Where a business loses money after the sale.
This business could tell you what it sold. It could not tell you which customers had actually got value, which were about to ask for their money back, or why. Six departments, eight systems, and no shared view of a single customer. I built the surface that answers all three questions — and keeps answering them without anyone maintaining it. Refunds came down by £500k.
The problem
This business was losing more in the months after a customer paid than in the weeks before — and had no instrument pointed at any of it. Nobody could say which customers had reached value, and nobody owned the question.
No 360° view of a customer. Answering "who is this, what did they buy, did they show up, are they at risk?" meant a human opening five tabs and stitching it together by hand — every morning, for every customer worth checking.
Refund prevention was manual and late. A large share of paying customers never logged in at all — the loudest possible refund signal, and nobody was systematically catching it. Refund "saves" were tracked on self-report, not verified against what the payment system actually showed.
Retention was unknowable. Leadership could not reliably say whether the refund rate was improving, whether onboarding was fast enough, or whether last month's product changes helped or hurt.
The trackers rotted. The spreadsheets that ran daily work went stale the moment someone stopped updating them — bloated with resolved rows and duplicates, and quietly trusted anyway.
Quality control didn't scale. Nobody could listen to every sales and coaching call, so delivery quality was spot-checked at best — and the calls that preceded a refund were never reviewed at all.
What I built
One operating surface that is both the single source of truth and the daily worklist — built around the customer lifecycle rather than around the tools.
A live value ladder. Every paying customer sits in exactly one state — never logged in, logged in but no value yet, reached value, or refunded / left — with a refund-risk score from a grounded, audited model rather than a guess. One number, the same everywhere it appears.
An activation funnel that names the leak. Every milestone from payment to delivered outcome, expressed as a share of paying customers — so the single biggest drop-off stage is obvious at a glance instead of buried three tools deep. The surface says where value is lost, not just that it is.
A post-mortem engine for every loss. When a customer refunds, the system reconciles every source — payments, product usage, calls booked and actually held, community access, email engagement — grades the calls they were on, and produces a causation verdict with direct links to the recordings. "Why did we lose this one?" became a question with an answer.
Department operating tabs. Leadership scorecard, onboarding, community, delivery ops, refund desk, call review — each mirroring how that team actually works, each with its own live worklist, and each separating what to do from why this is here.
Graders on autopilot. Every call scored for effectiveness and sentiment, hourly. Community replies audited against the real message history. SDR roundups generated rather than compiled.
Self-maintaining trackers, and push instead of pull. Trackers seed themselves from live signals, advance status automatically and archive resolved rows. Daily digests fire into the right Slack channel — booking misses to the SDR, onboarding lag to the onboarding lead, refund saves to the save desk — so the day's priorities land where people already are.
How it works
The point was never the dashboard. Dashboards rot. The point was an operating model that stays current without a person in the loop.
A surface that rebuilds itself. Scheduled jobs pull live data from every system every hour, re-run the scoring, regenerate the pages and redeploy. The header reads "live" with the current timestamp because staying current is the machine's job, not a person's.
One scoring model, reused everywhere. Refund risk, customer status and KPI logic live in shared single-source-of-truth modules — so the number a rep sees on a customer card is the number leadership sees on the scorecard. No two versions of the truth to argue about in a meeting.
Verified, not vibes. Wherever a claim could be self-reported — "we saved this refund", "the team replied" — the system checks it against the authoritative source instead of trusting the label. That check alone caught real cases where a "saved" customer had in fact been refunded.
Resilient by design. Every data pull falls back to a committed snapshot if a live key is missing, so one flaky API never blanks out the whole picture.
What it changed
Refunds fell by £500k. Customers who never logged in, went inactive, or crossed a risk threshold now surface into a save queue automatically — days before they would have quietly churned. That is the number the whole build was for: not a dashboard metric, money that stayed in the business.
The day runs off a live worklist. Each team opens one tab and sees exactly who needs attention and why, instead of reconstructing it from five systems every morning.
Leadership manages by real numbers. A live KPI and SLA scorecard, plus a deployment-impact analysis mapping a month of product releases to their measured effects — separating the one corroborated win from the changes that were still too early to call.
Quality control at full coverage. Every call graded, so delivery quality is measured rather than sampled — and a post-mortem can point at the exact call that contributed to a loss.
The team trusts the picture. Because it refreshes itself and verifies its own claims, people act on it instead of second-guessing it. That is the entire point of a single source of truth.
The win wasn't a dashboard. It was collapsing eight disconnected systems and six manual tracker cultures into one operating surface that scores itself hourly — one that names where customers stop finding value, explains why a specific customer left, and pushes the right slice to the right person before it turns into a refund. The business went from reconstructing what happened to acting on it, and from guessing whether things were working to measuring it. The refund line moved by £500k.
If revenue is growing and margin isn't, the leak is usually after the sale.
Thirty minutes. No brief required. We'll work out where your customer lifecycle is losing money and whether it's a problem I can fix — and if it isn't a fit, I'll tell you inside the first ten minutes.