Footfall is counted, conversion is guessed
Entries come from a door counter, bills come from the POS, and the two are joined by hand at month end. By the time a store manager sees a conversion number, the trading day it describes is six weeks gone.
Industries / Retail & Malls
Retail & Malls, run by systems that finish the work.
Conversion you can read by the hour, stock that matches the catchment, and shrinkage found while the trail is still warm.
Footfall analytics, shrinkage detection, demand forecasting and AI merchandising.
The picture on the ground
What Retail & Malls actually looks like from the inside.
Retail already measures everything and trusts almost none of it. The door counter reports entries, the POS reports bills, and the two are reconciled once a month in a spreadsheet that nobody defends in a review meeting. Category teams plan against national averages while a single store's catchment behaves nothing like the average. Shrinkage is discovered at stock take, six months after the process that caused it stopped working.
We work on the joins. Footfall tied to bill count gives a conversion rate per store per hour, which turns rostering from a habit into a decision. Sell-through tied to local weather, festivals and competitor activity gives replenishment a reason to override the average. Camera footage tied to exception transactions turns a shrinkage number into a specific till, a specific hour and a specific process gap.
For malls the unit of work changes but the shape does not. Leasing needs catchment and dwell evidence per zone, tenants need footfall they can act on, and marketing needs to prove what a campaign weekend actually delivered — all from cameras, parking systems and tenant sales declarations that already exist.
Where it hurts
Four pressures we hear in every review meeting.
Named the way your operators name them, not the way a vendor deck names them. If none of these land, we are the wrong partner for this problem.
Replenishment runs on last month's average
Sizes and colours break early in one store and sit dead in another. Inter-store transfers happen late, markdown absorbs the error, and the category review argues about a number instead of a decision.
Shrinkage surfaces at stock take
Voids, refunds without customer, discount abuse and back-door movement leave a trail in the POS journal and on camera. Nobody has time to watch forty stores, so the loss is only ever quantified after the fact.
Store teams live in the back office
Indents, GRNs, damage notes, visual-merchandising photo checklists, rosters and audit responses. Hours that were budgeted for the shop floor are spent on forms that a system could complete.
What we build
The work that pays for itself first.
Ordered the way we would sequence them in a retail & malls engagement. Each one links to the capability behind it.
Conversion and dwell from the cameras you already own
Edge inference on existing CCTV produces entries, dwell by zone and queue length, joined to POS bills for a conversion rate per store per hour. No new hardware in most sites — the NVR feed is enough.
Queue alerts that open a till before the walk-out
Live queue length at the cash counter triggers an alert to the floor manager at the threshold your own data says causes abandonment, rather than at a number picked from a manual.
Store-level demand forecasting and replenishment proposals
Sell-through, stock on hand, lead time, local calendar and weather produce a proposed indent per store per SKU. The buyer approves or overrides; every override is captured and improves the next run.
Shrinkage and SOP review at exception level
Exception transactions in the POS journal — voids, manual discounts, no-sale opens — are paired with the matching seconds of footage and queued for a loss-prevention reviewer instead of a full-day trawl.
A category assistant over sales, stock and supplier terms
A private assistant that answers from your own sales history, stock ledger, margin sheets and supplier agreements, with the source row cited. It replaces the request-and-wait loop between category and MIS.
Tenant and marketing reporting for malls
An agent that assembles the weekly pack — footfall by entrance and zone, parking occupancy, tenant sales declarations, campaign periods — and writes the commentary a leasing team would otherwise draft by hand.
The systems it plugs into.
Read-only to begin with, write access only where a workflow needs it, and every integration documented before it goes near production.
POS and billing systemsERP stock ledger and GRN historyCCTV, NVR and people countersLoyalty and CRM records
E-commerce and marketplace order feedsWarehouse and distribution dispatch dataMall lease, CAM and parking systems
How it gets judged
Agreed before the build, measured after it.
These are targets and mechanisms, not results borrowed from somebody else’s project. We baseline each one on your data in the first fortnight, and report against that baseline monthly.
Conversion rate, by store and by hour
The join between footfall and bill count. The target is a number a store manager can act on the same day, not a monthly reconciliation.
Stock-out days on A-class lines
Days where a top-selling line is unavailable in a store that had demand for it. Forecast quality is judged here, not on model accuracy in the abstract.
Shrinkage value identified per quarter
Loss traced to a specific till, hour and process rather than absorbed into the stock-take variance.
Back-office hours per store per week
Time returned to the shop floor once indents, GRN matching and checklist reporting are automated.
What we design around
The constraints come first, not at the security review.
Digital Personal Data Protection Act, 2023
Footfall and loyalty data are personal data once they can be linked to an individual. We design for purpose limitation, a stated retention window and a defensible notice at the store entrance, and keep the data map current so a request can be answered.
Restraint on face recognition
Counting, dwell and queue analytics do not require identity. Unless there is a specific, lawful and documented reason to identify individuals, we build anonymous pipelines that discard the frame after inference.
PCI-DSS scope
Nothing we build needs card data. Analytics consume the transaction record, not the payment instrument, so the cardholder data environment stays where your acquirer already assessed it.
This is how we scope and build, not legal advice. Your compliance, risk and legal teams stay the authority on what applies to your organisation, and we work to their reading of it.
Your first project
Start narrow. Prove the number. Then scale it.
Start with one cluster of five to eight stores and a single question: what is conversion, by hour, in each of them. Existing cameras, existing POS export, four to six weeks. It is the cheapest way to prove the data plumbing works, and the conversion series it produces is what every later piece of retail work — rostering, replenishment, campaign measurement — is built on.
Discovery Sprint
1–2 weeksWe sit with your operators, map the workflows, and score every candidate use case on value, data readiness and risk. You leave with a ranked backlog and a cost model.
- Use-case scorecard
- Data readiness report
- Costed roadmap
Proof of Value
3–4 weeksWe build the single highest-value agent against your real data and measure it against the humans doing that job today. If the number is not there, we say so.
- Working prototype
- Evaluation set
- Benchmark vs. baseline
Bring us one workflow from your floor.
Describe how retail & malls works in your organisation — the systems, the constraints, the part that goes wrong every week. We will tell you whether it is worth automating and what the first project would cost.