The problem, measured

Independent retailers lose 2–3% of revenue to shrink they never see.

National Retail Federation, National Retail Security Survey — shrink as a share of retail sales.

Sarva is the back office and AI security layer for convenience stores and fuel retailers — one platform for what is currently five vendors, three logins and a spreadsheet.

Talk to us See the platform

Built and run at three Georgia stores before it was sold to anyone.

The problem

The losses are known. The visibility is not.

None of what follows is new to an operator. What is new is being able to see it while it is happening rather than at the end of the quarter.

01
Shrink runs 1.4–1.6% of sales industry-wide, and higher in convenience.
Most of it is never attributed to a cause. Cameras record it; almost nothing analyses it, so the loss is discovered at inventory rather than prevented at the shelf.
National Retail Federation, National Retail Security Survey
02
A single store commonly runs five or more disconnected systems.
Back office, price book, cameras, loyalty, delivery — each with its own login, its own export, and no shared view. The reconciliation between them is done by hand, usually by the owner, usually at night.
Observed across the three Georgia stores this platform was built in
03
Third-party delivery commissions reach 30% of order value.
On a category already running thin margins, that is frequently the difference between a profitable channel and an unprofitable one — and few operators see the per-channel number at all.
Published merchant terms, major delivery marketplaces

The platform

Nine modules, one login, one bill.

Each replaces something an operator is already paying for, or already doing by hand. Two of them have no equivalent in this market.

AI Guard

Sarva exclusive

Cameras that analyse rather than record — loitering, register events and after-hours entry raise an alert while it is happening, not at review time.

Profit Pulse

Sarva exclusive

Margin by item, category and channel, including the delivery commissions and labour most reporting leaves out.

Daily Close

Sales, tax, tenders and lottery reconciled in one screen, with over/short derived rather than typed.

Price Book

One catalogue across registers and channels, with cost-to-shelf tracking and multi-store price groups.

Inventory & Shrink

Adjustments with a review queue, so a large write-off is a decision someone made rather than a number that moved.

Invoice AI

Photograph a vendor invoice; it is read, matched against the price book, and held for approval before anything changes.

Lottery

Pack lifecycle, per-bin counts and variance — the highest-shrink category in most stores and the least instrumented.

Fuel & Compliance

Tank readings, delivery reconciliation, and the licence and permit calendar that otherwise lives on a wall.

Ask Sarva

Plain questions against your own data. It answers from what it queried, and says so when the data is not there.

How it works

Three steps, then it runs.

Connect the register

An on-site connector reads your existing POS — Verifone Commander first, others behind the same interface. Sales, departments, tax and fuel arrive on their own. Until it is connected, the same screens accept figures by hand, and the changeover requires no migration.

Screenshot to come

Connectors — agent status and last sync

Work the day as normal

Close the day, count lottery, approve an invoice, adjust stock. The platform records what happened and where it disagrees with what was expected — over/short, variance, unmatched items — rather than asking anyone to compile it later.

The Daily Close screen: sales by department, tax on an invoice basis, tender breakdown and a derived over/short figure. — Daily Close — tenders, drawer and over/short · Demo store

Read the exceptions

Each morning, a brief: yesterday against the same weekday last week, anything unusual, and what has been waiting too long. One concrete thing to do. If the data is not there, it says so instead of estimating.

The Command Center: sales yesterday, tax collected, stock review queue, invoices awaiting approval, low stock, open alerts and last close over/short. — Command Center — the morning brief · Demo store

Evidence

Built and run at three Georgia stores.

Sarva was not designed for convenience retail from the outside. It was built by an operator, in his own stores, to solve the problems he had that quarter — and every module in it existed because something was costing him money or hours before it did.

BP 155, a fuel and convenience site in Georgia, is the live pilot: the first store where the register connector, the back office and the camera layer run together. It is also where the figures on the right will come from.

Those are blank because the measurement period is not finished. They will be filled in with what was actually observed, and if a number disappoints it will be published anyway — a case study that only ever reports good results is an advertisement.

The pilot, in numbers

Stores running the platform: 3Georgia, operated by the founder
Shrink before / after: Not yet measuredFrom BP 155 inventory counts, first full quarter
Hours per week on back office: Not yet measuredOwner time log, before and after
Vendor systems replaced: Not yet measuredCounted per store at onboarding

No customer logos are shown because there are no customers to name yet.

Pricing

Published, in full.

Four of the five platforms an operator would otherwise shortlist require a sales call before quoting a price. Ours is on this page.

Sarva pricing tiers and included capabilities
Plan	Starter$99/monthOne site, back office and price book.	Operator$199/monthOne site, everything including AI Guard.	Multi-storeTalk to usThree or more sites, or a bespoke integration.
Daily close, tax and tenders
Price book and inventory
Invoice AI	50 / month	Unlimited	Unlimited
Ask Sarva
Lottery
Fuel and compliance
Profit Pulse
AI Guard (cameras)
Register connector
Multi-store rollup
Onboarding	Self-serve	Guided	On-site
	Start	Start	Talk to us

Per site, billed monthly, no contract. Cameras are additional hardware and are quoted at cost — existing RTSP or ONVIF cameras usually work, in which case there is nothing to buy. Payment processing is not live yet and is not included in any tier.

Perspectives

Writing on the trade, not on the product.

Short pieces on the operational problems behind the software. Written from running stores, not from a category report.

Loss preventionIn preparation

Why cameras that only record don’t prevent theft

Recorded footage answers "what happened" after inventory has already been lost. The argument for analysis at the edge, what it can and cannot detect, and why the useful alert is the one that arrives during the incident.

Not published yet

OperationsIn preparation

The true cost of a fragmented back office

Five subscriptions is the visible cost. The larger one is the reconciliation between them — done by hand, at night, by the person whose time is worth the most. A method for counting what it actually costs your store.