Day 51 · Phase H

They aren't refusing to recommend you. They can't think of anyone.

Themes: not unwilling, just unable to retrieve · the window is the weeks they can state a result · cut risk, don't add bounties · the score predicts nothing·4 moves · 4 diagrams
Most referrals don't die on willingness. They die on a retrieval problem. You hand "who would be a fit?" to a busy person, their mind goes blank, and you get back "sure, I'll keep you in mind" — after which nothing happens.
Word of mouth is not the natural byproduct of satisfaction. It is an act, and someone has to walk the last mile. Four moves turn "they seem happy" into an actual introduction: one, give a persona and a scene so you do the remembering for them; two, the window is the few weeks after they can state a result, not signing day; three, their worry is lending you their reputation, so cut cost rather than adding reward; four, that 0–10 score predicts nothing — demote it to a switch.
MOVE 01

"Keep me in mind if anyone needs it" is an open question they cannot answer A free-recall problem, handed to a busy person, that they will fail.

cued requestsfree recallchannel factors
The principle in one line
What you want isn't "a referral", it's a name surfacing in their head — and those two things differ by an order of magnitude. Agreeing costs nothing. Retrieving, with no cue at all, the one person among several hundred acquaintances who happens to be struggling with exactly this is something human memory is close to incapable of. So the answer you get is a sincere, entirely inert "sure, I'll keep an eye out". Turn the open question into a fill-in-the-blank: a persona concrete enough to match against, plus a scene they personally witnessed, and the name surfaces on its own.
The second row is where it leaks
A hundred customers who genuinely want to help. Two ways of asking. Illustrative, not measured data: the shape comes from the mechanism Open: "keep me in mind" Cued: persona + scene + ready-made words Willing to help 100 100 A name comes to mind breaks here you did it for them Judges them a fit persona judges for them Actually says something words ready to send Willingness is identical on both sides. The whole difference is row two. One gives a retrieval cue. The other gives only a category.
Template · turning the open question into a blank to fill

Don't ask: "if you come across anyone suitable, send them my way." They will agree, and mean it, and never think of anyone.

Do ask (persona + scene + singular): "You mentioned those cross-border sellers you work alongside — several of them still reconcile accounts by hand. Which one of them complained about it recently? Not a list, one person is enough."

Three parts, all required: one, a persona specific enough to match against ("cross-border sellers", not "anyone who needs it"); two, a scene they personally heard — that's a retrieval cue, not a filter; three, ask for exactly one, because one is a completable task and "some" isn't.

If they say "let me think", don't stop. Offer a second cue: "who was in the room the last time you were all complaining about it?" The name often arrives on the spot.

Once you have the name, remove the last cost too: "I'll write something and send it over — if it looks fine, just forward it."

Why it works · mechanism and evidence
Why it works

Mechanism: memory is cue-driven. Being stored (available) and being retrievable right now (accessible) are different things, and what separates them is having the right cue. You think you are testing willingness; you are testing retrieval, and uncued retrieval fails at an absurd rate.

  • Hard evidence (classic memory experiment, very stable effect) · cued beats uncued by a wide margin: Tulving & Pearlstone (1966, Journal of Verbal Learning and Verbal Behavior) had participants learn category-organised word lists, then recall them either freely or with the category names supplied. The cued group recalled substantially more, though both groups had learned exactly the same material — the difference wasn't storage, it was access. That is the entire gap between "send anyone suitable my way" and "which of those cross-border sellers complained about it". (The original used word lists; extending it to recalling people is a mechanism analogy.)
  • Hard evidence (classic field experiment) · what blocks action is usually the path, not the motive: Leventhal, Singer & Jones (1965, Journal of Personality and Social Psychology) showed students tetanus-vaccination messaging; cranking up the fear appeal barely moved actual vaccination rates. What raised uptake from roughly 3% to roughly 28% was a map of the campus health centre plus asking them to decide when they'd go. Small things that decide outcomes like this are called channel factors. A message they can forward as-is is your map.
  • Boundary: cueing fixes "can't remember", not "doesn't want to". If they don't actually rate you, the name you extract just buys you a cold call and costs them one awkward moment — which is why this step belongs after the window below.

