Pipeline review from the rep side

Take your CRM's stage names, whatever they are, and write beside each one a single sentence that starts with the buyer. If you cannot, the stage is decoration. Here is the test applied to Northwind:

Notice that every left column entry has a date and a name attached, which means every one of them is already in the record you built in b8. That is not a coincidence. A stage definition you cannot evidence from the CRM is a stage definition you will argue about in the review.

Three signals. Any deal carrying all three is a zombie, and a zombie is not a deal you might lose, it is a deal you have already lost and are still paying rent on.

• No next step with a date. Not "following up next week". A time on a calendar that both of you can see.
• No reply in 21 days. Silence from a buyer who was engaged is information. Silence from a buyer who was never engaged is not even that.
• No second stakeholder. One contact is a conversation. A deal needs at least two people who would notice if it disappeared.

Reps resist killing deals for one reason: it feels like admitting the pipeline was fiction. It is the opposite. The rep who carries eleven zombies into every review has a forecast nobody believes and eleven weekly reminders to feel bad. The rep who kills them has a smaller number that people trust and four hours back to spend on Northwind. Closed-lost with a reason is a finished piece of work. An open deal with no next step is an unfinished one, forever.

Most of this course has been careful about where AI helps. Zombie detection is one of the places where it is straightforwardly better than you, and the reason is not intelligence. It is that the three signals are all structural facts sitting in your CRM - last inbound date, next-step date, count of distinct contacts on the thread - and a model reads them without wanting any particular answer.

• Recency is arithmetic. You remember the good call in week two. The system remembers that it was eleven weeks ago.
• Multithreading is a count. Humans mistake one enthusiastic champion for coverage. A count of two is a count of two.
• Hope is the bug. The rep is the only party to the deal with an emotional stake in it still being alive, which is exactly the wrong qualification for the job of auditing it.

Here is a CSV export of my open opportunities with these columns:
account, stage, amount, created_date, last_inbound_date,
next_step_text, next_step_date, distinct_contacts, trigger_logged.

Flag every row where ALL THREE are true:
next_step_date is empty or in the past,
last_inbound_date is more than 21 days ago,
distinct_contacts is less than 2.

Return: account, stage, amount, days since last inbound, and which
of the three signals fired. Sort by amount, largest first.
Do not tell me whether the deal will close. Do not write a summary.

Here is my pipeline. Which deals will close this quarter and what
should my forecast be? Be specific and give me a percentage for each.

The second prompt asks the model to do the one part of the job that is yours. It will answer, fluently, with numbers that look exactly like the good ones. Ask it for facts you can check and it is a great analyst. Ask it for a judgment and it is a very confident stranger.

An override log is four columns in a spreadsheet you own. Date, deal, model number, your number, reason. That is it. Nobody else needs to see it and it should not live in the CRM, because the moment it is scored by someone else you will start writing reasons that sound good instead of reasons that are true.

• The reason must be a fact, not a feeling. "Strong relationship" is not a reason. "Priya sent the security questionnaire on 14 May" is a reason, and in six weeks you can check whether it meant what you thought it meant.
• Both directions count. Overriding down is an override too, and it is the one that teaches you the most, because you will find you almost never do it.
• At quarter end, sort by direction. Count how many upward overrides came true and how many downward ones did. Most reps find their upward overrides land under half the time and their few downward ones land nearly always. That is the signature of an optimist, and it is the single most useful thing you can learn about yourself as a seller.

The correction is not to stop overriding. It is to notice the size of your own bias and shave it. A rep who says "the model says 40, I say 70, and I know I run 20 points hot, so put me at 55" is doing something no model can do and no leader can coach into you.

Sandbagging is calling a deal lower than you believe it is, usually by holding it out of commit until it is nearly signed. Every rep has done it. Let us be honest about why it feels good, because pretending it is simply dishonest is why the advice never lands.

• Why it feels safe. A missed commit is a public event with your name on it. A beaten commit is a quiet win. The asymmetry is real and it is not in your head.
• What it costs you, month one. Nothing visible. This is the trap.
• What it costs you, quarter two. Your leader stops using your number and starts applying a private multiplier to it. From then on you are not forecasting, you are being interpreted, and you have no say in the interpretation.
• What it costs the deal. Deals held out of commit do not get help. No exec sponsor call, no legal fast-track, no discount approval standing by. You hid the deal from the resources that could have closed it.
• What it costs you, year one. The rep whose number is trusted gets the territory, the escalation and the promotion. Trust in a number is a career asset that compounds, and sandbagging spends it a little at a time for a feeling of safety on a Tuesday.

