Basaltica builds software that understands physical spaces without identifying anyone in them.
Inclina, our first product, turns the cameras a retail chain already has into an experimentation system: change something in some stores, hold it back in others, and see what it did at each step between the door and the register.
The problem
For twenty years, online retail has measured every step of the funnel and tested every change before rolling it out.
Physical stores, where most retail sales still happen, know what was sold at the register and almost nothing about what happened before it.
Layouts, campaigns and staffing are decided by feel and judged weeks later, without knowing what caused what.
Inclina, our first product, in three layers
Every visit becomes an anonymous session from entry to exit: where it went first, what it stopped in front of, whether someone from the team approached, whether it reached the checkout. Counted by store, zone and hour, and comparable across stores.
Turn a decision into an experiment. We pair comparable stores, work out before the test how many stores and weeks it needs, measure both groups, and report the difference with the confidence the sample allows, step by step through the funnel, with every number traceable to its source.
Let the store respond while shoppers are still in it: open a register before the queue forms, bring light to a zone that has gone quiet, tell the manager what is happening now. Every reaction is tested as an experiment before it becomes a rule.
We don't promise a number. What a change is worth gets measured in the customer's own stores.
How it works
This is how the first version of Inclina is designed to work.
The cameras already in the store
Measurement runs on the store's existing IP cameras. No new sensors and no installation project to get started.
A capture agent on a store computer
A small agent captures video and sends it out encrypted. It does no analysis inside the store.
Processing in the customer's region
Video is processed in cloud infrastructure located in the customer's own region, with inference running on NVIDIA GPUs.
Video destroyed on a fixed schedule
Once video has been turned into events, it is destroyed on a fixed schedule. What remains are events and numbers, kept for as long as the customer decides.
Stores without IP cameras aren't eligible, and we say so up front. Processing inside the store is planned as a later module.
Privacy by rule
- Never identifies anyone.
No facial recognition and no biometric identification, in any module, at any stage of the product.
- Never links one visit to another.
A visit is followed from entry to exit and then forgotten. Returning customers come from the retailer's CRM, never from a camera.
- Never follows a gaze.
What a shopper considers is read from movement and posture: stopping, turning toward a table, reaching for a product.
- Never infers emotion or physical condition.
Not for shoppers, and not for staff.
These are rules of the product, not settings. No configuration or contract switches them off. The store is measured; people are not: team metrics exist only as store totals, and only where the team is large enough that no one can be singled out.
Who it's for
Retail chains with assisted selling
Fashion, footwear, beauty, chocolate and gifts, eyewear, jewelry: many similar stores, where the sale depends on what shoppers see and who serves them.
Marketing and trade marketing
The teams that decide campaigns, assortment and visual merchandising, and today make those calls with the thinnest data in the company.
United States and Brazil
The United States is our target market. Brazil is where we build, and where we plan to run our first pilots.
Team
Daniel Mendes
Co-founder, CEO and CTO
Physicist (PUC-Rio). Thirty years building high-performance, real-time computer vision and edge systems, from energy research to industrial safety.
Beatriz Ansay
Co-founder, Product
Architect (UFPR) with training in industrial design (UTFPR). Ten years designing digital and physical products for technology-based startups, across RFID, computer vision and thermal imaging.
Design partners
We are choosing the first retail networks to build this with. A design partner runs the first version of Inclina in a small set of comparable stores, tells us what the team actually uses, and shapes what we build next. In return: early access, setup included, and a direct line to the people writing the product.
We are looking for networks with many similar stores, a marketing or trade team that wants to test decisions, and IP cameras already in place.
Company
Basaltica, Inc. is a Delaware corporation.
Our engineering team is in Brazil by choice: strong computer vision and edge talent outside the usual hubs, and the market where we plan to run our first pilots.
We will be hiring computer vision and edge engineers in Brazil.