A good salesperson asks two or three questions and pulls out the right product. Online, your visitor is left alone in front of the whole catalog. Sense asks the questions in their place, reads their photo and what they write, and returns the products that suit them. For every visitor, with no volume limit.
Sense runs today on eight Shopify stores, across four different trades and at wildly different volumes.
Every answer feeds a scoring scheme you set yourself. You decide which profile comes out with which products, and you can change it without going through us.
The next questions depend on the previous answers. The visitor does not walk through ten screens that do not concern them, and they reach the end more often.
The visitor sends a photo. It is analysed, then crossed with their answers: what they declare and what can be seen do not always say the same thing.
Every order coming out of a diagnostic carries its identifier, on the line item or on the cart. You read real revenue, not an estimate rebuilt after the fact.
The first level reads the answers only: your scoring scheme turns them into a profile, always the same way. The second adds the photo and the free text, each with its own weight, and settles it when they contradict each other. You choose according to your catalog. Either way, the visitor sees the same page.
The three signals do not carry the same weight. Image analysis counts in proportion to its confidence: clear, it steers the profile; doubtful, it hands over to the answers.
Below the confidence threshold, the engine changes its tone: it offers a follow-up instead of a verdict. A visitor steered wrong does not come back, and they tell others.
The next questions are generated from what has already been answered. No two visitors take the same questionnaire.
Weights, thresholds and personas are settings, not a black box. Same answers, same result: the calculation can be replayed and broken down line by line.
Three views of the same engine, one per mode. Open the one you care about: the rest of the page does not depend on it.
Every answer pours a weight onto one or more letters. The letters are not summed in bulk: they compete inside their own axis, and only the strongest goes through. The weight also varies with the number of priorities ticked: a single concern counts for more than the third item on a list. The axes, letters and weights above are an example - they are written for your catalog.
A customer disputes their recommendation? You reopen the calculation and show them where it came from. You want a profile to land differently? You change a weight.
Five model calls in total, four of them launched in parallel once the persona is picked. Between them sit steps of pure code: a few milliseconds, and they frame what the model is allowed to propose.
The visitor does not have time to switch tabs. That is the point: the longer the wait, the more of them drop out.
Image analysis always returns two things: what it sees, and how sure it is. It is the second number that decides the weight given to the first. A clear photo steers the profile; a blurry one, badly framed or taken from too far away, sees its weight melt, and the answers take back control.
These three rules are code, not instructions sent to the model: they apply whether it agrees or not.
A failed photo does not invalidate the diagnostic: it makes it more cautious. The visitor walks away with a protocol either way.
A monthly subscription, with no commission on sales. Volume enters the price where it enters our cost: on the plans that call a model. Deterministic mode does not depend on it.
| Deterministicweighted letters, no AI, no model call at all | Per monthvolume does not enter the price |
|---|---|
| Up to 2,000diagnostics / month | 49€ |
| Unlimitedfair use · 25,000 / month | 199€ |
| Tierdiagnostics included | Text AIpersona, adaptive questions, free text | Most complete AI + Vision+ photo analysis |
|---|---|---|
| 200diagnostics / month | 49€ | 69€ |
| 500diagnostics / month | 129€ | 179€ |
| 1,000diagnostics / month | 189€ | 259€ |
| 2,000diagnostics / month | 299€ | 419€ |
| 4,000diagnostics / month | 489€ | 659€ |
| 6,000diagnostics / month | 649€ | 899€ |
| 12,000diagnostics / month | 1,190€ | 1,690€ |
| 25,000diagnostics / month | 2,290€ | 3,190€ |
| Beyond the plan | 0.12€ / diag. | 0.15€ / diag. |
Hosting, updates, questionnaire customisation and attributed-sales tracking are included in every tier. A bespoke result page is free from the 4,000 tier up.
The completed diagnostic. A questionnaire abandoned along the way is never counted. The running total is visible at all times in your dashboard.
At store configuration, based on the modules switched on, not diagnostic by diagnostic. Turning on photo analysis moves you to the Vision column for the current month.
Billed per unit, at the rate of your column. If your overages become regular, the next tier up works out better: we point it out rather than letting it run. Deterministic mode has no overage: above 2,000 diagnostics, you move to the unlimited plan.
The standard template is included. A bespoke result page, designed for your theme, is billed as integration work, quoted per project. It is free from the 4,000 diagnostics tier up.
We plug Sense into your store, adapt the questionnaire and the personas to your catalog, and you read the attributed revenue from the first week on.