Your customers stopguessing their size.
An AI size assistant that knows how every brand cuts. The right size on every product page, fewer returns, and a warm hand off to your team when a human should step in.
15 minutes, no slides. We fit it to your own catalogue on the call.
Built on a real luxury multibrand catalogue: 4,792 products, 122 brands, 14,712 size labels.
Find your size
The fit dossier is already built on a catalogue of 122 brands, and always current
Try it on a real product page.
Click the button on the product page, exactly where your shoppers will. Then write freely: the assistant is simulated here, so it costs nothing to play.
Slim through the thigh, mid rise, gently tapered leg. Eleven ounce Italian denim with a touch of stretch, garment washed for a soft hand. Made in Italy.
Find your size
The real panel, as it opens on the product page. It converts scales, knows the cut, and asks how you like to wear it.
A hundred brands, a hundred fits.
Your shopper does not know how your brands cut. Your size chart does not know your shopper. Somewhere in between, the wrong size gets bought, shipped, and sent back.
- DAY 0Picks the 48, the chart says so. Nobody asked how she likes them.
- DAY 3The parcel arrives. Incotex cuts slim on the thigh, and she likes a little ease: the 48 pulls. Back in the box.
- DAY 9Return request, second courier, the piece comes back out of season.
- DAY 14Refund issued. Two shipments paid, and a shopper who will not risk your store again.
- DAY 0Takes the 50, not the 48 the chart would have given her.
- DAY 3The parcel arrives. It sits exactly where she wanted. It never goes back in the box.
- DAY 14★★★★★Five stars on Google or Trustpilot, and a shopper who comes back.
Your shopper feels like they are talking to your best assistant.
Every reply follows the shape a good stylist uses on the floor: the size first, the why in one line, one useful fact, one question. Never two questions, never a wall of text, never a system word.
It is not a promise made to the model. The code trims what breaks the shape before the shopper sees it: one question per reply, two per conversation, 150 words at most, no markup.
- 1The size, firstSaid before anything else, never buried.
- 2The why, one lineHow this brand cuts, in plain words.
- 3One useful factPieces left or shipping, only when it is true.
- 4One questionOne that helps the advice. Never two. Then it offers a person.
307 fit rules. Every one traced to a source.
Brands drown in size charts: dozens of tables, split by product, in scales nobody agrees on. We wrote 307 fit rules across 55 brands, each recorded with where it comes from: an official size guide, a retailer fit guide, or at least three concordant product pages. Where sources disagree, we write nothing.
Rules that would move a size stay in draft until a person approves them. Your own store rule always wins over ours. If your brand is not in yet, we add it, reviewed before it goes live.
Five minutes to set up. Then it shows you everything.
Set it up once, in about five minutes. From there the console shows exactly what happens on the product page: which pieces raise doubts, how many shoppers open the panel, write, add to bag, or ask for a stylist.
Next on the roadmap: return reasons flow back into the dossier, so the advice gets sharper on its own.
Live in minutes, not a project.
The heavy work is already done in the dossier. Connecting your store and switching the panel on is the fast part.
Connect, read only
A custom Shopify app on read only scopes. It never writes to your store, creates no metafields, and can be removed without a trace.
Match your catalogue
Your brands map straight onto the dossier. Where a rule is missing we add it, reviewed by a person before it goes live.
Switch on
The panel goes live on the product page. The AI chat stays off until you decide, and turns on with a single switch.
Guided setup, 100% free
A specialist sets everything up with you on the first integration, at no cost.
First 30 days free
Run it live on your store for a month before you pay anything.
One avoided return is worth more than a month.
Priced on value, not usage. Start with a free pilot and measure the impact before you decide.
- ✓ Full assistant on the product page
- ✓ Brand dossier ready, nothing to load
- ✓ Console and guided 1:1 setup
- ✓ A report of what was measured at the end
Starts your free pilot. No card. Cancel anytime.
- ✓ Everything in Pilot, live
- ✓ Up to ~1,000 conversations a month
- ✓ Custom size rules for your store
- ✓ Italian and English
- ✓ Console and email support
- ✓ Everything in Starter
- ✓ Up to ~5,000 conversations a month
- ✓ Unlimited custom rules, all brands
- ✓ Deeper analytics and guided onboarding
- ✓ Priority support
Enterprise
Chains, large catalogues, multiple stores. High volumes, dedicated onboarding, bespoke brand dossier, SLA, and a revenue share option.
A single avoided return on a 390 euro piece covers eight months of Starter. The AI itself costs about 0.3 cents a conversation, so the price never chases your traffic.
Your store stays yours. Your customers' data stays quiet.
