What this looks like in a real business.
The pages before this one describe a method. This one shows it working. Below is a full worked example — six domains, one dimensional model underneath them, and an agent on each that decomposes a move before it interrupts anybody. Every figure is illustrative; the structure is not.
Daily brief · all markets · illustrative
Find the bad cohort while you can still act on it
Know a cohort has gone bad in six weeks, not six months.
First payment default is the earliest honest signal you have that something changed in underwriting or acquisition. Everything else about a vintage is confirmation.
- FPD 5 and FPD 30, number and amount based
- Acceptance rate against FPD, by cohort
- Vintage default curves by issue month
- Decline reason distribution and rule contribution
- Score population stability and characteristic shift
- Override rate and override performance
- Watches
- FPD 5 by cohort, market and acquisition source, daily
- Speaks when
- A cohort crosses three points above its trailing eight-week baseline
- First, though
- Decomposes the move across score band, decline-rule change, source mix and product before it says anything
- And says
- Names the cohort, the driver and the size of the contribution, with a link to the rows
Read the deep dive: First payment default, and the agent that watches it
Cumulative default by issue-month cohort
Months on book across, issue month down. Illustrative figures.
Aim contact effort at the accounts that will actually cure
Every euro of contact effort aimed at the accounts that will actually cure.
The recovery funnel is where a business either protects its receivables or spends money chasing people who were going to pay anyway. Self-cure is the number nobody wants to measure.
- Delinquent → right party contact → promise to pay → promise kept
- Kept promise rate and PTP quality by agent
- Aging distribution across DPD buckets
- Forward roll rate against cure rate
- Self-cure versus intervention mix
- Reminder channel efficiency: SMS, email, call
- Amount owed against amount collected, principal and interest split
- Watches
- Kept promise rate, forward roll and the self-cure share, by market and agent
- Speaks when
- Kept promise rate drops while contact volume holds, or self-cure share rises above the contact cost line
- First, though
- Separates a collections performance problem from a book quality problem, which look identical in the headline
- And says
- Which segment stopped curing, whether calls are still paying for themselves, and where to move capacity
Read the deep dive: The recovery funnel, and the number nobody wants to measure
Recovery funnel, 1–30 DPD
Stage counts and conversion off the previous stage. Illustrative figures.
Know you will miss the month in week two
Find out you will miss the month in week two, not in week five.
Issued against a pro-rata forecast, split by new and top-up, is the fastest read on whether the commercial plan is real. Most teams look at it monthly, which is three weeks too late.
- Issued volume against pro-rata forecast, by market
- New versus top-up and drawdown split
- Application status distribution: active, declined, cancelled, expired
- Counteroffer take-up and its effect on the book
- Average ticket, term and yield at issue
- Portfolio size and 0-bucket principal outstanding
- Watches
- Daily issuance against pro-rata forecast at market and product grain
- Speaks when
- Projected month-end lands outside a tolerance band you set, not a fixed percentage
- First, though
- Attributes the gap between demand, acceptance and conversion so the fix goes to the right team
- And says
- The projected close, the three biggest contributors to the gap, and what changed on the day it turned
Read the deep dive: Sales pacing, and why the monthly forecast review is three weeks too late
Issued against forecast, by market
Pro-rata to date, indexed so 100 is on plan. Illustrative figures.
Rank channels on the book they buy, not the leads
Stop paying more for the customers who default.
Cost per acquisition ranks channels one way and contribution per cohort ranks them another. The gap between those two orderings is usually the largest unclaimed margin in the business.
- Issuance and sales by UTM source, affiliate and broker
- Cost per funded loan against contribution per cohort
- FPD 30 and default by acquisition source
- Broker lead flow: seen, reached, duplicate, ruleset pass, funded
- Duplicate and lost-lead leakage by partner
- Decline value by rule, by source
- Watches
- Source-level issuance, cost and cohort default as each cohort seasons
- Speaks when
- A source moves in the contribution ranking, or its FPD drifts from its own baseline
- First, though
- Waits for enough seasoning before it speaks, which is the part humans get wrong under pressure
- And says
- Which channel is buying a worse book than its CPA suggests, with the cohort evidence attached
Read the deep dive: Acquisition source quality, or why your cheapest channel is your most expensive
Share of spend, share of issued, and default by source
Three views of the same six channels. Illustrative figures.
Catch a broken step the morning it breaks
See a broken step the morning it breaks, not in next month’s review.
Onboarding drop-off is where growth quietly leaks. A third-party identity check that starts timing out costs more in a week than most pricing decisions save in a quarter.
