Management reporting

What this problem is

Management reporting is the repeating work of assembling figures the leadership team uses to run the business: pipeline, utilisation, cash, service load, delivery. The problem is rarely “we need a dashboard”. It is that the numbers are rebuilt by hand, late, and argued over.

Is this your problem?

• Someone spends the end of the week copying exports into a pack
• Two directors bring different totals for the same metric
• Definitions (“active client”, “qualified lead”) change by meeting
• The pack arrives after the decision it was meant to support
• Fixes live in a personal spreadsheet that nobody else can refresh

Current workflow

A manager exports CRM, accounting, and a delivery tracker. They reconcile mismatches in a spreadsheet, paste charts into slides, and email a PDF. Questions in the meeting send them back to the exports. Next week they start again. The spreadsheet is the real model; the systems are inputs.

Where it fails

• Metrics are not defined, so every rebuild is a negotiation
• Source systems disagree and there is no owner for the difference
• Access is personal: if that person is off, the pack does not exist
• Manual steps mean the pack cannot be produced mid-week
• Dashboards exist but nobody trusts them, so the spreadsheet continues

How to measure the baseline

Time the weekly pack: hours spent, number of source systems, number of manual reconciliations, and how often a figure is later corrected. Use human_hours_per_week, minutes_per_case (one pack = one case), error_rate_pct on sampled figures, and wait_time_hours from period-end to pack sent. Do not count “dashboard logins” as success if leaders still use the spreadsheet.

Metrics that matter

• Human hours per week — Total staff time spent on this workflow in a typical week. (hours/week)
• Minutes per case — Average handling time from pickup to completion of one case. (minutes)
• Error rate — Share of cases that contain a defect requiring correction. (%)
• Wait time — Elapsed time a case sits idle between steps. (hours)
• Manual touches per case — How many separate human interventions a typical case requires. (touches)

Root causes

• No metric dictionary: the same word means three things
• Operational data is not captured in systems during the work
• Reporting is treated as a communications task, not a data product
• Tools were implemented without agreeing grain (job, client, week)
• Leadership asks for new cuts every week, so the pack never stabilises

What should be automated

• A short list of defined metrics with a single source each
• Scheduled extracts or live queries instead of copy-paste
• A reconciling check that flags when sources disagree beyond a threshold
• A pack or view that refreshes without rebuilding slides from scratch
• An audit of who changed a definition

What should remain human-controlled

• Choosing which metrics matter this quarter
• Explaining a variance — automation should not invent a narrative
• Decisions that follow from the numbers
• Adding a new metric only when the definition and source are agreed

Example intervention architecture

Illustrative only — not a client result. A metric dictionary names each figure, grain, owner, and system. Pipelines pull from CRM, finance, and delivery on a schedule. A check compares totals that should match. The output is a stable pack or a simple view, not a warehouse project. Humans still write the commentary. If a source is missing, the pack shows that it is missing rather than silently using last week.

Failure modes

• A dashboard nobody opens because the numbers feel wrong
• Automating a bad definition, faster
• A data project that takes longer than the reporting pain it was meant to remove
• Access that leaks personal or client-sensitive detail
• Weekly scope creep until the pack is a second business

When not to automate

• Leaders cannot agree the five numbers they would actually use
• Source systems are so incomplete that any automated figure would be fiction
• The pack is a one-off board exercise, not a repeating operating rhythm
• A spreadsheet already takes under an hour and is trusted
• The organisation will not stop asking for ad-hoc new cuts every week

Related problems

• Invoice intake

Sources

• Corporate reporting (FRC)
• Code of Practice for Statistics (UKSA) — quality and trustworthiness principles