Day 58 · Phase H

Deals rarely die inside a step. They die between two steps, where nobody caught them

Themes: the leak is at the seam, not in any stage · one deal from cold email to renewal · delivery is remembered by peak and end · judge the process before the outcome·4 moves · 4 diagrams
String every move so far into one chain and the weakest point isn't any single link — it's the joints between them. What follows walks one deal from the first cold email to renewal in month fourteen, joint by joint.
Four moves. One, the loss happens at the seams and it multiplies — every stage has to hand something specific to the next; two, walk a full example through seven stages and watch the exit conditions rather than the lines; three, closing only hands the risk to delivery, and what the customer keeps is the peak plus the ending; four, judge the process before the outcome, because the deals you won teach the worst lessons.
MOVE 01

The funnel doesn't leak inside a stage — it leaks in the handoff between them Six stages at 80% each do not add up to 80% end to end.

the handoff objectlosses multiplythe working-memory ceiling
The principle in one line
You debrief by asking which step you botched, but six stages at 80% each don't come out at 80% end to end — five handoffs losing a fifth apiece leave you a third of what went in. The variable that decides deals isn't performance inside any stage; it's whether each stage hands something specific into the hands of the next one.
Six stages and the five seams between them
Six stages: the loss happens between them, not inside them Hanging under each arrow is the one thing that actually has to be handed over. 01 Reach out the right person 02 Qualify worth chasing? 03 Discover pain priced in money 04 Decision chain meet who signs 05 Propose, close get a date 06 Deliver, renew they buy again their exact words which ICP box they tick a pain they can repeat who pays, who objects who · what · which day Five handoffs: the losses multiply, they don't add lose 20% each time 33% survives end to end lose 5% each time 77% survives Six stages at 80% each doesn't help. The difference isn't inside any stage — it's what fell out at the seams. Structuring the seam cuts the loss: Starmer et al. (2014, NEJM)
Checklist · five handoffs, one thing each

1. Reach out → qualify: hand over one specific line they said about their current state, not "they seemed interested". Keep the exact words; every later stage uses them.

2. Qualify → discover: hand over which criterion of your ideal customer profile they meet, and on what evidence. If you can't write it, the deal got in on a feeling.

3. Discover → decision chain: hand over one pain they can repeat back themselves, already converted into money or hours. If they can't repeat it, it doesn't exist.

4. Decision chain → proposal: hand over a list — who pays, who uses, who will object — with one line each on what they care about.

5. Proposal → close: hand over a next step naming a person, an action and a date. "Let's stay in touch" is not a next step.

How to run it: keep those five lines in one place (a CRM note is fine) and read them before every move. Where a line is blank, go back and fill it rather than advancing — advancing only carries the gap into a more expensive stage.

Why it works · mechanism and evidence
Why it works

Mechanism: each stage's input is the previous stage's output. Loss at the seam is the default state, not an accident — the brain holds only three or four chunks at once, so the customer detail you were sure you'd remember has been compressed into a vague impression two weeks later. Writing the handoff down as one sentence routes around the bottleneck. And the loss is lethal because it multiplies rather than adds: five factors of 0.8 leave 0.33.

  • Hard evidence (neuroscience, an EEG measure): Vogel & Machizawa (2004, Nature) measured visual working-memory capacity directly with contralateral delay activity: the signal saturates at three to four items, individual differences are stable, and the source localises near the intraparietal sulcus. Cowan (2001, Behavioral and Brain Sciences) reached the same "about four chunks" ceiling from behavioural evidence. The implication is blunt: hold handoffs in your head and anything past four is lost.
  • Hard evidence (large-sample pre-post): Starmer et al. (2014, NEJM) introduced a structured handoff bundle across nine children's hospitals and 10,740 admissions: medical errors fell from 24.5 to 18.8 per 100 admissions (about 23%) and preventable adverse events from 4.7 to 3.3 (about 30%), with no increase in the time handoffs took. Strength, honestly: a pre-post study in medicine, not a randomised trial in sales — what transfers is the mechanism, fixing the handoff into a few mandatory fields.
  • Classic but methodologically weak: Bartlett (1932) showed with serial reproduction that information is systematically simplified in retelling and assimilated into the reteller's existing frame. It doesn't meet modern standards; treat it as qualitative support for "spoken handoffs always deform", not as hard evidence.

