The pattern we most often see when we look inside SME reporting isn't the absence of data. It's the presence of too much of it — wired to nothing. A wall of metrics nobody's acting on, produced on a cadence nobody set, reviewed in meetings nobody prepares for. This case study is one firm's version of that pattern, and what changed when they stopped adding and started subtracting.
The firm is a small UK professional-services practice — ten staff, around £1.2M revenue, owner-operated, four-year-old business with a mix of retainer and project work. Details anonymised; structure and sequence accurate.
Where they started
An 18-metric management dashboard built in Power BI a year earlier by a part-time analyst who had since left. The dashboard was technically live and technically accurate. It was also technically ignored. The owner admitted — after some prodding — that she hadn't opened it in six weeks. The monthly management meeting used a separate three-page PDF produced manually in Excel by the operations lead.
Stated problem: "We have all this data and it doesn't help us make decisions." Real problem: the data never had a chance to help, because nobody had ever wired it to a decision.
What the Audit Found
Rather than starting from what to build, we started from the 5-step BI audit. Forty minutes with the owner and the operations lead. Findings were both predictable and uncomfortable:
- 12 of the 18 metrics on the dashboard had no defined decision attached. They were there because they were trackable, not because anyone had agreed what to do if they moved.
- Three metrics had inconsistent definitions. "Utilisation" meant one thing in the monthly PDF (billable hours ÷ contracted hours) and another on the dashboard (billable hours ÷ 40, regardless of contract). This had been the source of at least two heated meetings.
- Two major blind spots existed. The firm couldn't answer, from data, which client segment was most profitable after fully-loaded staff cost, or which project types were most likely to scope-creep. Both were first-order questions for the business.
- The monthly meeting reviewed the PDF, not the dashboard. The dashboard and the meeting had, in effect, been running parallel for a year, with no connection between them.
The audit surfaced what it almost always surfaces: a reporting portfolio that had accumulated rather than been designed.
The Intervention: Six Metrics, Four Loops
We didn't build anything new in the first four weeks. We subtracted. The 18 metrics went to six. Not because the other twelve were wrong — because they had no decision-linkage, and keeping them was paying a visibility tax for zero return.
The six that stayed:
- Billable utilisation — properly defined this time, one source of truth.
- Gross margin by client segment — closing the "which clients are actually profitable" blind spot.
- Project scope-variance — planned hours vs actual hours per project, the second blind spot.
- Cash runway (weeks) — already on the old dashboard, but now with a threshold.
- Pipeline value — next 90 days — replaced a lagging bookings metric.
- NPS (quarterly) — the one leading indicator the firm already had data for and had never used.
Each of the six went through the four-part decision loop: metric, threshold, conversation, action. Four of the six had loops defined; the other two (pipeline and NPS) were too new for meaningful thresholds and were flagged for review after 90 days of observation.
Metric: Billable utilisation = billed hours ÷ contracted hours, per consultant, weekly, rolling four-week average.
Threshold: Amber if any individual drops below 60% for two consecutive weeks; Red if any individual drops below 55% in any week, or if the firm average drops below 65% for two weeks.
Conversation: Operations lead reviews every Monday. Amber or Red is raised in the Thursday partner sync that same week.
Action: Amber triggers a same-week conversation with the individual (workload, pipeline, flagging). Red triggers reallocation of existing work or an immediate push on business development for that consultant's practice area.
What Changed in 8 Weeks
The behavioural shifts were specific and small — which is usually the shape of a working BI rollout. No dramatic turnarounds, because dramatic turnarounds in eight weeks are almost always someone lying.
Monthly management meeting averaged 90 minutes, mostly spent debating whether the numbers in the PDF agreed with the dashboard.
Monthly meeting at 45 minutes, most of it on decisions rather than on reconciliation. Both artefacts retired; a single six-metric view replaced them.
Utilisation conversations happened quarterly, usually after the problem had already cost billing capacity.
Weekly. Three conversations triggered by Amber thresholds in the first eight weeks — two resolved by reallocation, one by a targeted BD push. None escalated to Red.
The firm assumed its most profitable client segment was its largest by revenue.
Margin-by-segment analysis showed the largest segment was the second-most profitable after fully-loaded costs. The most profitable segment was a smaller one that the firm had been underselling. Pricing on two retainers in that segment was revised at renewal.
Scope-creep was handled reactively. The firm wrote off approximately 11% of project hours unbilled across the previous year.
The scope-variance metric surfaced three in-flight projects trending over scope in weeks 3, 5, and 7. Each was re-baselined with the client before the overrun compounded. Full-year write-off projection down to ~6% (signal still settling).
What Didn't Change
Two things we'd caveat clearly. First, NPS had three data points in the eight-week window, which is not enough to say anything meaningful about satisfaction trend — it's on the dashboard and ready to be acted on when the signal firms up, but claiming a NPS lift at week eight would be overstating. Second, cash runway didn't move, and wasn't expected to — the metric was there for threshold-alerting, not improvement.
And the firm isn't done. The dashboard was simplified; the habits started forming; the decision loops ran for the first time. But the Rhythm stage of the Method is a quarter-long discipline, not a two-month one. The case study stops at the point where the wiring works — it doesn't claim the system is finished.
What This Cost
Measured in external spend: nothing. No consultant fees, no new platform, no data-staff hire. The existing Power BI dashboard got rebuilt in-house in a weekend by the operations lead, using the six-metric scope. The definitional work happened in two one-hour sessions with the owner.
Measured in internal time: approximately 14 hours of the owner's time across eight weeks (setup, definitions, the weekly 15-minute check, the monthly review), and about 4 hours per week of the operations lead's time on the new rhythm (which mostly replaced time previously spent producing the retired PDF).
Measured in framework: the firm now has a six-metric view with defined loops, an SSOT protocol, and a weekly rhythm that — if maintained — will outlast any individual who implemented it. That was the point.
The dashboard that replaced the old one is half the size. That's the win. Subtraction was the intervention.
The Pattern This Illustrates
What this firm had before wasn't unusual. It was normal — normal in the sense that the majority of SMEs who have invested in reporting are in the same shape. More metrics than decisions, more reports than readers, more dashboards than discipline. The intervention wasn't sophisticated; it was structural. And that's usually what the work is.
For the contrast, the earlier lag vs. lead indicator case study shows the same principle from a different angle: not pattern-of-work around subtraction and loops, but pattern-of-work around the leading-vs-lagging problem that most reporting portfolios get wrong. Same underlying thesis — BI is built out of structure, not technology.