AI for Retail in Kenya
Head office decisions that fit the branch rather than the average.
What is retail?
AI for retail covers branch level performance visibility, till and mobile money reconciliation, pricing consistency and shrinkage detection for shops and multi branch retail operations.
A chain of eight shops is usually managed as one. Head office looks at aggregate sales, sets one price list and one range, and makes decisions that fit no individual branch particularly well.
The averages hide the interesting things. One branch sells a line that dies everywhere else, another has shrinkage twice the group rate, and a third is quietly the only reason the group looks profitable.
Alongside that sits reconciliation. Takings arrive across a till, a bank account and one or more M-Pesa numbers, and matching them daily by branch is the work that reveals both errors and losses.
Why we report by branch
Duara works with retail chains, supermarkets and multi branch shops from Nairobi on branch visibility, reconciliation and stock loss.
We report by branch rather than in aggregate, including when a branch is performing badly. Group averages are comfortable and they conceal the two branches that need attention, which is the whole reason nobody noticed sooner.
Who this is for
- You run several branches
- Reconciliation across till and M-Pesa is manual
- Shrinkage is suspected but not measured
- Head office decisions are made on group averages
Industries we serve
- Supermarkets and minimarts
- Pharmacy chains
- Fashion and apparel retail
- Electronics retail
- Hardware shops
- Fuel station shops
Problems this solves
These are visibility problems, and every one of them is hidden by looking at the group instead of the branch.
- Decisions made on averages. One price list and one range are set for branches with genuinely different customers, and the aggregate describes no actual shop.
- Reconciliation across payment types. Takings arrive by cash, card and several M-Pesa numbers, and matching them by branch daily is slow enough that it slips.
- Shrinkage suspected and unmeasured. Everybody believes stock is being lost and nobody can say which branch, which line or how much.
- Branch performance invisible until late. A branch declining for two months is noticed at a quarterly review rather than in week three.
What we deliver
Performance reported by branch rather than aggregate, daily reconciliation across till, bank and mobile money, shrinkage measured by branch and line, and pricing consistency checked across sites.
Performance reported branch by branch
Daily reconciliation across till, bank and M-Pesa
Shrinkage measured by branch and product line
Pricing consistency checked across branches
Early warning when a branch starts declining
Range recommendations based on branch demand
Features
- Branch level, never aggregate. Group averages conceal the two branches that need attention, which is precisely why nobody noticed sooner.
- M-Pesa reconciled with the till. Takings across several mobile money numbers matched against till and bank daily, which is where both errors and losses surface.
- Shrinkage located, not estimated. Which branch and which lines, rather than a group figure that tells you something is wrong somewhere.
- Price consistency checked. Differences between branches are flagged, since most are errors rather than deliberate local pricing.
- Decline flagged in weeks. A branch slipping is raised while it is still recoverable rather than at a quarterly review.
Benefits
- Problem branches surface early. Branch level reporting shows decline in week three rather than at the quarterly review.
- Reconciliation stops slipping. Daily matching across payment types catches errors while the day is still recoverable.
- Shrinkage becomes actionable. Knowing which branch and which lines turns a suspicion into something you can actually address.
- Range fits the customers. Branch level demand shows where the group range is wrong for a particular location.
Technologies we use
- Branch performance reporting. Sales, margin and movement per branch rather than group totals.
- Multi channel reconciliation. Matching till, bank and mobile money takings daily by branch.
- Shrinkage analysis. Locating stock loss by branch, line and period rather than reporting a group rate.
- Price consistency checking. Comparing prices across branches and flagging unintended differences.
- Branch decline detection. Identifying a branch moving away from its own normal pattern early.
Almost everything useful in multi branch retail is hidden by aggregation. The group figure is the one number guaranteed to describe none of your shops accurately.
How the process works, step by step
Retail projects usually take five to nine weeks, starting with reconciliation because it surfaces both errors and losses.
-
Step 1
Free branch review
We look at how branches are currently compared and how reconciliation is done today.