Boundary: don't impose five handoffs on a small deal that closes in one call — that's manufacturing process for yourself. The value of this rises with cycle length and headcount; for a two-week deal with one decision-maker, one note is enough.

MOVE 02

One deal from cold email to renewal, watched at every seam Watch the exit conditions, not the lines.

a composite exampleexit conditionsgroup decisions
The principle in one line
What follows is a composite example built from common shapes — not a real company's post-mortem. The numbers illustrate proportions and aren't benchmarks. Watch each stage's exit condition — what counts as done — and the one thing it hands forward, not the wording.
Seven stages, six handoffs, fourteen months
One deal, cold email to renewal A composite example, used to show the seams and the proportions. Not a real post-mortem, and not a benchmark. Day 0 Reach out three-sentence email one specific observation Day 6 Qualify twenty-minute call budget and timing checked Day 19 Discover 45 minutes of questions pain priced at 380k / yr Day 34 Decision chain the signer in the room plus an internal champion Day 61 Proposal one-page business case they can repeat it back Day 88 Close a go-live date not: let us think about it Month 14 Renew, expand 120k → 310k first value in week 6 Seven stages, six handoffs, not one magic line — each stage's output is the next one's opening. The longest stall was the decision chain, fifteen days. That is where most deals stall too.
The seam that matters most · from "they relay it" to "you're in the room"

The stall above almost always comes down to the same thing: your contact offers to take it inside for you. Relaying always attenuates — they won't use your words, and they won't remember your three numbers. The answer isn't better material for them to carry. It's getting yourself into that room.

Say roughly this: "This has gone well, but I don't want you relaying it internally on my behalf — usually half gets lost, and it's the finance half. Could we get thirty minutes next week with whoever owns the budget? I'll bring one page covering three things: what the current setup costs per year, what switching saves, and how fast it shows. You don't have to present it. You just have to be there and say what you think."

Why it lands: it asks them to do less rather than more, and it names a duration and an agenda. To refuse, they have to give a reason — and that reason is the information you were missing: either they don't actually want to advance this, or the budget line runs through someone else.

Why it works · mechanism and evidence
Why it works

Mechanism: a deal stalls between stages not usually because the buyer doesn't want to advance, but because nobody owns the next thing down to a named person and a named day. Fixing each stage's closing move as "who, what, which day" converts an intention that needs willpower to start into an action triggered by a situational cue — one of the most reliable interventions in behavioural science.

  • Hard evidence (meta-analysis): Gollwitzer & Sheeran (2006) pooled 94 studies covering roughly 8,000 participants: writing a goal as an implementation intention — "if situation X arises, I will do Y" — produced an effect size of d ≈ 0.65, medium to large, on goal attainment. That is the hardest support behind "end every step with a concrete next step".
  • Neural level (weaker, directional): fMRI work suggests implementation intentions shift action control away from self-initiated prefrontal processes (around the frontal pole, BA10) toward more automatic, cue-triggered pathways (Gilbert et al. 2009). Strength, honestly: a small literature with limited replication — supporting context for the mechanism, not a firm conclusion.
  • Industry data (moderate strength, not peer-reviewed): vendor-side research reported in The Challenger Customer (Adamson & Dixon, 2015) put the average B2B buying group at about 6.8 people, with the probability of "buy nothing" rising as the group grows. Use it for order of magnitude, not as a constant — the sample and method aren't public.

Boundary: seven stages are for deals measured in months with several people involved. For consumer purchases or a single decision-maker, compress to three — qualify, make the value clear, ask for a decision. Forcing seven stages onto a short cycle exhausts you and makes the buyer feel processed rather than helped.

