Guide
Inventory management for a small business
Most small-business inventory advice is written for warehouses with a planning team. This is written for a business where the person setting up the system is also the person serving customers.
Do these in order
The sequence matters more than any individual step. Nearly every failed rollout skipped ahead to the interesting parts on top of a foundation that was not there.
Step 1: fix the catalogue
One row per genuinely distinct sellable item. If a customer can order the 500ml and the 1-litre separately, they are two SKUs. If you sell the same thing under two names, that is one SKU with one name.
- Pick a SKU format and stick to it. Short, meaningful, and never reused.
SHM-500-LAVbeats1042for a human and works just as well for a machine. - Record the barcode separately from the SKU. The barcode belongs to the manufacturer; the SKU is yours. Keeping both lets you bulk-update from either your data or a supplier's.
- Categorise once, properly. Every report you will ever want groups by category. A flat catalogue of 800 uncategorised products cannot be analysed.
- Record cost as well as price. Without cost there is no margin report, and margin is the reason to do any of this.
- Delete what you no longer sell. Or archive it. Carrying discontinued lines in the active catalogue slows every screen and every count.
This is unglamorous work and it is where the value is. A tidy catalogue with mediocre processes beats a sophisticated system on a catalogue nobody trusts.
Step 2: one honest count
Everything downstream inherits this number. Some rules:
- Count what is there, not what should be there. The temptation to adjust toward the book figure is overwhelming and completely destructive.
- Close, or freeze movement, while you count. Counting a moving shelf produces a number that was never true.
- Two people on high-value items. One counts, one records.
- Write down the variances. The gap between book and physical is your first real measurement of shrinkage, and you will want the baseline later.
- Do not investigate every discrepancy. Investigate the expensive ones. Absorb the rest and move on.
Step 3: record movements as they happen
This is the whole ball game, and it is a habit change rather than a software problem. Stock moves at the counter, at the loading bay, in the van — not at a desk at 6pm.
- Make recording take seconds. If it takes a minute, it will not happen when the shop is busy, and busy is exactly when the movements matter.
- Put it on the phone. The person who knows what moved is holding a phone, not sitting at the office desktop.
- Record reasons, not just quantities. "Minus 3" is useless in a month. "Minus 3, damaged in transit" is a supplier conversation.
- Never let anyone overtype a quantity. Corrections should be adjustments with reasons, so the history survives. This single rule is what separates a stock system from a spreadsheet.
Expect this step to take a few weeks to bed in, and expect the first month's data to be imperfect. That is normal and it is still worth far more than no data.
Step 4: set the numbers that matter
Only now do you have enough history to set anything sensibly.
- Run ABC analysis. Twenty minutes in a spreadsheet.
- For A items, calculate a proper reorder point at a 98% service level.
- For everything else, use lead time demand plus a few days of cover.
- Load the reorder point as each product's low-stock threshold.
- Check turnover by category and look at what is not moving at all.
Nothing here needs a consultant, and none of it needs more than a spreadsheet plus the calculators on this site.
Multiple people, one truth
The moment two people touch stock, three requirements appear that a shared spreadsheet cannot meet:
- Simultaneous access without overwriting each other.
- Attribution — every change carrying the name of whoever made it.
- Permissions — a shop assistant recording sales without being able to alter costs or delete products.
Being able to answer "who changed this, and when?" removes an entire category of recurring workplace argument. It is worth more than most feature lists suggest.
Why implementations fail
| Failure | What actually happened | The fix |
|---|---|---|
| "The system is always wrong" | Movements are not being recorded at the moment they happen | Make recording fast and mobile; audit for a fortnight |
| Nobody uses it but the owner | Staff have no permissions, or the tool is desk-bound | Give roles and put it on phones |
| Abandoned after two months | Everything was migrated at once, including 300 dead SKUs | Start with the products that actually move |
| Reports look wrong | Opening count was adjusted to match the books | Recount honestly; accept the variance |
| Alerts are ignored | One global low-stock threshold, so alerts are meaningless | Per-product reorder points |
| Data stuck in the old sheet | No import path, so it was retyped and abandoned halfway | Bulk import from Excel, matched on SKU |
Notice how few of these are software problems. The common thread is a process that asks people to do something inconvenient at the moment stock actually moves.
If you want to see what this looks like implemented, SmartShelfKart's feature list maps directly onto these steps — catalogue and Excel import, stock takes, typed movements with reasons, per-product thresholds, roles, and reports derived from the movement record. It is free on every tier during the launch period.