Evidence strength: both are peer-reviewed classics and the direction is solid, but neither was run on sales referrals, so this is mechanism transfer. Don't read "3% → 28%" as a promise that your referrals will multiply tenfold.

MOVE 02

The window is the few weeks they can state a result in their own words — not signing day Ask when they have something to say, not when your calendar says so.

timingintention-behaviour gapsaying is believing
The principle in one line
Signing day is when they feel best and hold nothing. All they can say is "I just bought something, looks good", which is zero information to a friend and arguably negative — it sounds like vouching for something unverified. What they can pass on is never the product, it's a result they can retell. And the window closes at both ends: too early and they have nothing to say; too late and they no longer remember how bad the old way was, because a solved problem rapidly starts to feel like it never existed. So the moment to ask is right after they have said the result out loud themselves.
Both lines are only high for a few weeks
Ability to state a result rises; how much the old pain stings falls Illustrative, not measured data: the timeline stretches with your cycle, the shapes don't change the window How much the old problem still stings (motive) Can state a result in their own words (ability) Ask here they just named the result, and still recall the old pain Later: the problem feels like it never existed, no contrast left, only "it works fine" Signing day first result confirmed in their words 3 months 9 months Signing day is peak mood and zero material. Mood is not something they can pass on.
Template · say it right after their own sentence

First, make them say the result — don't summarise it for them. Ask: "compared with three months ago, what's actually different here?" Their sentence is the raw material for everything that follows. Yours is useless: the person who has to retell it is them.

Second, catch it on the spot: "can I quote that?" Take the smallest permission first, then ask for a person.

Third comes the fill-in-the-blank: "You just said reconciliation used to take two days and now takes half an hour. Who in your circle is still living the two-day version?"

Three trigger moments, all of them right after they've said something good, none of them a separately booked meeting: the call where the first result gets confirmed; any time they spontaneously thank you — the strongest signal and the most commonly wasted; and within 48 hours of them praising you publicly.

The line not to cross: don't ask before there's a result. At that point you aren't asking for a referral, you're asking them to vouch for something unproven — they'll be embarrassed later, and the embarrassment gets charged to you.

Why it works · mechanism and evidence
Why it works

Mechanism: "I'll introduce you" is an intention, and the gap between intention and action is systematic, not dishonesty. There are only two ways to narrow it — drive the cost of acting to near zero (previous move), or catch the moment when the trigger conditions happen to be complete (this one). On top of that, making them say the result out loud changes them: the version they voiced firms up their own judgement.

  • Hard evidence (meta-analysis of meta-analyses) · saying you'll do it and doing it are far apart: Sheeran (2002, European Review of Social Psychology) pooled multiple meta-analyses and estimated that intentions account for roughly 28% of the variance in behaviour, with large numbers of "intenders" never acting. So logging "said he'd refer someone" in your CRM as a lead is manufacturing a fake pipeline — what belongs in there is whether they did a specific thing for a specific person on a specific day.
  • Hard evidence (classic social-cognition experiment, repeatedly replicated) · saying it changes the speaker: Higgins & Rholes (1978, Journal of Experimental Social Psychology) had participants describe a person to an audience whose views were known; they tuned their wording to the audience — and crucially, their own subsequent memory and judgement shifted toward the version they had said. So getting a customer to put the result into a sentence isn't only harvesting material, it also makes them more certain of it. Summarising it for them produces no such effect. (The "shared reality" line of research has tested this repeatedly.)
  • Industry data (single firm, not a randomised experiment) · referred customers really are worth more: Schmitt, Skiera & Van den Bulte (2011, Journal of Marketing) compared real customer records from a German bank's referral programme and found referred customers were better on both contribution margin and retention, with long-run value higher by roughly 16%–25%. The direction is credible, but this is observational data from one bank, and how much of it comes from referrers simply picking the right people is hard to strip out. Don't treat it as a universal multiplier.