An AI that talks to your customers touches their words and their measurements. Here is exactly what it does with them, and what it never does.
Read only on Shopify
Read only scopes, by choice and for good. No metafields, no writes to the catalogue, nothing to migrate, nothing to undo. Remove the app and no trace remains.
No stored store token
We never keep a token to your store. A short lived one is minted per request and held in memory only. Every call to the panel is signed and verified, or refused.
Spend you can cap
The chat is born switched off. When on: six messages a session, a hard daily token ceiling, and two separate API keys so a runaway job can never touch the assistant facing your customers.
Retention you decide
Conversations contain the measurements a customer typed. How long they are kept is a setting per store. Open contact requests are never deleted, handled ones age out on your schedule.
No personal data in events
Usage events carry a salted daily visitor hash, never a name, an email or an IP. Product titles and descriptions are treated as data the model reads, never as instructions it follows.
Model provider disclosed
The chat model is provided by Anthropic and receives only the rows our code chooses for that product: no tools, no web, no database access. Full details in the privacy notice.
Others size the body. We size the brand.
In a luxury multibrand catalogue that is the only thing that matters, and it is exactly what tools built on the average body cannot do.
Not a statistical average
Others map your measurements onto a generic chart, or guess from what similar shoppers bought. We keep a dossier per brand: which scale it labels in, how it cuts, where it runs tight. Knowledge no single store could build alone.
No cold start
Machine learning rivals need mountains of purchase and return history before they work. Our dossier makes a new store useful from the first product and the first customer, with nothing to load.
It advises, it does not push
The engine is deterministic and the AI is caged. It advises only when it is sure, never invents measurements, prices or stock, never promises free returns, and tells the shopper not to buy when the size is not there.
Not fuller carts
Almost everyone optimises for add to bag, which quietly grows returns. We optimise for the right size, even at the cost of a sale. In luxury a return costs more than a missed sale.
- AaHeight never picks a size. It governs length, not girth. If only height was given and the answer still names a size, the answer is replaced.
- $0Returns are never called free. The one line that would create an obligation. Tested in two languages, zero promises made.
- >1More than one size away, it says do not buy. The engine measures the distance to the nearest size in stock. Beyond one step, it tells the shopper the size is not here.
- </>No system words. Verdict, database, model, prompt. The out gate drops any answer that leaks them, or a link, or a price.
- 1?One question, warm and short. Two per conversation at most, then it offers a stylist.
What a merchant asks first.
How is this different from True Fit or Fit Analytics?+
Those tools size the average body, from your measurements on a generic chart or from what similar shoppers bought. We size the brand: a dossier of 307 fit rules across 55 brands, each traced to a source. In luxury multibrand, low volume and a hundred brands, the statistical approach has nothing to learn from. Ours works from day zero.
Does it write anything to my store?+
No. The Shopify app runs on read only scopes and never creates a metafield or writes to your catalogue. All the sizing knowledge lives in the TrueToSize database, which is also what makes it portable to the next store without touching theirs.
Where does my customers' data go, and which AI model do you use?+
Conversations and contact requests live in the TrueToSize database, not in Shopify. The chat model is provided by Anthropic and receives only the rows our code chooses for that product, with no tools and no database access. Retention is a setting per store. The full notice is on the privacy page.
How does the free pilot work?+
Thirty days, the full assistant and a brand dossier already built, with guided 1:1 setup. At the end, a report shows what was measured over the period, so you decide with the data in front of you. No card required.
Can the AI recommend the wrong size?+
The size is chosen by a deterministic engine, not the model. The model only writes the surrounding sentence, and an out gate throws the answer away if it fails to name the engine's verdict or breaks a rule. In 751 tested answers it never invented a size and never contradicted the engine.
My brand is not in the dossier. Can you add it?+
Yes. 55 brands already carry a rule, and where a rule is missing the engine still uses the garment's own measurements and the brand's typical ones. New rules are written only from a specific source and reviewed by a person before they go live.
How is the AI spend controlled?+
The chat starts off. When it is on, it is capped at six messages a session, rate limited per visitor, and stopped hard by a daily token ceiling. Measured cost: about 0.3 cents per conversation.
Why is pricing not usage based?+
Because the value is in the returns avoided and the extra sales, not in the number of chats. A fixed subscription is predictable and never gives you meter anxiety. Over the included conversations nothing switches off, we simply talk.
Which languages does it answer in?+
Italian and English, following the language the customer writes in. If a shopper writes in another language, the assistant says in one line which two it speaks.
Start free. Measure the impact.
Put the right size on every product page and see what moves on your own catalogue.
30 days, no card, brand dossier included. A 15 minute call to start.