- Step funnel: attempts, completions, drop-off, cumulative
- Average time spent per step, by device and market
- Attempted against completed over time
- Bank connection and identity verification success rates
- OTP delivery and success by provider
- Repeat versus new applicant behaviour
- Watches
- Per-step completion and time-on-step against a seasonal baseline, hourly
- Speaks when
- A step moves beyond its own historical variance — not a global threshold, because steps differ
- First, though
- Correlates against release timeline, device, market and third-party provider before raising anything
- And says
- The step, the size of the leak in funded loans, and the change most likely to have caused it
Onboarding funnel, share of starters reaching each step
Drop-off at each step shown against the bar. Illustrative figures.
One set of numbers finance and the funder both recognise
One set of numbers that finance, the board and the funder all recognise.
Cash in, cash out, yield and outstanding principal — reconciled to the ledger every month, with the gap reported as a standing line item rather than explained when somebody notices it.
- Total incoming and outgoing, by month and market
- Revenue and yield on average outstanding principal
- Principal outstanding, active, closed and issued loans
- Repayment curve by client cohort, day 0 to day 720
- Reconciliation to the general ledger, by named component
- Funder and covenant reporting pack
- Watches
- Daily cash movement against the schedule, and the ledger reconciliation gap by component
- Speaks when
- A reconciliation component moves outside its normal range, or yield diverges from the book mix
- First, though
- Names the component rather than the total, because a total tells finance nothing they can act on
- And says
- Which component moved, by how much, and whether it is timing or something real
Read the deep dive: Cash, yield and the reconciliation nobody wants to build
Cash in, cash out and yield
Twelve months, in millions, with yield on the right axis. Illustrative figures.
Different nouns, identical structure
Every business has a way of acquiring demand, converting it, delivering it, getting paid for it and losing some of it along the way. The domains do not change. Only the vocabulary does.
Subscription and SaaS
MRR pacing against plan, activation and onboarding funnels, retention and expansion cohorts, CAC payback by channel, failed-payment recovery, and net revenue reconciled to the ledger.
Marketplaces
Supply and demand balance by category, liquidity and fill rate, take rate by segment, seller cohort quality, fraud and abuse, and payout reconciliation.
Insurance
Quote-to-bind funnel, loss ratio by underwriting cohort, claims cycle time, fraud triage, renewal retention, and reserve movement against the ledger.
Retail and e-commerce
Trading against plan by channel, basket and margin mix, checkout funnel, return-rate cohorts by source, availability, and promotional lift measured properly.
Operations and service
Demand against capacity, first-contact resolution, backlog aging, SLA breach risk, cost to serve by segment, and the leading indicators of each.
B2B services
Pipeline pacing and coverage, utilisation and bench, project margin by engagement, delivery risk, and days sales outstanding by client cohort.
If your business is not on this list, the first call is the cheap way to find out whether the pattern holds for it. Occasionally it does not, and we will say so.
One write-up per metric
Each of these takes a single KPI, explains what it means commercially, how to define it so it survives an argument, the model underneath it, and exactly what the agent on it does.
First payment default, and the agent that watches it
Of all the numbers a lender reports, first payment default is the one that arrives soonest and hurts most. It is also the one most often looked at monthly, in a…
ReadThe recovery funnel, and the number nobody wants to measure
Collections is the part of a lending business where effort and outcome are least correlated, and where almost everybody measures the wrong thing. The headline number…
ReadSales pacing, and why the monthly forecast review is three weeks too late
Every lender has a monthly sales number and a monthly forecast. Most compare them at the end of the month, in a meeting, where the only available action is an…
ReadAcquisition source quality, or why your cheapest channel is your most expensive
Every marketing review ranks channels by cost per acquisition. It is the wrong ranking, and almost everybody knows it, and almost nobody changes it — because the…
ReadOnboarding drop-off, and the identity check that costs more than a pricing decision
Growth leaks quietly through onboarding. Nobody complains, because the people affected simply leave. There is no ticket, no angry email and no spike on any dashboard…
ReadCash, yield and the reconciliation nobody wants to build
Cash is the one part of a lending business where being approximately right is not a position anybody will accept. It is also, in most lenders we meet, the number…
ReadThe KPI framework we build for a consumer lender in the first 30 days
Every consumer lender we meet has a dashboard. Most of them report the same three things: applications, approvals and arrears. Those are the numbers that are easy to…
ReadPutting AI in a credit decision without losing the argument later
There is a conversation that happens about eighteen months after an AI system goes into a lending decision. Somebody — a supervisor, an ombudsman, a claimant's…
ReadCredit risk reporting that survives the questions after it
There is a moment in every funder diligence process where someone asks how a number was produced. Not whether it is right — how it was produced. Which system it came…
ReadDatabricks or Snowflake for a lender's data platform
We get asked this most weeks, and the honest first answer is that either will work. A mid-sized consumer lender with a few million rows a day is nowhere near the…
ReadWhich number would you want watched first?
Pick one. We build it as one working thread — source data, dimensional model, agent, guardrail, brief — and put it in production before we talk about anything else.
Usually a reply the same working day · IST (UTC+5:30)