TL;DR: A London B2B SaaS company with £3.8M ARR and 420 active accounts was monitoring client health manually across five disconnected tools. The customer success team of four held a 90-minute Monday meeting where each CSM reported on their 105 accounts from memory. Last year, 76 accounts churned. The team anticipated 31. The other 45 were surprises. £414,000 in revenue. The signals were there (declining usage, souring ticket sentiment, failed payments, dropping NPS). They were in five separate tools that nobody was aggregating. We built a four-stage agent that pulls signals from all five sources daily, calculates a weighted health score per account, alerts CSMs when accounts cross below threshold, and analyses churn patterns to refine scoring over time. Monday meeting: 90 minutes down to 20. Surprise cancellations in Q1: down 62%. Agent cost: £280/month.
The Monday Meeting
Monday morning, 9:30. Sophie's customer success team gathers in the meeting room. Four CSMs. 420 accounts between them. Roughly 105 each.
They go round the table.
"Brightfield Analytics has gone quiet. I haven't seen anyone log in for a couple of weeks. I should probably check in."
"Meridian HR raised a ticket last week about the API. Seemed frustrated. I'll follow up."
"I think Conway & Partners' renewal is coming up. Let me check the date."
This is the Monday morning bottleneck. Four people monitoring 420 accounts by checking five separate tools (Mixpanel for usage, Intercom for support tickets, Stripe for payment status, HubSpot for account details, Typeform for NPS and satisfaction surveys), noting signals that seem concerning, and reporting them weekly from memory. No aggregated view. No scoring. No alerts between meetings. If an account drifts from healthy to at-risk on a Tuesday afternoon, nobody knows until the following Monday, assuming the CSM remembers to mention it alongside the other 104 accounts in their portfolio.
Sometimes the first signal is the cancellation email.
Last year, 76 accounts churned. Sophie's team saw 31 of them coming. The other 45 were, in Sophie's words, "surprises."
Forty-five surprise cancellations at an average annual contract value of £9,200 each. £414,000 in revenue that left while four people were checking dashboards one at a time and attending a weekly meeting to report what they could recall.
The Company
B2B SaaS platform based in London. Thirty-four employees. £3.8M in annual recurring revenue across 420 active accounts, a mix of SMB and mid-market clients.
Sophie heads customer success. Her team of three CSMs handles everything from onboarding to renewal. Their tools are capable individually: Mixpanel tracks feature usage, login frequency, and session duration per account. Intercom captures ticket volume, response times, and conversation sentiment. Stripe holds payment status, contract dates, MRR per account, and failed payment flags. HubSpot stores account details, CSM assignment, renewal dates, and call notes. Typeform collects quarterly NPS scores and post-support CSAT ratings.
Five tools. Five sets of signals. Zero connection between them. The system for synthesising all five into an account health picture is Sophie's team, checking each tool separately, holding the patterns in their heads, and delivering the synthesis once per week at the Monday meeting.
The cost of this arrangement, stated plainly: the CS team spends an estimated 35-40% of their time on monitoring work (dashboard checking, ticket reading, NPS review, internal reporting). Four people at an average loaded cost of £55,000 each. That's £77,000 to £88,000 per year in manual health monitoring. The Monday meeting alone: 90 minutes, four people, 52 weeks, 312 hours per year of reporting from memory.
The 18% annual churn rate cost the company £699,200 in lost revenue. Sophie estimated 40-50% of churned accounts left for preventable reasons (not budget cuts or business closures, but unaddressed frustrations, declining engagement that nobody caught in time, or renewal conversations that started too late). Saveable revenue: £280,000 to £350,000 per year. Sitting in five tools. Visible only to whoever had time to check.
What the Monday Meeting Missed
The meeting's structural problem wasn't effort. Sophie and her CSMs care about their accounts. The problem was scale versus memory.
A CSM responsible for 105 accounts checks what she can during the week. She opens Mixpanel, scans for declining logins. Opens Intercom, reads recent tickets. Opens HubSpot, checks upcoming renewals. She notices some patterns. She misses others. The account that logged in three times this month instead of its usual twelve? She might spot it. The account whose support ticket sentiment shifted from neutral to frustrated across three conversations over six weeks? Unlikely. That pattern exists across 200 messages. Nobody reads 200 messages looking for sentiment shifts.
The 45 surprise cancellations shared a profile. In retrospect (and retrospect was all they had), most showed declining usage 60-90 days before cancellation. Most had increased support ticket volume. Several had failed payments that were resolved but never followed up on. A few had submitted NPS scores that dropped from 8 to 5 over two quarters.
Every signal existed. In one of five tools. Visible to whoever logged into that specific tool and looked at that specific account at the right time. The Monday meeting captured whatever fraction of these signals a human brain retained from a week of scattered dashboard checks across 105 accounts. Sophie estimated the meeting surfaced about 40% of the at-risk accounts. The other 60% remained in the tools, unconnected, until the cancellation arrived.
The design is next. The design, the build guide, the running costs and the take-away are free to subscribers. Enter your email below and the rest of this Blueprint opens. New ones arrive every Tuesday and Friday.