MOVE 03

Closing only hands the risk to delivery — and they'll remember two points The peak, and the ending. Duration barely registers.

peak-end rulereward prediction errortime to first value
The principle in one line
Whether a customer renews doesn't track the average of their year with you. It tracks the peak — the worst or best moment — plus how the last stretch felt. Which is why one extra promise at signing costs so much: it raises the baseline the whole year gets compared against, and what the brain records is actual minus expected.
How memory samples, and where the expectation line sits
One experience curve. Memory keeps two points. experience peak end Dashed = the average of the whole experience. It barely enters memory; duration hardly counts at all. One extra promise, and the same delivery turns negative. the extra promise puts expectation here a restrained promise stops here actual delivery: identical in both cases − + Peak and end: Kahneman et al. (1993, Psych Science); Redelmeier & Kahneman (1996, Pain); randomised follow-up, Redelmeier et al. (2003, Pain) Prediction error: Schultz, Dayan & Montague (1997, Science) — midbrain dopamine encodes actual minus expected, not absolute good or bad So over-promising costs twice: it wins nothing now, and pushes every later delivery into the zero-or-negative range.
Scene: the call sixty days before renewal
Two months left on the contract. Same task, two openings, two completely different conversations.
The usual opening

"Hi — the contract's coming up for renewal. When can we get some time to talk about it?"

The problem: this frames the topic as "do you want to pay again", which invites them to re-evaluate you from zero. And they have no evidence to hand, only impressions — and the impression that surfaces most easily is the two ugly weeks during rollout.

Putting the peak and the end back where you control them

"Two months to go, so I did a version of the year's accounts myself first. In week six you ran the whole process end to end for the first time, and you're currently saving roughly forty hours a month — that's exported from your own back end, not a number I invented; detail is attached. I'd like to hear two things from you: which stretch of this year was worst, and what you most want solved next year. Does Wednesday afternoon work?"

What changed: you replaced the peak with a checkable fact (week six), then put the worst stretch on the table yourself — pain that gets named out loud is far milder than the version they reconstruct alone. And the last line moves the topic from "renew or not" to "what are we solving next year".

Checklist · put the peak and the end where you control them

1. Define first value as a checkable event. Not go-live — the first time they actually get a result. Put it on line one of the delivery plan and agree to look at the numbers together on that day.

2. Warn them about the ugly stretch in advance, and roughly how long it lasts — data migration, process change, internal training. An expectation flattened beforehand doesn't get charged to you; the same explanation afterwards reads as an excuse.

3. Replace project sign-off with twenty minutes of reading the results out loud. Three numbers plus one sentence they said themselves. Twenty minutes, and it's the twenty minutes of the year they'll remember most clearly.

4. At signing, say one thing less than you want to. The extra promise doesn't make them keener to sign; it just lifts the baseline that the next twelve months get measured against.

Why it works · mechanism and evidence
Why it works

Mechanism: the brain doesn't integrate an experience, it samples it — the peak and the ending carry enormous weight, and how long it lasted barely counts at all ("duration neglect"). Meanwhile the reward system encodes prediction error rather than absolute value: better than expected fires, exactly as expected barely moves, worse is suppressed. Stack the two and the felt value of delivery is a difference, which is why over-promising costs twice — it sells nothing extra now, and pushes every later delivery toward zero or negative.

  • Hard evidence (experiment plus a randomised trial, aimed straight at repeat purchase): Kahneman, Fredrickson, Schreiber & Redelmeier (1993, Psychological Science) found people preferred to repeat a longer cold-water episode with a gentler ending; Redelmeier & Kahneman (1996, Pain) reproduced it in colonoscopy — extending the procedure with a less painful final stretch lowered remembered pain; and Redelmeier, Katz & Kahneman (2003, Pain) ran the randomised follow-up, in which the improved-ending group was more likely to return for a repeat procedure.
  • Hard evidence (neural mechanism): Schultz, Dayan & Montague (1997, Science) recorded midbrain dopamine neurons (ventral tegmental area and substantia nigra pars compacta) and established that their firing encodes reward prediction error — a burst when the outcome beats expectation, little change when it matches, suppression when it falls short. That's the physical basis of "delight is a difference".
  • Boundaries and disputes: peak-end holds most reliably for short experiences of clear intensity. Stretching it across a months-long B2B delivery shared by several people is a reasonable extrapolation, not a demonstrated result, and some studies fail to reproduce an ending effect in particular settings.

