Choosing inventory management software is the easy part. The hard part is the fortnight after you switch it on, when your system says 240 units and your shelf says 191, and nobody can tell you which number is wrong.
Most inventory rollouts in Kenya fail in that fortnight. Not because the software was wrong, but because the migration was done in a hurry — an approximate opening count, a product list imported straight from a messy spreadsheet, and no rule about who is allowed to change a stock figure.
This guide covers the migration itself. If you are still comparing tools, start with our guide to inventory management software in Kenya, which has the vendor comparison and the reorder-point and stock-turnover formulas worked in shillings.
Why the first month decides everything
Inventory software only ever knows what you told it. Every number it shows you traces back to two things: the opening count you gave it, and the movements it has recorded since.
Get the opening count wrong and every report is wrong for months, in a way that is almost impossible to unpick later. Worse, your staff learn within a fortnight that the system numbers cannot be trusted — and once they stop believing the screen, they stop using it, and you are back to the notebook with a subscription attached.
So the migration is not an IT task. It is a counting task with a software step at the end.
Decision 1: what counts as a SKU
Before you import anything, settle this. A SKU is one distinct sellable thing. If you sell a shirt in three sizes and two colours, that is six SKUs, not one product — because you need to know that the blue mediums are finished while the red larges are gathering dust.
Get this wrong in the obvious direction and you carry a single "shirt" line that tells you nothing useful. Get it wrong in the other direction and you create 400 variants nobody will ever count accurately.
The working rule: split it into separate SKUs when you would reorder them separately. If you would ring the supplier and ask specifically for blue mediums, it is its own SKU. If you would just ask for "twenty shirts", it is not.
Decision 2: who is allowed to change a stock number
In every business that ends up with untrustworthy inventory data, the same thing happened: everyone could adjust stock, and nobody had to say why.
Decide now:
- Who can receive stock against a purchase order
- Who can adjust a count after a physical check
- Who can write off damaged or expired goods
- Who can transfer stock to another branch
In sell.ke these are separate permissions, so a counter attendant can sell without being able to quietly correct a shortfall away. That single control does more for data quality than any feature on a comparison table.
Decision 3: your cut-over date
Pick a specific date and time — ideally the end of a trading day when stock is at its lowest, which for most Kenyan retailers is a Sunday evening or the day before a delivery.
Do not migrate "over the course of a week". Stock moving while you are counting it is the single largest source of phantom variances, and a week-long cut-over guarantees it.
Step 1: clean the product list before you import it
Export whatever list you have. Before it goes anywhere near the new system, fix these:
Duplicate entries. Almost every spreadsheet that has been maintained for more than a year contains the same product twice under slightly different names. Sort alphabetically and read through it — you will find them.
Inconsistent units. Decide whether you sell sugar by the kilo or by the packet and make every row agree. Mixing units is what produces stock figures that are wrong by a factor of twelve.
Missing cost prices. Without a cost price the system cannot calculate your margin, your stock value, or your GMROI. If you genuinely do not know, use your most recent supplier invoice rather than leaving it blank.
Dead products. Anything you have not sold in a year and do not intend to restock should not be migrated. Archive it. Bringing dead stock into a new system just means counting it forever.
A realistic clean-up on a 500-line catalogue takes an afternoon. It is the highest-value afternoon in the whole project.
Step 2: do a real opening count
Not an estimate. Not last quarter's figures. A physical count, done properly:
- Close, or count a closed section at a time. Nothing moves while it is being counted.
- Print count sheets without the expected quantities. A counter who can see the system figure will confirm it rather than count it. This is human nature, not dishonesty.
- Count in pairs. One counts, one records. Scanning barcodes rather than reading labels removes the transcription errors that make a manual count untrustworthy by the third aisle.
- Count high-value items twice. The expensive lines are where an error costs you real money.
If your catalogue is large, count in ABC order: the 20 percent of SKUs that make up most of your stock value first and most carefully, the long tail after.
Step 3: import the fields that actually matter
Bulk import from CSV or Excel. The fields that earn their keep on day one:
| Field | Why it matters from day one |
|---|---|
| Product name | How staff find it at the counter |
| Barcode | Removes manual entry errors and speeds checkout |
| Cost price | Every margin and stock-value report depends on it |
| Selling price | Obvious, but check for stale prices while you are here |
| Opening quantity | The number from your physical count, not an estimate |
| Reorder level | Set it now or you will never come back and set it |
| Supplier | Turns a low-stock alert into an action you can take |
Reorder level is the one people skip. A low-stock alert with no threshold set never fires, and the single biggest benefit of the software quietly never arrives.
Step 4: run parallel for two weeks
Keep your old method running alongside the new system for a fortnight. Yes, it is double work. Do it anyway.
At the end of each day, compare a handful of fast-moving lines: what does the system say you sold, and what did you actually sell? Differences in the first week are almost always process, not software:
- Goods received but never booked in
- A multi-pack sold as a single unit
- Staff selling without ringing it through
- Returns handled informally
Every one of those is worth finding in week one, while you can still trace the cause. The same discrepancy discovered in December is unsolvable.
Step 5: lock it down and start counting cycles
Once the parallel fortnight is clean, switch the old method off completely. Running both indefinitely means neither is trusted.
Then start cycle counting: count a small rotating subset every week rather than closing for a marathon count twice a year. Your highest-value and fastest-moving lines get counted most often. Discrepancies found within days are traceable; discrepancies found in six months are just losses.
What to measure after 30 days
Once you have a month of real data, these four numbers tell you whether it is working:
- Shrinkage rate. Book value minus counted value, over book value. Retail typically runs 1 to 2 percent. Above that is a process problem.
- Stock turnover. Cost of goods sold over average inventory value. Track the direction more than the number.
- Stockout frequency. How often a fast-moving line hit zero. This is the figure that shows up as lost sales you never see.
- Dead stock value. Shillings sitting in items that have not moved. Almost always larger than owners expect.
The formulas, with worked Kenyan examples, are on our inventory management page.
Common failure modes
| What happens | Underlying cause | Fix |
|---|---|---|
| System and shelf never agree | Opening count was estimated | Recount properly; do not adjust your way out of it |
| Nobody uses the reorder alerts | Reorder levels never set | Set them per product during import |
| Stock figures quietly drift | Everyone can adjust stock | Restrict adjustment permissions by role |
| Reports look wrong | Mixed units in the catalogue | Standardise units before importing |
| Staff bypass the system | Checkout is slower than the old way | Scan barcodes; fix the till workflow first |
The honest timeline
A single shop with a few hundred SKUs: catalogue clean-up one afternoon, count one evening, import one hour, parallel running two weeks. Live and trusted inside a month.
A multi-branch business: the same again per branch, plus a decision about whether branches share one catalogue. They should — one catalogue, per-branch stock counts — but that is a discussion to have before you import, not after.
Nobody regrets spending an extra day on the opening count. Plenty of businesses regret not having.
Ready to move off the spreadsheet? Start a free sell.ke trial — bulk import your catalogue, set reorder levels, and the team will do the first import with you over WhatsApp.