-
Step 2
Reconciliation first
Daily matching across till, bank and M-Pesa goes first, since it finds errors and losses together.
-
Step 3
Branch reporting built
Performance is reported per branch, which usually changes which shops management worries about.
-
Step 4
Shrinkage located
Loss is measured by branch and line once stock and sales data are matched reliably.
-
Step 5
Pricing and range
Consistency checking and branch level range recommendations follow.
Quality assurance and how we work
The commitments attached to systems that measure branches and the people in them.
- We report by branch, including the bad news. Group averages are comfortable and they conceal the branches that need attention.
- Shrinkage findings are not accusations. Stock loss has many causes including recording error and damage, and we present findings as something to investigate rather than as evidence against staff.
- Reconciliation differences are surfaced daily. A difference found the same day is recoverable. The same difference found at month end usually is not.
- We do not build till level staff surveillance. Measuring transactions is operational; monitoring individuals continuously is a different thing and we would decline it.
- Branch comparisons account for context. Comparing a town branch to a mall branch on raw numbers produces a conclusion that is simply wrong.
Frequently asked questions about retail
Why report by branch rather than group?
Because the group figure is the one number guaranteed to describe none of your shops accurately, and averages conceal the branches needing attention.
Can it reconcile M-Pesa with the till?
Yes, across several mobile money numbers alongside cash and bank, matched daily by branch.
Why does daily matter?
A difference found the same day is recoverable. The same difference found at month end usually is not.
Will it tell us where shrinkage is?
By branch and by line, rather than a group rate that only tells you something is wrong somewhere.
Does that mean accusing staff?
No. Stock loss has many causes including recording error and damage, and findings are presented as something to investigate rather than as evidence.
Can it monitor individual till operators?
We measure transactions at branch level. Continuous monitoring of individuals is a different thing and we would decline building it.
Does it check pricing across branches?
Yes, and most differences turn out to be errors rather than deliberate local pricing.
Can it recommend what each branch should stock?
Yes, from that branch own demand, which frequently differs from the group range more than management expects.
How does branch comparison work?
Against context rather than raw numbers. Comparing a town branch to a mall branch directly produces a conclusion that is simply wrong.
When would we know a branch is declining?
Within weeks, against its own pattern, rather than at a quarterly review when recovery is harder.
How long before reconciliation improves?
Five to nine weeks, with daily matching going first since it surfaces both errors and losses.
Does it work with our point of sale system?
Most common systems, reading sales and stock movement rather than replacing the till.
Can small chains use it?
Yes, and chains of three to ten branches often gain most, since they are large enough to lose visibility and small enough to lack analysts.
What about a single shop?
Reconciliation still helps. Branch comparison obviously does not, and we would say so rather than selling the full thing.
Does it forecast demand?
Basic patterns from your own sales, and we would fix stock record accuracy before promising anything more than that.
Can it handle promotions?
Yes, where you tell it in advance. An unannounced promotion looks like an anomaly and gets flagged as one.
Will it reduce stock loss?
It locates it. Reducing it depends on what you do with the finding, and we would not claim the system does that part.
Do reconciled takings reach our books?
Yes, posted through to your accounting system rather than being entered a second time branch by branch.
Who can see branch performance data?
Whoever you authorise. Sales and margin figures by branch are commercially sensitive and stay inside your systems.
What does a retail project cost?
Branch count and transaction volume determine it, quoted after the branch review.
Glossary of terms
Retail terms in plain language.
- Shrinkage
- Stock lost without a recorded sale, through theft, damage or recording error.
- Reconciliation
- Matching takings across till, bank and mobile money against recorded sales.
- Branch level reporting
- Measuring each shop separately rather than reporting a group total.
- Range
- The products a branch stocks, which should reflect its own customers rather than the group.
- Basket
- What a customer buys in one transaction, which differs sharply between locations.
- Margin
- Profit per line after cost, which branch level reporting frequently reveals as uneven.
Related services in Industries
Talk to us about retail
Compare your best and worst branch on the same measures for last month. The gap is usually larger than the group figure suggests.
Start your project