Boundary: don't read peak-end as licence to let the middle rot — a bad middle grows its own negative peak, which is exactly what memory grabs. And all of this only holds where you did deliver value; using experience design to paper over what the product didn't do buys you one renewal cycle at most.

MOVE 04

Judge the process first, the outcome second — the deals you won teach the worst lessons Debrief by outcome and what you learn is the shape of your luck.

hindsight biasoutcome biasstructured debriefs
The principle in one line
Debrief by outcome and what you learn is the shape of your luck. Go stage by stage first — was the exit condition met? — and only then bring in the result. Reverse the order and a win locks in the steps you skipped, which is precisely what wrecks the next deal.
The four boxes
The four boxes: judge the process first, the outcome second every exit condition met one condition missed WON LOST The box you can copy Write the whole run down as it happened. What copies is the sequence and the exit conditions, not your clever lines. The dangerous box Winning locks in the steps you skipped. Write down which external condition carried it. Next time it is gone. Change the bar, not the play Every condition met and still lost: either it should never have entered the pipeline, or it was luck. Fix the entry standard. The only box you study alone And the easiest to learn only the surface of. Name the stage whose exit condition failed, not that they found it expensive. Outcome bias: Baron & Hershey (1988, JPSP) — the same decision gets rated differently depending only on how it turned out.
Checklist · the one-page end-to-end debrief

Step one (no peeking at the result): for each of your stages write "exit condition met: yes / no / can't tell". "Can't tell" counts as no — you had no evidence at the time.

Step two: find the first "no" and date it. Most deals actually died a month earlier than you think.

Step three (now bring in the result): place the deal in one of the four boxes above and do what that box says.

Step four · three questions: (1) Next time, at that same stage, which single question could I ask earlier? (2) What in this deal was outside my control and shouldn't be booked to me? (3) If it was lost from the start, what's the earliest stage I could have seen that?

The part that makes it work: a debrief only means something if a prediction was written down first. When a deal enters the quote stage, write one line — "probability of closing within 30 days: __%, because ____" — and pull it out at debrief time. Without it you'll just read back a story you have already rewritten.

Why it works · mechanism and evidence
Why it works

Mechanism: once you know the outcome, the brain automatically rewrites what you thought at the time and makes the result look as though it was obvious. So a debrief that starts from the outcome can only learn the contaminated version. Judging the process first locks in the judgement before the contamination gets in — and counting "can't tell" as "no" blocks the same flattering rewrite.

  • Hard evidence (classic plus meta-analysis): Fischhoff (1975, JEP: Human Perception and Performance) established hindsight bias — once told the outcome, people systematically overestimate how predictable it was. Guilbault and colleagues (2004) pooled more than a hundred studies meta-analytically and found the effect reliably reproduced.
  • Hard evidence (outcome bias): Baron & Hershey (1988, JPSP) had participants rate decision quality: an identical decision process was rated significantly differently purely because the outcome differed. That is exactly why the order of the two questions cannot be reversed.
  • Hard evidence (meta-analysis, with conditions attached): Tannenbaum & Cerasoli (2013, Human Factors) pooled 46 samples: structured debriefs improved performance over controls by roughly a quarter, d ≈ 0.67 — with the effect concentrated in debriefs that were structured, focused on specific behaviours and conducted without blame. A loose round of "how did everyone feel" isn't in that effect.
  • Optional strengthener (single study): Mitchell, Russo & Pennington (1989) showed that prospective hindsight — assuming the event has already happened and reasoning backwards — significantly increases the number of correct reasons people generate. Klein's premortem is built on it. Strength, honestly: one study; don't treat the commonly quoted improvement figure as an exact number.

