous store-to-store differences. A chatbot that lifted conversion 20% for a fashion retailer in another country tells you very little about what it will do for your store, your visitors, this quarter. If you buy on that promise and it doesn’t hold, you’ve spent months and a subscription fee finding out the hard way.
Corelex takes a different approach. Instead of asking you to extrapolate someone else’s success, we measure the actual impact on your store, with your visitors, before you commit to anything. That measurement is called a proof-of-value pilot, and here’s exactly how it works.
What is a proof-of-value pilot?
A proof-of-value (PoV) pilot is a short, time-boxed A/B test that runs directly on your store. During the pilot period, every visitor is automatically and consistently assigned to one of two groups using a deterministic hashing algorithm tied to an anonymous visitor ID:
- Control group: the Corelex widget is hidden. The visitor experiences your store exactly as it is today.
- Treatment group: the Corelex widget is active. The visitor can interact with the AI shopping assistant.
The split is persistent: the same visitor always sees the same experience, so the test isn’t polluted by the same person switching groups between visits. That persistence is what turns this into a genuine controlled experiment rather than a rough before/after comparison.
What gets measured?
For every visitor in both groups, Corelex tracks:
- Exposure: was the visitor shown the widget, or not?
- Add-to-cart events: did the visitor add a product to their cart?
- Source: did the add-to-cart happen directly via the chat widget, or did the visitor navigate to the product page themselves?
Every event is logged automatically and shown in full in your Corelex admin panel — nothing about the result is a black box.
What the results show
At the end of the pilot, the Corelex admin panel shows a side-by-side comparison:
|
Metric |
Control (no widget) |
Treatment (widget active) |
|---|---|---|
|
Visitors exposed |
N |
M |
|
Add-to-cart events |
X |
Y |
|
Conversion rate |
X/N% |
Y/M% |
|
Lift |
– |
+Z% |
The conversion lift is the headline number: how much more likely a visitor is to add to cart when the widget is present. Because both groups are drawn from the same visitor pool, on the same store, in the same period, the widget is the only variable that changes. That’s what separates a real conversational commerce conversion rate measurement from a marketing claim.
For the treatment group, a further breakdown shows how many add-to-cart events came directly from a chat interaction versus organically – the visitor chatted, then browsed to the product page and added it themselves. That distinction answers a genuinely useful question: is the widget helping visitors decide, or just acting as a shortcut button? Both are valuable, but they tell you different things about how your visitors actually use the assistant.
Why this matters
A proof-of-value pilot removes the biggest objection in any software purchase: will this actually work for us? Instead of projecting other stores’ results onto your situation, you get a controlled experiment on your own traffic, and the number that comes out isn’t a promise – it’s a measurement. If you’re trying to build an honest business case for AI chatbot e-commerce ROI, a number generated by your own visitors is worth more than any vendor’s benchmark deck.
It’s also not an unusual ask on your part. More than 60% of B2B buyers now use some form of trial, sandbox, or proof of concept before purchasing software – and that rises to 78% for deals over $10 million. A PoV pilot is simply that same due-diligence step, applied to conversational commerce.
It’s also worth knowing that the measurement is deliberately conservative. Treatment-group organic add-to-carts – where the visitor chatted and then converted from the product page – are only counted if a standard WooCommerce event fires. If it doesn’t, those conversions go uncounted rather than over-reported. In practice, that means the real AI shopping assistant conversion uplift is at least as high as what the panel shows, and often higher.
How to get started
Running a proof-of-value pilot is a simple, guided process:
- You get in touch with the Corelex team to request a PoV pilot for your store
- You get a discounted rate on your Corelex subscription for the agreed pilot period – typically one or two months
- You subscribe at that discounted rate
- Corelex is installed and synced with your WooCommerce store, as usual
- Our team switches on the pilot from the Corelex admin. From that moment, the split test runs fully automatically – no dev work, no manual tagging, nothing for you to maintain – and results build up in real time on your Corelex admin pages for the full duration of the pilot.
If you’re evaluating a WooCommerce AI chatbot and want a number you can actually take to the rest of the business, that’s exactly what a PoV pilot is built for. See Corelex pricing to find the plan that fits your store, or get in touch to set up your pilot.
FAQ
How do I start a proof-of-value pilot?
Building and running a fair A/B test yourself would normally take real technical effort. With Corelex, it doesn’t — you get in touch with our team, agree on a discounted rate for the pilot period (typically one or two months), subscribe, and once Corelex is installed and synced with your store, our team switches the pilot on. From that point, the test runs and reports fully on its own.
Can a visitor end up in both the control and treatment group?
That would invalidate the test, which is exactly why it can’t happen. Visitor assignment uses a deterministic hashing algorithm tied to an anonymous visitor ID, so the same visitor always sees the same group, every visit, for the duration of the pilot.
Can I see the proof-of-value results myself, or do I have to take Corelex’s word for it?
You don’t have to take anyone’s word for it. Every exposure and add-to-cart event is tracked automatically and shown in full, in real time, on your own Corelex admin pages — nothing about the lift calculation is hidden from you.
Does the conversion lift number overstate the results?
Vendors have an incentive to make their own numbers look good, so it’s fair to ask. Corelex does the opposite: organic add-to-carts in the treatment group only count when a standard WooCommerce event fires, so uncertain conversions are left out rather than assumed. The real lift is therefore at least as high as what the panel shows.
How long does a Corelex proof-of-value pilot run?
It’s short by design — a defined start and end date you set yourself, typically just long enough to gather a meaningful sample of visitors at your normal traffic level, not an open-ended trial.
Written by the UCPro BV team (Hasselt, Belgium), creators of Corelex – the AI-powered conversational commerce assistant for WooCommerce.
