ANALITIX · physical store intelligence

Let us start with the basics

What is ANALITIX?

A system that counts the people walking into your store and tells you what they did inside: which door they came through, what stopped them, how long they waited, whether anyone helped them, and whether they ended up buying. All of it inside your own Odoo, next to your sales.

A camera and a small computer

The camera is usually already on the wall — analog, webcam and IP cameras all work. The computer is the size of a book, it lives in the back room and it needs nothing special.

Inside your Odoo, not alongside it

The numbers land in the ERP you already use and match themselves against your point of sale. It is not another system with another password, or a dashboard nobody opens.

No image ever leaves the store

Recognition happens on that computer and the video frame is destroyed right there. What travels to Odoo are numbers — never photos, never video.

In one sentence: it gives your physical store the numbers an online store has had since day one. How many people came in, what they looked at, where they lingered, and the moment they left without buying.

So there is no misunderstanding

And what ANALITIX is not

It is the first question everybody asks, so it comes before anything else.

It is not surveillanceIt does not record, does not store and does not replace your CCTV. They are different things and they can coexist.
It does not identify anyoneIt knows the person at 11:04 is the same one at 11:40, without knowing who they are. Putting a name to a customer is a separate feature — optional, and off by default.
It does not police your teamStaff are excluded from the count precisely so they do not pollute the figure. What it measures about a salesperson is whether the alert reached them and what they did with it.
It is not a people counterThe counter is the floor, not the ceiling. The question it answers is who you failed to sell to, and why.

An ordinary afternoon

At 2:02 pm a customer walked in. At 2:11 pm she left with nothing.

Nobody in the store noticed. It will not show up in any report you have today, because nothing happened — and "nothing" is exactly what an ERP cannot record.

2:02 pm

She walks in

A virtual line across the doorway: what counts is which direction you cross it. The salesperson stepping out for a coffee is recognized and not counted — otherwise ten trips out is ten visitors.

Overhead view of the doorway with the virtual line, tracked people in boxes and the salesperson flagged as staff

2:04 pm

She stops at the ring case

Zones are pieces of your own floor, drawn once over a photo of your store. From then on it measures how long people spend there and whether anyone approached them — by zone, by hour and by day.

The store seen from a high corner with four zones drawn over the counters and their average dwell time

2:06 pm

Four minutes. Still on her own.

Plenty of time, nobody helped her, no receipt. Each of those three things on its own is completely normal; together they are the only reliable alert a store ever gets. This is the minute everything else exists for.

A customer waiting with her arms crossed in front of a counter, her expression read as sustained negative

2:11 pm

She leaves

Nine minutes inside, no bag and no receipt. Tomorrow this will be indistinguishable from a slow afternoon.

A visitor leaving empty-handed after nine minutes while another customer is handed a bag at the counter

That was one. Last month it happened twenty-nine times in that store, and in nine of them nobody even approached. That is the number that appears in no report you have today, and the one you will have every morning.

The same afternoon, with ANALITIX on

Nothing changes for the customer. A phone buzzes.

Odoo chat as the salesperson sees it, with the alert that someone has spent four minutes at the ring case with nobody approaching — **2:06 pm — the alert.** In your own Odoo chat, to the salesperson covering that zone on that shift. Short enough to read mid-shift and act on without opening anything. The customer never knows.

Monthly report per store: visitors, sales, conversion, walkouts detected and recovered, and estimated revenue recovered against what the store pays — **End of month — the tally.** Walkouts detected, walkouts recovered and roughly what that was worth, next to what the store pays. Labeled **estimate** everywhere, because that is what it is.

The number nobody has

You sold $80,000 on Saturday. Was that good or bad?

Without knowing how many people walked in, that figure means nothing. Maybe 40 people came in and 30 bought — an outstanding Saturday. Or maybe 400 came in and 30 bought, and you lost 370 opportunities without ever knowing. Same sales figure, two completely different businesses.

Your store’s traffic, hour by hour and door by door.
How many of the people who came in actually bought — your real conversion.
What every person crossing your door is worth, whether they buy or not.
And the most expensive of all: who left without anyone helping them.

What you start measuring

Conversion rateVisitors to receipts

Average transaction valueWhat each sale leaves behind

Units per transactionHow well they upsell

Sales per visitorWhat each person who walks in is worth

Visitors per transactionHow many you must attract to make a sale

Density per m²Sales and traffic per square meter

Counting properly, not just counting

A badly built people counter lies, and it lies in your favor

An inflated figure gives you false conversion and expensive decisions. These six things are what separate a real number from a flattering one.

Every one of your doors

One entrance, three, or however many the store has: doors are defined in configuration, never in code. A wizard creates the store, its doors and its devices in a single step.