The alternative is not bravado. It is calling the number you actually believe and saying out loud what would have to be true for it to slip. That sentence - "this is a commit, and the thing that kills it is the IT review Jonas has not scheduled yet" - is what a good rep sounds like, and it is impossible to fake without a real record underneath.

The difference is visible in the first question asked. "Why is this still open?" is an interrogation and it teaches you to hide deals. "What would have to be true for this to close, and what do you need from me?" is coaching and it teaches you to surface them early.

You do not control which one your leader runs. You do control what you bring, and bringing the honest version changes the meeting more than you expect. A rep who opens with "I killed two, here is why, and here is the one I need help on" has moved the room from audit to problem-solving before anyone else has spoken. Gartner's finding that organisations giving sellers AI-enabled next best actions are 2.6x more likely to achieve commercial growth is really a finding about this: the review is worth something when it ends in an action, and worth nothing when it ends in a justification.

Four sentences. No slides. Northwind, week eight:

• The number and the stage evidence. "Northwind is a commit at 48k. Priya wrote back on 14 May with a finance sign-off date."
• The single risk. "The thing that kills it is Jonas's data residency review, which is not scheduled."
• The override, named. "The system has this at 40 because our thread went quiet for eleven days. That is because Mira was on site. I have it at 75 and the reason is in my log."
• The ask. "I need our security pack in front of Jonas by Thursday."

Every one of those sentences is checkable, which is exactly why they are believed. Compare with the version that gets a follow-up question: "Northwind is looking really strong, Mira loves us, I think we get it this quarter." Same deal, same rep, zero information.

Some of tonight's zombies are not late-stage failures. They are Target failures from b2 that took eleven weeks to become visible. Bright Harbour has twelve staff and sits under your ICP floor - it was never going to buy at your entry price, and the simulator capped it at 1.2% in b1 for exactly that reason. It did not become a bad deal. It was created as one.

• Wrong seat. Right company, contact with no budget and no pain. It can survive four stages on politeness alone.
• Under the floor. A company that cannot buy at your price. Every hour spent here is an hour Northwind did not get.
• No trigger at creation. If nothing had changed at the account when you opened it, you were selling to a steady state, and steady states do not buy.

This is why the homework sends you back through ten closed deals. The pattern you are looking for is not in this quarter's pipeline. It is in the moment each deal was created, which is a stage-1 habit and the reason b2 sits where it does in this track.

Nobody will build this for you, and if they did you would not trust it. One tab, one row per deal, filled in twice: once when you call it, once when it resolves.

• Twelve weeks is the minimum. Under about twenty resolved deals you are reading noise and will conclude whatever flatters you.
• Slipped is wrong. A deal that closed one quarter late was not a good call that took longer. It was a wrong call about timing, which is the call you were actually being asked to make.
• Keep it out of the CRM. The instant it is visible to someone who scores you, it becomes a performance and stops being a mirror.

In b8 you added one field that most CRMs do not have: the trigger you used to open the account, with its date and source. It looked like admin. It is the highest-value field on the record, and this is the session where it pays out.

Sort your closed-won deals by trigger type. New facility or lease. Funding round. Exec hire. Job reqs. Regulatory change. Nothing, we just called them. Then do the same for closed-lost. What you are looking for is not the trigger that appears most often in wins, because that is just the trigger you use most often. You want the trigger with the best ratio, and it is usually not the one you assumed.

• The benchmark says timeline hooks beat problem hooks by roughly 2.3x at the reply stage. Your own record tells you something the benchmark cannot: which timeline hooks turn into revenue at your price point, in your segment.
• Watch for the trigger that replies well and closes badly. It is real, it is common, and it will quietly eat a quarter. High reply rate plus low win rate usually means the trigger creates curiosity rather than a deadline.
• Then take it back to b2. The output of this analysis is not a report. It is a change to what you go looking for on Monday morning.