Research commentary · Food systems
The delivery that never made the ledger.
A farm can weigh every crate it harvests and still not know whether it kept its promises. A pickup log can be accurate too. It will tell you who collected food, provided there was something to collect and someone there to record it. The household whose box was never packed cannot sign for it. That is precisely the household a measure of reliability needs to count.
Start with the promise
Suppose a programme promises a weekly box of vegetables and eggs to 40 eligible households. Set out the contents and collection window before the week begins. At the end, 35 households have a confirmed complete box, two received partial boxes, two are known to have missed out and one has no surviving record. A receipt-only ledger can make the 35 complete receipts look like 35 successes out of 35. The promise gives a different denominator: 35 of 40, or 87.5%, were verified complete. If the unrecorded household did receive its complete box, the number could be 36 of 40, or 90%. Neither treating the unknown as a known miss nor dropping it from the count is warranted.
The two partial boxes count as partial under the declared rule. If the programme wants to distinguish a missing carton of eggs from a missing week of food, it can publish both measures. What it cannot do is change the rule after the fact and call that an improvement in delivery.
Find out where delivery broke
The remedy depends on where the record stops. If the crop never came in, look at production. If the box was packed but no offer reached the household, look at notification. If the offer went out but the box could not be collected, inspect the collection window and transport. And if there is no evidence after packing, preserve the uncertainty rather than choosing whichever explanation suits the report. Those distinctions are useful in a meeting because each suggests a different change for next week.
Put the costs on the same footing. If the illustrative week cost £700 across all 40 commitments, that is £17.50 per commitment. Dividing by the 35 verified completions gives £20 per verified complete box, including the costs incurred for unsuccessful or unresolved commitments. This second number is not the cost of producing one more box. It answers what the programme spent in total relative to the service it can show it delivered.
What the pilot does and does not show
In my published Methods paper, the household-week ties production, distribution, loss, labour, cost and water records to a service obligation fixed in advance. The partial pilot shows what can be learned from 18 weeks of production records. It did not record household deliveries or the complete costs needed to calculate the figures above. A better harvest is worth reporting; it is not, by itself, evidence that households received a reliable service, that their food security improved, or that the programme saved money.
The practical change is small. Open next week’s ledger with a row for every household you have promised to serve. When a delivery falls through, or the evidence goes missing, leave the row in. The blank is part of what you need to explain.