Visits, not crossings

The customer who steps out to take a call and comes back is one visit, not three. Counting it three times would report a third of the true conversion.

Buying units

A family of four walking in together is one buying decision, not four. A store that counts them separately sees 25% conversion where the truth was 100%.

Your staff left out

Your team goes in and out all day through any door. Their crossings are excluded so they neither inflate footfall nor sink your conversion.

Device health

Every camera reports that it is still alive. If one goes down or starts sending odd figures, the system raises the alert — you do not find out three weeks later staring at a gap in the chart.

Kill switch

One button per store stops capture instantly. And every device has its own key, revocable in a click, valid only for its own store.

The sales floor

Where people stop, which display holds them, and who left without being spoken to

Zones and displays

"Twelve minutes inside" tells you nothing

You split the floor into the zones the store actually has and mark the displays in each one. Then the sentence changes: "eleven of those twelve minutes at the ring case, and nobody spoke to her." That tells you exactly where the money went.

Attention by zone and by display, next to the sales of the products on show there.
The two findings that are worth money: the display that draws a crowd and never sells, and the one that sells well from a cold corner.
Receipt attribution: which visit produced which sale, anonymously.
Behavioral signals: repeated entries and exits, long unattended stops, groups that split up.

Lost sales

The number that makes it worth it

A lost sale is not "somebody who did not buy." It is somebody who stood in front of a product for a long time, was never helped, and left without a receipt. All three conditions together, because each one on its own is perfectly normal.

It is detected while the customer is still inside, which is when something can still be done.
Each one is recorded with its zone, its duration and whether anyone got to it in time.
The ones that were not recovered can open a lead in the CRM of your own Odoo.
And the recovery is measured: how many were flagged, how many were attended and how many ended in a sale.

The alert to the salesperson

Discreet by design: the customer never knows

Detecting the lost sale is useless if nobody moves. ANALITIX alerts whoever is covering that zone at that moment — through the channels the store chooses, and with three of them the customer notices absolutely nothing.

Three discreet channels: the Odoo mobile app, internal chat (Discuss) and WhatsApp. Only the assigned salesperson sees them.
Two the customer might notice: an audible chime and a message on the screen. Both ship disabled, and each one spells out in its own help text what it costs you in discretion.
Rostering by zone, day and hour, escalating to a supervisor when nobody is covering.
Response time is recorded, along with which channel actually delivered the alert.

The screens react

And they react on the right screen

A bundle promotion belongs on the screen next to where the group is standing, not the one at the entrance. The rules engine resolves where first and what second. The store edits the rules without calling us.

Ready-made triggers for groups, known customers, dead hours, profiles and people on their way out.
Personal greeting with a social rule: a recognized customer is greeted by name only if they came in alone. If they are with someone, the screen stays neutral and the greeting goes to the salesperson instead — putting a name on a screen in front of a companion tells that companion something.
The context is assembled from what you already have: CRM, point-of-sale history and the customer’s special dates.
It leans on DISPLAX for the content, and on the signage connector we also publish for Odoo.

For the owner and for the team

Floor coachingAlerts received, response time and close rate

Visit frequencyWho comes back, and how often

LeadsLost sales that open a lead in the CRM

AttendanceWritten to Odoo’s own hr.attendance

Monthly reportIn plain language, to the owner’s inbox

Coaching, not employee surveillance

Measured against what actually reached them

Per salesperson: how many alerts they received, how fast they responded and how many of the ones they took ended in a sale. Measured against what they responded to, so nobody is marked down for an alert that landed while they were serving another customer. Attendance is written to your own Odoo attendance module, and never on top of a manual correction.

The monthly value report

To the owner’s inbox, in plain language: visitors, conversion, lost sales detected, how many were recovered and roughly what that was worth — labeled an estimate every single time, because inflating that figure is the fastest way to lose the client who eventually checks it.

Manuals and videos inside the app

A manual for the store and an installation guide for whoever mounts the cameras, in English and Spanish, served according to each user’s language. Plus videos of every flow recorded on the real interface. The documentation lives inside the product, not in an email nobody can find six months later.

Privacy by design

Not a single image leaves your store

This is the question everybody asks, and the answer is in the architecture, not in a promise. The camera processes on site and sends only counts and irreversible numbers: you cannot rebuild a face from them. No video, no photo, no name.

The anonymous signatures are stored encrypted and delete themselves within the retention window the store sets.
Demographic data (age range, apparent gender, expression) is optional and ships disabled.
Recognizing is not identifying: the system knows it is the same person from ten minutes ago, not who they are.
Identifying a customer by name only happens if they left their details at the register themselves, it is optional, and every later lookup against that list is audited.
Face matching never leaves the store, and is never run against the whole database.

The watch list

The only feature that names a person, and the one we deliberately built with the most friction.

