TestCart AI is a virtual shelf testing platform for CPG brands, retailers and research agencies. You photograph a real shelf, TestCart rebuilds it as a walkable 3D store, and real shoppers and simulated shoppers shop it so you can see what they notice, consider and choose before you change packaging, pricing or the planogram.
Shoppers do not rate images. They walk an aisle, search, pick up products, compare and put items in a cart, so you measure behavior at the shelf, not stated opinion. Short tasks and a survey after the walk add the why.
A laptop or desktop with a web browser. Nothing to install: shoppers open a link, give consent and start. Phones are not supported for the store walk.
No. Webcam eye tracking is optional per study. When it is on, shoppers can still decline, and attention results are reported only for the visits that used it. Choice, search and cart data are always collected.
No. Gaze is estimated in the browser and only the estimated gaze points are stored. The video is never uploaded.
Shoppers generated by a behavioral model from the audience you define. Once each version of a study has 20 real visits, they can extend it with a directional read. Real shoppers are the evidence. Simulated and real shopper results are always reported separately and never merged.
Packaging and design changes, shelf placement and planograms, assortment changes, price changes between versions, and findability of a product. Each study can compare two or more versions of the shelf.
There is no magic number. The report shows the 95% margin of every figure for the visits you have, and more visits narrow it. A report is stamped decision grade only when its key comparison is settled; until then it says about how many more visits would settle it.
Four ways: share the study link with your own customers or community, ask TestCart to recruit from its tester panel, send it through your survey or panel provider with a redirect we set up with you, or, once each version has 20 real visits, extend the study with simulated shoppers, always reported separately.
A report with choice share per version, time to find, the shopping path, answers to your tasks and survey, and attention heatmaps when eye tracking is on. You can share it with a link or export a PDF.
Every visit starts with an explicit consent step, shoppers can request access to or deletion of their data, and builders only ever see data from their own studies.
Plans are Pilot, Team and Enterprise, priced on a quote. Signed-in users can request an upgrade in the app and are invoiced directly.
What we measure, how the protocol runs, and what we do not claim. Real and simulated shoppers kept separate, every figure with its margin.