Smarty Bloodstock

in production

From scattered data to a research tool for horse buyers.

Smarty Bloodstock brings sale catalogs and results, natural-language questions, horse comparisons and notes into one experience built for the phone in a buyer's hand. It is an Equitechverse product, built with Daniel Menéndez: we designed and built it, from idea to production.

Type
Client · Equitechverse
Platform
Installable phone web app
Scope
Data, applied AI, product and payments

How it works

A buyer's morning before the ring: the screening question, the horse's sheet with a grade and a note, and the buyer's notes quoted back.

  1. 1The morning screenWhat sells today out of black-type mares, with consignor and barn.
  2. 2The whole hip in your handConsignor and barn, dam and sire, every figure with its sample size.
  3. 3Your grade and your noteA, B, C or Pass, and what you saw at the barn.
  4. 4What you marked, in your wordsYour A's quoted back, before you walk to the ring.
Illustrative demo · example horses and figures

Where it started

At a sale, a buyer reviews hundreds of lots in a few days. The information exists —catalogs, past results, every mare's and sire's record— but it is spread across PDFs, spreadsheets and each sale house's website.

The decision is made standing at the barn, phone in one hand. A desktop tool does not work there.

It began as an internal tool for a buying team: read the catalogs, recognize the same horse across sales and score every lot. The Sale Desk grew out of it, a product open to any buyer.

The problems and the decisions

  1. 1An AI answer that invents a figure costs a buyer money.

    What we decided
    The model understands the question, but cards, comparisons and figures are built by code from the database. Every statistic carries its sample size and, when there is no data, the answer says so.
    What we gave up
    Less free-form answers, in exchange for every number being checkable.

  2. 2It is used at the barn, one-handed, with no time to type.

    What we decided
    A web app that installs on the phone, with voice notes that are recorded, transcribed —live where the browser allows— and cleaned on save.
    What we gave up
    Without a native app we depend on each browser, Safari on iOS above all; the recording always stays as the fallback.

  3. 3The data comes from third parties, with their own terms of use.

    What we decided
    Each sale house's terms are reviewed with legal counsel before any data is exposed. The public site and its demo use only example horses and figures, with no house logos.
    What we gave up
    A plainer public site, in exchange for not putting the product at risk.

  4. 4Every AI question has a cost and a response time.

    What we decided
    The product measures spend and response time per model and per day in an admin console we built, and each plan has its question quota.
    What we gave up
    More parts to run: quotas, counters and a dashboard to maintain.

What we have achieved so far

We do not publish usage or business metrics yet: they will be added once measured.

  • In production, with every release tested in QA before it ships.
  • Paid plans with Stripe, cancellations and refunds.
  • Sign in with Apple and Google; English and Spanish.
  • Admin console with users, spend and response times.
  • AWS infrastructure defined as code.

How we built it

The same way we would work with you: nothing reaches production without QA, every important decision is written down with its why, and whoever arrives tomorrow can carry on without asking us.

  • 3sale houses in one database
  • 5months from idea to production
  • 2languages from day one

Facing a similar challenge?

Tell us what you need to solve. We can help you define the next step.

Let's talk about your project