Chapter 8
Chaos in the Pipeline
Sixty-One Days
The Post-it note says DALLAS — proposal? and Marcus does not know what it means.
He knows it's his handwriting. He knows it's about a mid-size lender whose head of ops had a good conversation with him about servicing transfers. He does not know whether he sent the proposal or only meant to, or when he wrote the note, because at the time he wrote it he knew.
That was three weeks ago. There are eleven other notes on the monitor.
Forty-three unread emails. Fifteen are live conversations. Three are follow-ups he owed people last week. One is Karen Walsh, asking whether he can move Thursday. One is Mike Donovan, four lines, acknowledging the assessment and asking to circle back in a few weeks — which is either enterprise politeness or the beginning of something, and Marcus can't tell which, because he can't remember exactly what he sent Mike or when.
The calendar is worse. Discovery calls stacked against each other. A demo for a lender in Colorado. And in three weeks, forty-five minutes in a ballroom in Fort Worth that he agreed to in four minutes and has been carefully not thinking about since.
David stops at the desk and looks at the monitor.
"Busy?"
"Busy month."
"That's not what I'm looking at." He nods at the notes. "What's that one?"
"I don't know. That's the problem."
David pulls up a chair, which he doesn't usually do.
"This is the opposite of a system," he says. "This is chaos with good intentions."
"It's working, though. Four months ago I had nothing."
"Right, and now you have twenty conversations and a wall of adhesive paper." David leans back. "Do you remember year one, when you wanted me to ship the officer app in six weeks and I said no?"
"You said we'd build it twice."
"I said no technical debt. Build it right or build it twice, and building it twice costs four times." He points at the monitor. "So explain to me why you're building sales debt."
Marcus opens his mouth and finds there's nothing in it.
Sales debt. He's been so pleased with himself for generating opportunities that he never built anything to hold them. Every I'll organize this later, every conversation living in his head instead of somewhere he can look — all of it borrowed, and the interest is showing up as a yellow square that says DALLAS.
"I'm dropping things," he says.
"You're going to drop Karen." David says it flatly, which is worse than saying it hard. "Karen isn't a deal. Karen is the proof that any of this works."
Sixty-one days left on Rachel's second ninety, and one signature to get.
"So what do I do?"
"Same thing we did with the product." David stands. "Stop adding. Fix the foundation."
Refactor
Marcus calls Tom that night and opens with "I think I'm failing," which Tom apparently finds funny.
"How many live conversations?"
"Fifteen. Maybe twenty. I don't know exactly, which is the answer to your next question."
"And in February?"
"Zero."
"So you've gone from no pipeline to more pipeline than you can hold. That's a good problem, and people lose companies to good problems every year." A pause. "What does David do when the code gets past what he can hold?"
"Refactors."
"So refactor. Map what you actually have, then build something to hold it." Another pause. "This weekend, not next month. Sales debt comes due on somebody else's schedule."
"Karen's."
"Karen's."
Marcus writes that on a Post-it, and recognizes even while he's doing it that it ought to be the last one.
The Machine Is Not the CRM
The software is not the system. The division of labor is the system. A CRM is a filing cabinet with opinions. What makes it worth anything is deciding, explicitly and in advance, which work is volume and which work is judgment — and then giving away all of the first kind.
The machine's half: logging what happened. Remembering what you sent and when. Noticing that somebody went quiet nineteen days ago. Drafting the follow-up that says the obvious thing. Scheduling, reminding, transcribing, scoring, sequencing, and pulling the context back up before a call so you walk in knowing what they told you last time.
That is the work that used to require an SDR and a coordinator. It is why Marcus can hold twenty conversations this month instead of the nine he was managing badly. Not the writing. Not the selling. The remembering.
Yours: which of these is real. What to say to this specific person. When to push, when to wait, when to walk away. And the one that matters most — when the system is wrong.
Get that division right and twenty conversations is a Tuesday. Get it backwards and you have automated the only part that was ever yours.
Stages, so nothing lives in limbo. Any CRM ships with six stage names; they are the least useful thing in the box. What earns its keep is the entry criteria. A deal cannot move up because you feel good about it. Write down what has to be true, in evidence, before a deal advances, and you have cured most of what inflates a forecast.
BANT-Plus. Budget, Authority, Need, Timeline — and two the original four now miss badly.
Fit: does this solve their problem, or one adjacent to it? Selling to somebody who buys and then fails costs you the referral, the case study, and the next three deals in their network.
Trust: do they believe you can do it? It belongs in the qualification because it is the variable most likely to be assumed. You will assume it precisely when it is absent, because when it is present you can feel it.
The machine fills in most of the first four from what has already been said — it read the transcript, it knows whether a number came up. It cannot fill in Fit. Fit is a judgment about whether their problem is the one you actually solve, and every incentive you have points at yes.
Which fields earn their place. Not the ones the software ships with. The ones that predict your deal, which you can only learn from the ones you have already won and lost. For Marcus that turned out to be four: how many loan officers, whether they can see their own retention number, what they have already bought and abandoned, and who inside the building owns the problem. Two of those are not standard fields anywhere. They are the ones that call the outcome.
Scoring, and the thing it will do to you. Score your leads so you know where the hours go — fit, engagement, budget, authority, timeline, and how much the problem actually hurts. Let the machine keep it current, which it will do without getting tired or optimistic.
