What 625 checked deliveries can tell us.

What counts as a delivery problem?

A fully checked delivery has a recorded problem when at least one item is marked missing or damaged, or its recorded actual quantity is lower than its expected quantity. Each delivery counts once, even if several items have issues.

The benchmark describes what was recorded in Unpickle. It does not measure money saved, credits received or the financial value of the affected goods.

How we calculated the benchmark

Unpickle verified the aggregate against the source account on 9 October 2026 at 14:26 BST, using read-only queries. Individual delivery records are not published. The café is Matté Coffee + Chill, which shares its story, with the owner’s permission, in our case study.

1. Scope the records to one authorised shop. Retrieve all deliveries and delivery items for that shop. Order each query by a stable record ID, fetch in pages of 500 and request exact row counts.
2. Check that the extraction is complete. The delivery query returned 628 rows across 2 pages. The item query returned 2,795 rows across 6 pages. Both fetched totals matched their exact counts.
3. Include only fully checked deliveries. A delivery must contain at least one item, and every item must have status ok, missing or damaged. All 625 included deliveries also had a recorded check timestamp for every item. The 3 deliveries with pending items are excluded.
4. Count distinct deliveries with a problem. Include a delivery if any item is missing or damaged, or if a non-null actual_quantity is less than expected_quantity. Count that delivery once, regardless of the number of matching items or issue types.
5. Calculate the completion-date range. For each included delivery, use the latest item checked_at timestamp. The earliest and latest of these delivery completion timestamps define the displayed period. The first individual item check was on 21 October 2025; that delivery was not fully checked until later.
6. Divide problem deliveries by fully checked deliveries. Use 72 ÷ 625 = 0.1152 = 11.52%. The denominator excludes incomplete deliveries. Estimates use the unrounded ratio.

What was recorded?

These categories overlap: a delivery can appear in more than one row. They must not be added together. The distinct total remains 72 problem deliveries.

What this evidence cannot tell us

• How often every café has delivery problems. This is one venue’s experience, not a representative sample of cafés, suppliers or the hospitality industry.
• How much money was recovered. No prices, credit amounts or refund outcomes were retrieved or used. A recorded problem does not guarantee a refund or credit.
• Whether Unpickle caused a financial improvement. There is no before-and-after comparison or control group, and no measured causal effect.
• Every problem ever found at this café. These are current item records, not an immutable audit history. Later edits can change statuses, quantities and check timestamps. The date range describes the currently recorded completion dates.
• A transactionally frozen database snapshot. The source queries and pages were read consecutively. Both fetched totals matched their exact counts, but the extraction did not hold a single database snapshot transaction.

How this relates to the savings calculator

The Unpickle delivery savings calculator applies this 11.52% observed rate to the monthly delivery volume you enter. It then multiplies the estimated problem deliveries by your assumed recoverable value per problem delivery and deducts the subscription cost for the net estimate.

The editable £10 starting value is an illustrative assumption, not a measured result from this café. The calculator is a way to explore an estimate; it does not predict guaranteed savings for your shop.