Boundary: this timing logic assumes you actually delivered something observable. In businesses where the value only shows up a year later (insurance, consulting), the window has to be manufactured — periodically put the baseline and the present side by side for them. As the customer-success issue put it: what works goes invisible.

MOVE 03

What stops them isn't the effort. It's lending you their reputation. Cut what it costs them; don't raise what it pays them.

reputational risksocial normswhen rewards backfire
The principle in one line
Making one introduction stakes their standing as a good judge in front of a friend. Deliver and they gain a little; blow it and they lose a real relationship — the ledger is inherently asymmetric, so the default answer is to do nothing. See that, and you know which side to push: cut their cost and risk rather than raise their reward. Most people instinctively add to the reward side (bounties, commissions), which neither offsets the risk nor stays neutral — it converts "doing a friend a favour" into "a piece of business", and people follow completely different rules in those two modes.
The two columns they weigh
The only column you can really move is the left one Illustrative: what they actually weigh before deciding to speak up What it costs them staking their own judgement remembering who, then judging fit cleaning up if you blow it looking like they're selling for you one bad call costs a real relationship What it pays them genuinely helping that friend being seen as the well-connected one you owe them one (sometimes) a bounty however high, it can't offset the left works: shrink the left common but weak: add to the right And adding money flips it from "a favour for a friend" to "a transaction" — different rules entirely.
Checklist · four sentences that cut their risk

1. Underwrite the worst case: "if it turns out not to be a fit, I'll say so myself and go away — I won't chase him." That one sentence addresses the thing they're actually worried about.

2. Make it reversible: "I'll write something you can forward; if he's not interested, forget it happened, and you don't have to follow up for me." A guarantee downgrades into a message.

3. Let them vouch for their experience, never for your capability: "just say it's been working for you and let him judge" is far lighter than "tell him it's great" — the second is a warranty in your name, the first is sharing an experience.

4. Close the loop, unprompted: after the meeting, tell them how it went, deal or no deal. This is the most skipped step and it decides whether there's ever a second one — people don't mind helping, they mind helping into silence, never learning whether they helped or created a mess.

Why it works · mechanism and evidence
Why it works

Mechanism: an introduction lives in social norms — favours, reciprocity, a long relationship. Put a price on it and it switches into market norms, where it's a one-off trade judged on whether the money is worth it. The switch is one-way and hard to reverse, because pricing something also hands over a reason to decline: not worth it.

  • Hard evidence (experiment contrasting the two norms directly) · a small payment is worse than none: Heyman & Ariely (2004, Psychological Science) had participants do the same tedious task: effort was high when it was framed as a favour with no payment, high again for a substantial payment, and lowest of all for a small cash payment — the money moved it out of the favour world into the marketplace, where the price wasn't worth it. That is exactly what "$30 for every successful intro" does: it prices a favour as a badly paid side job.
  • Hard evidence (field experiment, and the effect didn't reverse) · a price replaces the norm it displaced: Gneezy & Rustichini (2000, Journal of Legal Studies) introduced fines at Israeli daycare centres for parents who collected their children late; lateness went up, not down — the fine replaced "it's embarrassing to keep the staff waiting" with "lateness is a service I can buy". Worse, lateness did not return to its old level after the fine was removed. Price a referral and you may not get back to favour mode.
  • The honest counterweight: none of this means referral rewards never work — the bank study above was itself a study of a cash-incentivised referral programme, and it worked. The difference is relationship type: in weak-tie, already transactional settings (consumer acquisition, channel partners) money is a clean incentive; with the strong-tie customers you care most about, money is a contaminant.