Then be clear about what you have built. You chose those weights. You chose them, whether you noticed or not, from the deals you have already closed — which means the model is a very fast, very confident machine for finding more of what has already worked on you.
It will therefore mark down the large opportunity that is early, because early enterprise deals are structurally incomplete: no confirmed budget, no named authority, no stated timeline, all for perfectly good reasons that look identical to a bad lead. And about the genuinely unfamiliar deal it has nothing to say at all — which is not the same as saying nothing. It will produce a number anyway, and the number will look like the others.
Use the score to allocate your Tuesday. Don't use it to decide what matters.
The weekly review. One hour, same slot, and it has four questions in a fixed order.
What moved? Not what happened — what changed stage, and on what evidence.
What stalled? Anything untouched for two cycles is stuck, and a stuck deal is usually waiting on a person nobody has identified yet.
What is the single next action on each live deal, and who owes it? If the answer is "follow up," there is no next action. There is anxiety with a date on it.
Which of these am I lying to myself about? Every pipeline contains at least one deal that exists because you want it to.
The machine prepares all four. It hands you the stage changes, flags everything untouched for a fortnight, lists the deals with no scheduled action, and shows you which forecasts you got wrong last quarter and by how much. It will never volunteer that a deal is a fantasy, because nothing in it wants anything.
The Weekend
Saturday morning, Marcus clears the calendar.
He has been using the free tier of the CRM he signed up for in year one as an address book with delusions. He rebuilds it properly — the four fields that actually predict a VaultPath deal, stages with real entry criteria, sequences for the people who aren't ready.
Then the archaeology. Every email, every message, every square on the monitor, entered with context. Eleven hours across two days, and the least glamorous work he has done all year.
By Sunday evening he can see his own business.
Three real opportunities. Five qualified, including Apex Lending, whose pilot has been quietly running since before any of this and which nobody has asked to convert. Twelve suspects worth nurturing. Seven that were never opportunities at all, which he archives, and which had been costing him something every day just by sitting in his peripheral vision.
Dallas never got the proposal. He'd drafted it and not sent it. He writes a short note that doesn't pretend otherwise and gets a reply Monday that is warmer than he has earned.
Then he turns on the scoring, because he built it and it's satisfying, and the model ranks everything in about a second.
Karen Walsh: 85. The Colorado lender: 78. Mike Donovan, First National: 58.
Marcus reads that twice.
Fifty-eight is under the line he drew on Saturday for what gets his actual hours. The system he spent the weekend building has just proposed, politely, that he stop spending real time on the largest company that has ever written to him.
The model isn't wrong on its inputs. Mike has no confirmed budget, no timeline, and an evaluation process nobody has described to him. Every field is soft. Fifty-eight is fair.
It's that Mike Donovan wrote four paragraphs about his own problem to a stranger, and there is no field for that, because Marcus didn't build one, because until three weeks ago it had never happened to him.
He sits with what he nearly did.
He doesn't change the score. Changing it to get the answer he wants would make it useless for every other row. He adds a column instead, one he fills in by hand and the machine can't compute — conviction — and writes high beside First National with two sentences explaining why, so that in October, when he's forgotten this evening, he can see what he thought and whether he was right.
Monday morning the dashboard is clean.
His phone rings.
"Marcus, about Thursday," Karen says. "My CEO wants to meet you. Can you do Tuesday at two instead?"
He opens her record. It's all there — every conversation, the twenty-question score she ran on herself, the three answers she couldn't give, the working session notes, who in her building owns what.
"Yes. What does he care about?"
"Implementation. He believes the problem now. He wants to know whether it's six weeks or six months and whether it breaks anything."
"That I can do."
After she hangs up he logs it, sets the prep task, and notices that his desk is empty for the first time since March.
Structure creates capacity. The capacity isn't for more deals. It's for being entirely present in the one in front of you, which is the only thing that has ever worked.
Except the desk isn't quite empty.
One note is still stuck to the bottom of the monitor, where it has been for three weeks.
TMBA — 45 min. Nineteen days.
He'd stopped writing notes on Sunday. He never took this one down, and he understands, holding it, that he hasn't taken it down because he doesn't know which system it goes in. It has no stage. No score. No close date. By every rule he spent the weekend building, it isn't a deal and doesn't belong anywhere.
It is also the only item on the desk that came from the file — the invitation came from a post, and the post came from the file — which makes it, on the evidence, the most valuable thing he owns and the one thing his new machine has no idea what to do with.
He puts it back on the monitor.
Nineteen days, forty-five minutes, and a room full of people who have never heard of him.
Endnotes
The first draft of this chapter carried a figure — that salespeople using structured pipeline management close 18% more deals — attributed to industry benchmarks and to CSO Insights. I have cut it.
The organization is real and does serious work. The number is the problem: I could not trace it to a specific study, a defined population, or a stated methodology, and every version I found in circulation cited another version in circulation. That is the same failure as the 60% recall figure in Chapter 4, with one difference that decides it. The 60% figure is load-bearing lore, kept with a warning attached. This one was decoration on a claim that doesn't need it. Nobody who has run a pipeline out of their own memory for six months needs a percentage to be told it doesn't work.
Marcus's scoring weights, and Mike Donovan's fifty-eight, are invented. The failure they illustrate is not: a hand-built scoring model encodes the shape of the deals you have already won, and it will hand you that assumption back as a number, looking exactly as authoritative as every other number on the screen.
Draft completed: December 2025 | AI-era rewrite: August 2026