Case study
August 2026
Delivered
What the pipeline was actually doing.
Seven views over one CRM, built to answer questions the reports could not. The first one it answered: reported wins had tripled in six years, and the number the team had actually sold had not moved at all.
- Views over one CRM
- 7
- Companies analysed
- 44K
- Growth in reported wins
- 3x
- Growth in worked wins
- 0%
Context
Five pipelines, sixteen years of history, and roughly 44,000 company records that had been through a CRM migration. Standard reporting could describe any single slice of that and could not describe the shape of it.
The questions leadership had were the ones reports are worst at: is the growth real, which accounts have we quietly lost, and where should the next quarter of effort go.
The finding
A deal that closes within two days of being created was not sold in the CRM. It was entered after the sale, or it was a renewal, or it was an order someone typed up. That is a legitimate thing to record. It is not the same activity as working a deal, and counting them together makes a sales team look like it is growing when it may not be.
Splitting the two apart is one date comparison. The result reframed the whole conversation:
- Reported wins
- 238 728
- Worked wins
- 161 161
- Recorded wins
- 15 398
2019 → 2025
unchanged
all the growth
Every unit of reported growth over six years came from recording, not from selling. Nobody had been hiding it. The two categories had simply never been separated, so the number everyone quoted was measuring two different activities at once.
Seven views
One tool, one visual language. Two chart primitives and one figure strip across all seven views, and colour reserved for exactly one meaning: how long something has sat.
01
Stage health
Where open deals are parked, and how long they have sat there
02
Movement
Won and lost over time, with the pipeline cutover marked
03
Funnel
Outbound touches through meetings, opportunities, and wins
04
Worked vs recorded
Which wins were actually sold and which were entered after the fact
05
Territory
Where the revenue actually is, by state and by year
06
Accounts
Penetration, lapse, and the win-back list
07
Rep load
Open pipeline by owner, against their assignment on the account record
Drive it
The real tool, running here. Move between the seven views, filter to a single product line, and drag the year slider to change the window. Worked vs recorded is the one to open first. Product lines appear as Core, Line B, and Line C; rep names are pseudonyms. Every figure is the real measured value.
Loading the live artifact…
What else it surfaced
A win-back list that was mostly unreachable. 1,284 accounts had not bought since before 2024. Only 202 of them had been contacted in the previous four months. 743 were cold and 178 had no logged contact ever. That turns a vague retention worry into a routing decision about which 202 to work first.
A system change masquerading as a business change. Wins appeared to collapse in Q1 and recover across new pipelines. The business had not changed in January; deal creation had moved. The view marks the cutover so nobody reads the artifact as a trend.
Volume and value pointing in opposite directions. One state carried more open deals than another, 91 against 73, worth $183K against $1.49M. Deal count alone would have sent effort to the wrong place.
A denominator worth trusting. Of roughly 44,000 company records, 4,171 had ever had a deal and 2,089 had ever bought. Penetration numbers are usually hand-waved because the account list is dirty. A cleanup against the public district registry is what made this one defensible.
Saying what it cannot say
The tool states its own limits in the places a reader would otherwise over-trust it. Deal counts are reliable everywhere, but open value is understated by an unknown amount because a large share of deals carry no value at all. Loss reason is blank on 6,319 closed-lost records, so the reason breakdown describes about 1,700 stated reasons rather than the population. 32,861 companies have no contact date. Outbound email barely exists before the migration, so the funnel view refuses to start earlier than 2024 no matter where the slider is set.
None of that weakens the analysis. It is the reason the strong claims in it survive being argued with.
Outcome
The worked-versus-recorded split changed how growth was discussed internally, because it separated a real operational question from a reporting artifact that had been flattering the numbers for years.
Built in React as a single component, delivered as a self-contained snapshot rather than a live dashboard. That was deliberate. This is an argument with evidence attached, meant to be read in a room and disagreed with, not a surface somebody has to keep maintaining.