Spare Parts Inventory: 4,000 Parts, 10,000 Names
Kirinyaga Road runs on memory. Ten thousand part numbers, four naming systems for the same alternator, and a counter man who is the only person who knows where anything is. What actually fixes that, in what order — and the vehicle record this system does not have.
A spare parts business is the hardest inventory problem in Kenya, and it is hard for a reason that has nothing to do with volume. A hardware shop selling ten thousand items has ten thousand distinct things. A parts dealer selling ten thousand items has perhaps four thousand distinct things wearing ten thousand names — the OEM number, the aftermarket equivalent, the number stamped on the box, the number the mechanic asks for, and the shorthand the counter man uses.
Everything difficult about this trade follows from that. Stock figures cannot be trusted because the same part is counted under three codes. Reordering is guesswork because demand is split across duplicates. And the business is genuinely dependent on one or two people who hold the mapping in their heads — which is a commercial risk nobody has priced.
The one decision that fixes most of it
Pick a primary identifier and make every other name an alias of it. Not a synonym list in someone's head — a structural decision, applied once, in a week of unglamorous work.
-
Choose the primary code, and choose it for the counter, not for the catalogue
Usually the OEM number where you deal in one or two makes, and the aftermarket number where you deal in many. The test is which one your best-selling supplier prints on the box, because that is the code that will appear on the goods arriving.
-
Put every other name in searchable fields on the same record
The barcode field for the number on the box, and custom fields on the item for the OEM equivalent, the aftermarket cross-reference, and the shorthand people actually ask for. One record, many ways in. The custom fields also flow into your import templates automatically, so this is loadable rather than a manual second pass.
-
Merge the duplicates before you count
The count is the moment your figures become true. Counting three records for one part gives you three true figures for a thing that has one quantity, which is worse than one wrong figure because it looks precise.
-
Name in a fixed order, every time
Component, then make or application, then the distinguishing detail: "Alternator, Toyota 1KZ, 24V". Not "24V Toyota alternator 1KZ", which sorts nowhere useful and is unsearchable by anyone who does not already know the answer.
-
Then bin-locate everything
A part with a location is a part any member of staff can find. A part without one is a part only the counter man can find, which is where the dependency comes from — and it is fixed by shelf labels rather than by software.
The business is not dependent on one person because he is irreplaceable. It is dependent on him because the mapping between four naming systems exists only in his memory, and nobody has ever asked him to write it down.
What a counted, coded parts store looks like
A 6,000-line parts business, first count
The merge is the part with no visible payback and the largest real one. Demand split across three duplicate records looks like three slow-moving parts; consolidated, it is one fast-moving part you have been under-ordering for years. That is invisible before the merge and obvious afterwards.
Serial numbers, and where they genuinely earn their place
Most parts do not need serial tracking. Four categories do, and in this trade they are the ones with warranty exposure: batteries, tyres, alternators and starters, and anything electronic.
For those, serial-level records carry the supplier and purchase order they arrived on, when they were received, and a warranty expiry date. Which converts the warranty conversation from an argument into a lookup: the customer brings a battery back, you read the serial, and you know when it was sold, who supplied it and whether the claim is live. Serial and warranty tracking covers this properly, and it is the single highest-value capability in this cluster for a dealer who sells batteries.
What is built for this trade, and what is not
What AWRA OpsHub does today
- Items with barcodes plus custom fields you define, so OEM, aftermarket and box numbers all live on one record and are all searchable.
- Bin locations within warehouses, so a part has a place and anyone can find it.
- Serial-level records for the parts that need them, carrying supplier, purchase order, received date and warranty expiry.
- Batch and lot tracking with expiry, which matters for oils, filters and anything with a shelf life.
- Landed cost on imports, allocated across a shipment by value or by quantity.
- Two-way matching at receipt — what was ordered against what arrived — with an over-receipt guard at the door.
- Counter sales through POS with availability enforced before the sale, so you cannot sell what is not there.
- Customer accounts with credit limits, for the garages who buy on account.
What it does not do
- No vehicle record. No vehicle, no registration number, no VIN, no make-model-year entity. A parts business can hold vehicle information as custom fields on a customer or a job, and there is nothing to hang a vehicle history on.