Evidence strength: both are peer-reviewed classics and the mechanism is solid, but how widely "motivation crowding out" holds is genuinely contested (meta-analyses disagree). Don't take it as a law that paying always backfires — take it as a real risk worth designing around.

MOVE 04

That 0–10 score doesn't predict growth — its one real use is telling you who to ask today Demote the number to a switch. The score is worthless; the moment isn't.

net promoterfailed replicationwhat word of mouth really does
The principle in one line
Net Promoter asks how likely you are to recommend, counts 9–10 as promoters and 0–6 as detractors, and subtracts. Two problems: it measures an intention, not an act (the 28% gap from the previous move lands exactly here), and it is not better than conventional satisfaction measures at predicting growth — that's not an opinion, someone checked properly. So stop running meetings on the number. Demote it to a switch: whoever just scored 9 or 10 is the person to ask, specifically, today. The score has no value; that moment does.
When the flywheel won't turn, find the broken link
The four links of a referral flywheel — and how far the score actually reaches Illustrative: every link needs a person doing a specific thing, or it's just a poster ① a result worth mentioning ② they put it into a sentence ③ it becomes one real intro ④ referred customers ramp faster, win sooner referral flywheel stalled? find the broken link ②→③ breaks most often the score logs willingness to speak, not whether they ever spoke A survey only carries you to link two. With nobody owning link three, all you collect is good moods.
Checklist · rebuild the survey as a switch

1. The score triggers, it doesn't report. Every 9–10 gets contacted by a named person within 48 hours — with the fill-in-the-blank question, not a thank-you note. Past that, the energy is gone.

2. The real payload is the open follow-up. "What was the main reason?" The sentence they write is your next piece of copy, your next case-study headline, and the words they'll use when describing you later. The score is one digit; that sentence is raw material.

3. Never appraise a team on it. Once the number drives rewards, what you receive is a groomed number: hand-picked recipients, a heads-up call before the survey goes out. You get the performance version of whatever you measure.

4. Don't compute it on a small sample. A score built on thirty responses is almost entirely noise, amplified by response bias — the people who fill in surveys skew to the extremes. With few customers, count "how many real introductions happened this month" instead.

Why it works · mechanism and evidence
Why it works

Mechanism: a single-item intention scale measures how someone is willing to present themselves at the moment of asking. Three things sit between that and revenue — the intention-to-action gap, whether they ever spoke to a specific person, and whether that person bought. With any link unowned, a high score never turns into money.

  • Hard evidence (peer-reviewed, a direct test of the original claim) · it isn't a better predictor of growth: Reichheld (2003, Harvard Business Review) proposed Net Promoter as "the one number you need to grow"; Keiningham, Cooil, Andreassen & Aksoy (2007, Journal of Marketing) redid that test on longitudinal data and found no evidence that Net Promoter beat conventional satisfaction measures at predicting firm revenue growth. The strong original claim did not replicate. It isn't useless as a rough, consistently-worded gauge — but don't treat it as causal, predictive, or as an operating target.
  • Hard evidence (consumer research across countries and categories) · good news travels more, and matters more: East, Hammond & Lomax (2008, International Journal of Research in Marketing) systematically measured how often positive and negative word of mouth occur and how each affects brand choice, and the result runs opposite to industry folklore: positive word of mouth is more frequent than negative, and has the larger effect on choice probability ("an unhappy customer tells ten people" has no traceable source). Negative word of mouth still needs handling, but your real leverage is turning the good words that already exist into one specific introduction.
  • Why "just scored you a 10" is a genuine moment: they have just stated a position publicly, so a small specific request pointing the same way costs them almost no new decision — commitment and consistency, covered in the influence-principles issue. Boundary: this only holds when they actually rate you.

In fairness: criticising Net Promoter doesn't mean satisfaction is unimportant. The real problem is that a badly over-compressed opinion got promoted to predictor and performance metric. To forecast revenue, watch behaviour — renewals, expansion, and introductions that actually happened.

