Methodology
Every figure I publish ships with how it was measured. This page is the longer answer — the four rules a number must pass before it appears anywhere on this site, in a case study, or in a report to your board.
When the WhatsApp CRM beta beat email on merchant ad spend, the sample was a UK pilot — too small to claim a precise multiplier. So the site says >80×: a lower bound, not a headline. When the honest answer is a range or a floor, I publish the range or the floor.
Forecasts are labeled forecasts. The €25M is an FY26 model co-authored with a central operations team — it is not booked revenue, and I won’t pretend otherwise.
Holdout groups wherever volume allowed a clean control; tests read at significance when the sample reached it — never stopped early because the numbers looked flattering. Where a clean control wasn’t possible, the provenance note on the number says so — that’s what the ⓘ how this is measured disclosures are for.
And I read deltas, not absolutes, when the instrument is known to be biased: Apple’s Mail Privacy Protection inflates open rates, so 43.9% → 52.7% matters as a +20% shift on the same programs — never as “52.7% open rate”.
Opens, clicks, even conversions are diagnostics — useful for steering, never the verdict. The question that decides whether CRM is working is economic: lifetime value against acquisition cost, contribution margin, and the shape of the retention curve.
That’s why every engagement starts the same way: in the first two weeks we agree on 2–3 P&L-linked metrics — and those, not a vanity dashboard, decide whether I’m earning my retainer.
My marketing philosophy in one line: retention compounds; attention doesn’t. Fewer, more relevant messages, sent when the human is actually ready for them, beat any volume strategy on a long enough curve — and the curve is exactly what we measure.
A message that isn’t worth measuring isn’t worth sending. Honesty about results is the same discipline pointed at myself — it’s cheaper than churned trust, in marketing and in consulting alike.
What the function actually owns
“CRM” gets used to mean “the email tool.” The real function is five disciplines that have to move together — and on most teams they’re scattered, under-owned, or quietly missing. Here’s the whole surface, and what each one is for.
Which audiences, which lifecycle moments, which outcomes — tied to the P&L, not to a content calendar. The decision of what not to do lives here too.
The build: segments, triggers, journeys, templates, the QA before anything fires. Where strategy becomes something that actually runs — reliably, at scale, in every market at once.
The measurement layer — the joins from the send platform to what people did next, the holdouts, the P&L-linked metrics a board can read. If you can’t see it cleanly, it isn’t owned.
The rules of the road: frequency caps, priority between competing messages, brand and legal guardrails, a single source of truth for who can contact whom and how often. The discipline that keeps volume from eating trust.
When deliverability slips, a journey misfires, or a number stops making sense — the unglamorous work of finding the defect and fixing it at the root. On a 65M-user product, fixing one platform-level defect can outweigh a quarter of new campaigns — the leak is bigger than anything you’d build on top of it. The numbers and how they were measured →
Hire five people and these fragment. The reason the fractional model works is that one operator can hold all five — and the seams between them are where most of the value, and most of the leaks, actually are.
FIG · The closed loop — a score is not the point
01
Ask
tied to an event
02
Route
to a person
03
Resolve
by someone who can
04
Return
to the same customer
05
Feed back
reason becomes a fix
Most NPS programmes stop at step 01 and report the number. The loop is the product: step 05 is what turns a survey into a defect list, and step 04 is what makes the next answer honest.
Where a customer actually is
Before any send, I place the customer on a single spine — because the right message, and the right metric, are completely different at each stage.
A lead needs a reason to start; a new customer needs to reach first value fast; an early-lifecycle customer needs the habit to form; a mature one needs to be left alone unless you have something genuinely worth their attention; a churning one needs a reason rooted in their behavior, not a reflex discount. Same person, five different jobs. A program that sends everyone the same thing is, almost by definition, wrong four times out of five.
So the metric moves with the stage. Onboarding is measured on time-to-first-value, not opens. Mature is measured on retention and margin, not volume. Churn is measured against a holdout, never against “we sent the win-back, they came back.”
The discipline nobody budgets for
Most fatigue isn’t one bad campaign — it’s five well-meaning teams each sending “just one more,” with no one counting the total a person receives. Contact governance is the count: a hard cap on how often anyone gets contacted, a priority order when messages compete for the same slot, and a rule for who gets to override it.
Treated as a setting, it’s an afterthought. Treated as a metric — measured, capped, defended — it’s one of the highest-leverage things in the stack. The send you hold back is often worth more than the one you make.
See the numbers these rules produced —