- No fitment or compatibility engine. Nothing answers "will this fit a 2014 Probox". Cross-references are searchable text you maintain, not a supported relationship.
- No work orders, so no job card as a first-class record. A workshop job is modelled as a project or a task, which works and is not the same thing.
- No parts catalogue data. Nobody supplies you a TecDoc-style reference — every code and cross-reference is one you enter.
- No exchange or core-return tracking for reconditioned units.
- Landed cost allocates by value or quantity only, with no weight or volume basis, which mis-spreads freight on a mixed container of heavy and light parts.
The first two are the ones to weigh seriously. If your business runs on fitment lookup — if a customer arrives with a registration number and expects you to find the part — then the honest position is that this system holds your stock, your money and your customers accurately and does not answer that question. Fitment and compatibility sets out what can be built around the gap and what cannot.
The order to do this in
First
Merge duplicates and fix naming
One week, before anything is loaded. Nothing downstream is worth doing on a duplicated item list, and this is the only step no software and no vendor can do for you.
Second
Bin-locate and count
Shelf labels, then a blind count. This is what removes the one-person dependency, and it is physical work rather than software work.
Third
Counter sales and customer accounts
Availability enforced at the point of sale, and credit limits on the garages who owe you money. The commercial payback starts here.
Fourth
Serials on the four warranty categories
Batteries, tyres, rotating electrics, electronics. Not everything — the discipline collapses if you try to serialise a bolt.
Fifth
Buying against reorder points that now mean something
Because demand is no longer split across duplicates. Wait sixty days for consumption history before trusting any reorder point you typed at setup.
Our take
Spend the first week merging duplicates and fixing names — it is the whole game in this trade and no system will do it for you. Then bin-locate and count blind, which is what ends the dependency on one person's memory. Serialise only the four warranty categories. And be clear-eyed about fitment: if your business is a lookup service with a store attached, this holds the store honestly and does not answer the lookup.
Fix the item list first, then the system
One record per part with every name searchable, bin locations so anyone can find it, serials with warranty dates where they matter, and availability enforced at the counter — with the fitment gap stated plainly rather than discovered.
See plans & pricingFrequently asked questions
What makes spare parts inventory harder than other retail stock?
The same part carries several names — the OEM number, the aftermarket equivalent, the number on the box, and the shorthand the counter man uses. So a store with ten thousand lines usually has around four thousand distinct parts wearing ten thousand codes. Stock figures cannot be trusted because one part is counted under three records, reordering is guesswork because demand is split across duplicates, and the business depends on whoever holds the mapping in their head.
What is the first thing to fix?
Merge duplicates and settle a naming convention, before anything is loaded and before you count. Pick a primary code — usually the number your best-selling supplier prints on the box — and put every other name in searchable fields on the same record, using custom fields for the OEM, aftermarket and shorthand versions. It is a week of unglamorous work and it is the only step nobody else can do for you.
Can the system tell us whether a part fits a specific vehicle?
No. There is no vehicle, registration, VIN or make-model-year entity and no fitment or compatibility engine, so nothing answers "will this fit a 2014 Probox". Cross-references are searchable text you maintain rather than a supported relationship. If your business is essentially a fitment lookup with a store attached, that is a limitation worth weighing before anything else.
Which parts should be serial-tracked?
The four categories with warranty exposure: batteries, tyres, rotating electrics such as alternators and starters, and anything electronic. Serial records carry the supplier and purchase order they arrived on, the received date and a warranty expiry — which turns a warranty argument into a lookup. Do not serialise everything; the discipline collapses the moment somebody is asked to serialise a bolt.
How much dead stock does a parts business typically hold?
Eight to twenty per cent of stock value in parts that have not moved in twelve months, and it is invisible before the first count because dead stock is defined by the absence of a movement. For a KES 9 million store that is KES 900,000 to 1.8 million, of which perhaps a third is realistically recoverable through discounting, supplier returns and simply stopping the reorder.
Can we run workshop job cards?
Not as a first-class record — there is no work order entity. A workshop job is modelled as a project or a task, which genuinely works for costing parts and labour against a job and is not the same thing as a job-card system. What that model does well and where it stops is worked through in the job cards and parts-versus-labour posts in this cluster.