Your Day 51 Action

Sixty minutes, turning "get customers to spread the word" into four things you can check.

One (15 min) · Write your cue lines: pick the three types of person most likely to convert, and write one question each with a persona, a scene, and a request for exactly one name. Read it back — if they could answer from memory on the spot, it passes; if they'd first have to work out who counts as suitable, rewrite.

Two (15 min) · Write the forwardable message: under 120 words, third person, one concrete result, and not a single sentence that makes them warrant your capability. Then ask yourself: forwarded verbatim, would this make them look like they're selling? If yes, cut more.

Three (15 min) · Pick two customers who recently got a clear result, and send it today. Ask for the sentence first ("what's different versus three months ago"), then for the person.

Four (15 min) · Audit your existing survey and rewards: does anyone read the open-ended answer? Has every 9–10 been asked something specific within 48 hours? Is the reward an awkward small cash bounty? Turn it into a discount for the person being referred.

Boundary: all of this assumes you actually delivered a result. A referral machine without results just spends your reputation faster — one bad introduction costs you every future referral from that person.
Think It Through
1. I'm just starting out with almost no customers. Where would referrals even come from?
Early on you're not asking for a customer referral, you're asking for an introduction to one conversation — a far cheaper request: they don't have to vouch for your product, only that you won't waste twenty minutes of their friend's time. Same three steps: give a persona, give forwardable words, say explicitly that you'll walk away yourself if it isn't a fit.

What matters more is changing who you ask. The most effective early referrers are usually not customers but the people who deal with your target customers daily: other vendors serving the same group, community organisers, former colleagues. They've never bought from you, so they stake less reputation, yet they know who is currently struggling with the thing — which is the scarcest information there is.

And one resource nobody uses: the people who explicitly turned you down. They know you and have no stake. "We're not a fit this time — but who in your circle is stuck on this right now?" often gets an answer, because someone who has just declined one request is more likely to say yes to a much smaller one.
2. I can't bring myself to ask. It feels like it makes the relationship transactional.
That discomfort comes from filing the request under "taking". Test it: do you genuinely believe that friend of theirs will be better off knowing you? If yes, the introduction is a gift. If no, the discomfort is an accurate signal — the thing to fix is the product, not the script.

If you do believe it and still can't ask, three things cut the awkwardness right down: one, ask for one person, never a list — one reads as a favour, a batch reads as an assignment. Two, hand back the veto explicitly: "if it doesn't feel right, just say so — I won't push and won't raise it again." Three, close the loop: tell them how the meeting went, even if nothing came of it.

One more layer: you almost certainly underestimate how often people say yes. The networking issue covered this — people systematically under-predict whether others will help, because you feel your own awkwardness in asking vividly and never see the awkwardness of declining. You think you're imposing; they often just feel trusted.
3. Management wants a referral rewards programme. Should we, and how much?
First sort out which kind of relationship your referrals come from, because the answers are opposite. Weak-tie, transactional settings (consumer acquisition, channel, affiliates) — money is a clean, effective incentive; be generous and be explicit. Strong-tie customers — money is a contaminant: it prices a favour as a small piece of business, and hands over a reason to decline (not worth it).

If you do build one, three rules: One, put the benefit on the referred person ("your friend pays X less in year one"). The referrer is forwarding a discount, not collecting a commission — socially far safer, and they never have to explain whether they got paid. Two, either make it unmistakably large and treat it as business on both sides (a real revenue share, in a contract), or use no money at all — early access, a properly written case study about them, an introduction they'd value. The worst option is the awkward middle: small cash, which completes the switch from favour to transaction and sets a low price while it's at it. Three, whatever you give, settle it only after the referred person has actually got value, or you've bought a list rather than a recommendation.

One signal not to misread: if referral volume jumps and close rate collapses the moment rewards go live, that isn't a lead-quality problem — you've swapped referrers who recommend because it works for referrers who recommend because it pays.