What the work looks like, before there is a client to point at

Fictional local small business

A local business with a handful of staff, a point-of-sale system, a separate accounting package and a supplier spreadsheet that one person maintains. The owner works in the business daily and knows it well by feel. Turnover is healthy enough that nobody has felt the need to look closely at where it comes from.

The owner can say what the business took last month. They cannot say which products earned the money, which day of the week is genuinely worth staying open for, or whether a line that feels busy is actually profitable once the buying price is accounted for. The information exists — it is spread across three places that do not agree with each other, and assembling it takes an evening nobody has.

At month end, someone exports a sales summary from the point-of-sale system, opens the supplier spreadsheet and types the figures into a third spreadsheet by hand. Categories are grouped from memory and grouped slightly differently each month. The total is compared against the bank, and when it does not match, the difference is usually written off as a timing issue rather than investigated. The result is a number that is roughly right and cannot be broken down.

• Point-of-sale transaction export
• Accounting software sales and purchase summaries
• Supplier cost spreadsheet maintained by hand
• Opening hours and staffing roster

One reporting view that reconciles the three sources, with product categories defined once and applied consistently. Revenue can then be broken down by category or by day or by hour without anyone assembling it. Margin appears alongside revenue wherever the cost data supports it, and where it does not, the view says so rather than showing a figure that looks complete.

• Agree what each figure means before building anything — whether revenue includes tax, whether a sale counts when it is rung up or when it settles, then what happens to a refund
• Measure the current process honestly: time the month-end assembly with a clock rather than estimating it, and count the manual steps
• Reconcile the point-of-sale export against the accounting figures and identify every difference before trusting either
• Define product categories once and apply them to history so past months become comparable
• Build the reporting so it refreshes from the exports rather than from retyping
• Add checks that flag an anomaly before the report is read, rather than leaving the owner to notice
• Hand over written documentation of every calculation, so the refresh can be run without outside help

What would be measured — before the work starts, and again afterwards. No values appear here, because there is no engagement and therefore nothing to measure.

• Time to produce the month-end summary, timed before and after
• Number of manual steps between export and finished report
• Number of separate systems touched per reporting cycle
• Difference between the reported total and the accounting total
• Days after month end that the report becomes available
• Reporting errors found after distribution per cycle
• Share of revenue that can be attributed to a defined product category

Before anything resembling this could be published about a real business, all of the following would have to be true.

• The job is complete and the client has accepted the work
• A baseline was captured before anything changed — a measurement taken afterwards is an estimate
• Results are measured the same way before and after, over a stated period, with confounding factors identified
• The client gives written permission for the exact wording and chooses the level of attribution
• Any screenshot is recreated from synthetic data, or fully sanitized and reviewed by the client
• Any testimonial is written by the client, in their own words
• Every claim is entered in the claims register with its evidence
• Where the work is not arm's-length, the relationship is disclosed in the case study itself