The Scorecard Nobody Fills In
There is a supplier performance record in our database with three fields on it: on-time delivery rate, quality score, and pricing trend. They are the right three. Nothing in the product has ever written a value into any of them.
The verdict
An empty table with the right columns is a peculiar thing to find in your own product, and it tells you something honest about how software gets built: the modelling is the easy part and the part that gets finished. What it costs you is specific — if you buy in a market where supplier reliability varies enough to matter, we cannot tell you who is reliable, and your buyers are defending their choices from memory. Ask us for the numbers rather than the feature, and ask every other vendor the same. The table is easy to show. The populated one is the product.
While auditing our own procurement module for this series, we found a model we did not know was there. It is called supplier performance. It has a link to the supplier, and three measures: how often they deliver on time, a quality score, and the direction their pricing is moving.
Those are, more or less exactly, the three things anybody buying from anybody wants to know. Somebody thought about this properly.
And then it was never connected to anything. The class is imported into one controller and never referenced. No code creates a record, updates one, or reads one. No screen displays it. It has been sitting in the schema, complete and empty, waiting for a calculation that was never written.
It is not an unfinished feature. It is a finished data structure with no feature attached to either end.
What we do have, which is not nothing
It would be misleading to conclude from this that suppliers are unmanaged here. Several genuinely useful things are built and working.
Suppliers go through prequalification before they can be used, so eligibility is a decision somebody makes and records rather than an accident of who was available. A supplier can be blacklisted, and the blacklisting carries a reason, a date and the person who did it — which means an exclusion is defensible afterwards. Suppliers have a portal where they can see the requests, quotations and orders that concern them.
All of that is about status: may we buy from this supplier, and what have we sent them. None of it is about behaviour: what happened the last forty times we did.
The data exists. It has never been added up.
This is the frustrating part, and the reason the empty model is worth a post rather than a shrug. Almost everything those three fields would need is already recorded, on every order, going back as far as your history goes.
What each measure would need, and where it already sits
- On-time delivery needs the promised date and the actual receipt date. Both are on the purchase order today — a delivery date, a promised lead time, and receipts recorded against the order with their own timestamps.
- Delivery accuracy needs ordered quantity against received quantity. The two-way match already computes exactly this, per line, and keeps the result on the order.
- Price behaviour needs what you were quoted against what you were charged, over time. Quotations are recorded per request; orders carry the agreed price.
- Quality is the one genuinely missing input — an incoming quality result exists per consignment but is not tied back to the supplier who sent it.
- Split-delivery behaviour needs receipts per order, which the matching already sums. A supplier who routinely delivers in three parts costs you more than one who delivers once, and nothing captures that.
Four of those five are computable today from data already in the system. The measurement is not blocked by missing information. It is blocked by nobody having written the aggregation.
Why supplier measurement matters more where reliability varies more
If every supplier in your market delivers within a day of promising, a performance score is a curiosity — the variance it would measure is too small to act on, and you would be choosing on price anyway.
The value of the measurement rises with the spread. Where the same product can be sourced from suppliers whose delivery reliability differs by weeks rather than days, and where the cheapest quote and the most dependable supplier are frequently not the same company, the score is not a nice-to-have. It is the entire basis on which a buyer defends choosing the more expensive option — and without it, that defence is one person's recollection against a spreadsheet showing they paid more.
That is also the situation in which buyers are most often asked to justify themselves, which is what makes the missing record expensive rather than merely absent.
The supplier scorecard you should demand from any procurement system
Ours has the table and none of the numbers. Ask to see the numbers, not the screen.
On-time delivery rate, computed from promised against actual
Make them prove it: Ask for your worst supplier by this measure, ranked.
Fill rate — how much of what you ordered actually arrived
Make them prove it: Ask which supplier short-delivers most often.
Deliveries per order
Make them prove it: Ask who splits deliveries, and how often.
Price movement against quote
Make them prove it: Ask whether quoted and invoiced prices are compared at all.
Quality results attributed to the supplier
Make them prove it: Ask whether a failed inspection is traceable to who sent it.
The score visible when choosing between quotes
Make them prove it: Ask to see it on the comparison screen, not in a report.
The last row is the one that changes behaviour
A supplier score in a report is read once a quarter by somebody in finance. The same score shown next to the quotes at the moment somebody is choosing between them changes the decision. If you are evaluating any system on this, ask where the number appears rather than whether it is computed — those are different questions and only one of them affects what gets bought.
Two, and the first is largely arithmetic on data we already hold
The unusual thing about this gap is how little new recording it needs. Four of the five inputs are already captured on every order.
Compute the three measures that already have inputs
On-time delivery from promised against actual receipt, fill rate from the matching that already runs, and deliveries per order. Populate the record that exists rather than designing a new one, and backfill from history so the score is useful on the day it ships rather than in a year.
Put the score on the quotation comparison
Next to the prices, at the moment somebody is choosing. This is the smaller half of the work and the one that changes what gets bought — a score that lives only in a report informs an annual review, not a decision.
We would not build a weighted composite supplier rating out of the three. A single blended number invites arguing about the weights instead of looking at the behaviour, and the three raw measures are more useful to a buyer than any score derived from them.
Talk to us about supplier performanceWhat AWRA OpsHub does today
- Prequalification, so eligibility is decided and recorded before a supplier is used.
- Blacklisting with a reason, a date and the person responsible, which is an exclusion you can stand behind later.
- A supplier portal, where a vendor sees the requests, quotations and orders that concern them.
- Quotation comparison across suppliers for a given request, on price and per line.
- Receipt matching per order, which already computes the raw material for a fill-rate measure.
What it does not do
- No supplier performance measurement of any kind. The record exists in the database and has never held a value.
- No on-time delivery rate, though both dates needed to compute it are recorded.
- No fill rate, though the matching already compares ordered against received on every order.
- No price-movement tracking from quote to invoice.
- Quality results are not attributed to the supplier who sent the goods.
- No score on the comparison screen, so nothing informs the choice at the moment it is made.
Not ours, by choice
- We publish no lead-time norms or trade statistics for any market. What a reasonable delivery window is in your sector is your knowledge, not ours.
- The empty model is dead code, not a roadmap commitment. We are not implying it is nearly finished — nothing has been built at either end of it.
Bring us two years of purchase orders
From whatever system you use. We will compute on-time delivery and [fill rate](/glossary/fill-rate) by supplier by hand and show you the ranking — which tells you more about your buying than any demonstration of ours would.
Talk to us about suppliers