Boundary: debriefs cost real time, so don't run one per deal. Pick three kinds — the largest, the longest, and the ones that "just evaporated for no reason". The third yields the most information. And "without blame" isn't a pleasantry, it's a condition of the effect: once a debrief becomes an accountability hearing, people submit the version that protects them.

Your Day 58 Action

Ninety minutes to walk one real deal end to end.

1 (20 min) · Draw your own stages: don't use my seven. Write down how many stages your business actually has, one exit condition each — what counts as done. More than seven, merge some.

2 (25 min) · Fill in the handoffs: take a live deal and write, stage by stage, the one line this stage hands to the next. Wherever you can't write it, that's your leak — go fill it today rather than advancing the deal.

3 (25 min) · Debrief a closed deal: judge only the exit conditions first (yes / no / can't tell, where can't tell means no), find the first "no" and its date, then look at the result and place it in a box. Do this on one you won — that's where the skipped steps are already hardening.

4 (20 min) · Define one first value: pick an existing customer, define "the first time they actually get a result" as a checkable event on a specific date, and send a three-line email to align on it.

Boundary: all of this is built for deals measured in months with several people involved. If yours close in one call, compress to three stages — completeness for its own sake backfires, because heavy process makes buyers feel you're advancing your funnel rather than helping them decide.
Think It Through
1. My deals close in a single call. Does any of this apply to me?
The number of stages doesn't apply to you. The seams do. On a short cycle, seven stages compress into three: qualify (is this someone I should serve), make the value clear (can they repeat it back), ask for a decision. The seams are still there — they've just shrunk from two weeks apart to two sentences apart.

The seam on a short cycle looks like this: the second between finishing your questions and starting your pitch. Most people skip the handoff entirely here, summarise the need in their own words and launch. The fix is to say the handoff out loud: "Let me check I've got this right — the thing that's costing you most is X, and it's burning about Y a month. Is that it?" They say "yes, and one more thing", and you've just recovered the part that would otherwise have been lost.

The other seam is close to delivery. What you sell on one call often gets delivered by someone else. That line you gave on the phone — "we'll take care of that for you" — turns into the negative prediction error described above, two weeks later, unless it becomes a specific handoff note to whoever does the work.
2. Nobody reads my handoff notes and I work alone. Do I still write them?
Yes — but you're writing the wrong object. Nobody reads them because you're writing a record, everything that happened. Records genuinely don't get read, including by you.

A handoff isn't a record, it's the opening line of your next move. Five lines maximum, and every line has to be sayable out loud: their exact words, the number you converted the pain into, what each person on the list cares about, and the who / what / which day. The test is simple: if those five lines can't be read in thirty seconds before your next call and then spoken, they're written wrong.

Working alone makes it more necessary, for the reason in the mechanism above: working memory can't hold more than about four items, so what you retrieve two weeks later isn't the detail — it's an impression you have already reorganised. You aren't handing off to a colleague. You're handing off to the version of yourself in two weeks who has already misremembered it — and no colleague will be there to ask "what exactly did they say?"
3. When I ask lost customers why, the answer is always "budget". Now what?
"Budget" is the cheapest polite answer available, and the answer won't change until the question does. Don't ask "why didn't you choose us" — that asks them to appraise you, and everyone gives a version that doesn't sting.

Ask about the process instead of the reason: "How did this actually get decided internally? Roughly when did the direction firm up? At that point, which two things were being compared?" Process is far more honest than reason, because it asks for recall rather than judgement. You'll often hear the real moment — much earlier than you assumed, usually while you were still following up enthusiastically.

More importantly, look at your own evidence before you ask anyone. Pull out which stage the deal sat in longest and the date of the first "can't tell". Go in with that and you're asking a specific question, not inviting an open-ended confession. Strength, honestly: loss interviews aren't strong evidence in themselves — reconstructed memory and social politeness are both in there. Treat them as leads, not conclusions; your own stage-by-stage record is more reliable than their recollection.