Nobody is added automaticallyManual entry, with a dated reason

Two-person controlSomeone else approves; whoever added the entry cannot

It expiresExpiring is the default; staying takes a deliberate act

Never on a screenA silent alert to a restricted group

A separate roleSecurity, outside the commercial hierarchy

Audit trailReads are logged, not just changes

When it is no longer one store

Chains and regions, with no cap on locations

Everything above works for a single location and needs none of this. This part is for whoever has twenty, or two hundred.

The head-office console

Every location measured by the same yardstick: conversion, average transaction, units per transaction, lost sales detected and recovered. Head office already knows which one sells most — this is the first time it can see which one converts worst.

Compared against its own region

Each location is measured against its region, on the same day. That strips out the week, the weather and the season, and leaves what is genuinely particular to that store.

Permissions that do not go stale

A regional manager is given a region, not a list of stores — so it covers the locations that open next year without anyone remembering to update anything.

What nobody shows you in a demo

Four things that actually change the month, and almost never get explained

A lit window marked as a display, with two people stopped counted as one buying unit, and the day’s stops against the receipts that came out of what it shows — **Displays: what stops people.** Zones tell you *where people go*. A display tells you *what stops them* — a window, a case, a table. On its own, that attention is a vanity metric, so it is never reported alone: it is always crossed against **the sales of the products sitting there**. A window where everybody stops and nobody buys is a completely different problem from one where nobody stops, and only the cross-reference tells them apart.

Two screens in the same store: the one in the zone where the customer stopped is the chosen one and shows her message; the one facing the entrance is explicitly excluded from this rule — **The right screen, not "the screen."** The screens in a store are not interchangeable: the one at the counter, the one at the exit and the one facing the street do different jobs, and a message sent to "the screen" is a message sent to the wrong one. Every rule resolves **where** before **what**, and the destination comes from the zone where the thing happened.

The empty sales floor after closing, with the counter reporting one person inside and a visit open for six hours with no exit recorded — **Came in and never left.** The alert the others cannot see: they check visits opened in the last hour, and a visit open since the morning dropped out of that search a long time ago. It almost always means an exit the camera missed — and that matters, because a visit that never closes inflates occupancy for the rest of the day. Sometimes it means somebody is still inside. Either one is worth thirty seconds.

A camera view almost entirely blocked by a cardboard display, with the device reporting online and zero crossings in 45 minutes — **The camera that is online and blind.** Somebody left a display under the lens; the device kept reporting in, perfectly happy.

What you can adjust yourself

Every trigger, on one screen

What counts as a lost sale, how long it waits before repeating an alert, how long before it gives up on it as unattended, which channels it may use, and the five behavioral conditions — including how long a visit may stay open before somebody reviews it. None of that is in the code.

The store’s floor and alerts tab: lost sales, alert routing, channels separated by discretion, sustained-expression thresholds, the five behavioral conditions and the per-zone threshold table — **And right at the bottom, the important part: every zone carries its own.** A fitting room is not a ring case, so they do not share a number. Interest and possible-lost-sale timings are set zone by zone.

Odoo Settings with the Analitix section: data retention in days and a summary of stores configured and activated — **The little that is global lives in Odoo Settings.** One single thing: how many days crossings are kept before the nightly job deletes them. It is a personal-data policy and it has to sit where the owner will find it, not buried in technical parameters.

How it gets deployed

We come out, install it and calibrate it with you

ANALITIX installs on top of your Odoo and we mount the cameras ourselves — it is exactly the kind of work our name is built on: on-site technology. If you do not have the ERP yet, that is what INOOVIX is for.

Assessment and floor plan: how many doors, which zones and which displays matter.
Installation: edge cameras at the doors and on the floor, with whatever networking is needed.
Calibration: staff exclusion, visit windows and thresholds, store by store.
Follow-through: we review the report with you until the decisions genuinely change.

Three tiers, per location

Each tier contains the one below it, and it is licensed per store: your flagship can run on Actions while the small ones run on Counting.

CountingHow many people come in, and your real conversion against the point of sale

Visual analyticsEverything above, plus everything the camera can measure: visits, groups, demographics, zones and displays

ActionsEverything above, plus everything that acts: lost sales and the discreet alert, screens, greetings, attendance and coaching

Moving up a tierIt is configuration: nothing to reinstall and nothing to remount

A rule we imposed on ourselves: if something changes from one store to another, it is configuration and never code. There is an automated test that fails if somebody forgets — the same one that verifies the system behaves identically with one door and with seven.

The next step

Let us look at it with your store’s numbers

In the demo we walk through the system with realistic sample data, look at how many doors and zones your store has, and the scope and price come out of that. There is nothing to install for that conversation.

Book a demo See all the modules

Your physical store, with the numbers an online store takes for granted