Deals do not die on price. They die on a promise nobody wrote down.
Stalled deals usually wait on a promise, not on price. Track every commitment either side makes as it is spoken, then draft what is owed for a person to send.
Jared Hawkins-Young · Founder, Lost Star TechnologyOctober 8, 20264 min readPart 5 of 13Key takeaways
- Every commitment either side makes on a call, in a thread or in a rep note becomes a tracked issue with an owner, a due date and the moment it was said.
- The ledger tracks the buyer's promises too. On the Northwind deal the close date depends on the one the buyer owes: an introduction to the security lead.
- After the call an agent attaches promised documents, builds the quote from stated scope and routes unanswered questions to a named person. Everything stops at a draft, and a person presses send.
- Dixon and McKenna's study of 2.5 million sales conversations puts 40 to 60 percent of the average rep's pipeline lost to no decision, a figure from the authors' own research.
Most stalled deals are waiting on something somebody said they would do: a SOC 2 report, a revised quote, an introduction to the security lead. Track every commitment either side makes as an issue at the moment it is spoken, with an owner, a due date and the timestamp, and have an agent draft what is owed for a person to send. A forgotten promise never shows up in a pipeline review, because it is not a field. It is a sentence somebody said out loud and everyone assumed someone else caught.
Why do deals stall after a good call?
Price gets the blame because it is visible. The more common ending is quieter. Matt Dixon and Ted McKenna built their 2022 book, The JOLT Effect, on a study of 2.5 million sales conversations, and they put the share of deals in the average rep's pipeline lost to no decision at 40 to 60 percent. Their explanation is indecision: a buyer afraid of making the wrong call. The figure comes from the authors' own research, not an independent audit, but the shape will be familiar to anyone who has run a pipeline review.
A missed promise does not cause indecision. It gives indecision somewhere to hide. A buyer who is unsure, and who has not received the security report you promised, has a reason to wait that nobody on your side can see.
What goes on the ledger?
| What was promised | Owner | Heard at | Due | State |
|---|---|---|---|---|
Send the SOC 2 report Report attached from the approved folder | MarcusUs | REC 39:05 | Tue | Drafted |
Revised quote at four seats, onboarding included Built from stated scope, two seats removed | MarcusUs | REC 18:02 | Tonight | Drafted |
Confirm in writing which HubSpot objects it writes Not in approved material, so a person answers | LegalUs | REC 22:15 | Thu | Routed |
Introduce the security lead before the kickoff on the 14th Short reminder drafted for the due date | Dana ReyesThem | REC 38:40 | Before the 14th | Waiting |
quote.build4 seats + onboarding$24,000 · 6-seat line removeddocs.attachSOC 2 reportFrom the approved folderemail.draftFollow-up to DanaIn Marcus's drafts, not sent
Every commitment either side makes becomes a row: what was promised, who owns it, when it was said, when it is due, and its state. On the Northwind deal from part 1 that is four rows from one call. At 39:05 Marcus promised the SOC 2 report by Tuesday. The revised quote follows from the scope cut at 18:02. At 22:15 Dana asked for written confirmation of which HubSpot objects the integration writes, before her security lead sees it. And at 38:40 Dana herself promised to bring her security lead in before the kickoff on the 14th.
That last row is the easiest one to miss. A ledger that only tracks your own side cannot see the half of the deal the buyer owes. On Northwind, the buyer's row is the one the close date hangs on: no security lead means no review, and no review means no January.
What does the agent do with the ledger?
The ledger is only worth something if something acts on it. Each row has a type, and each type has a known next step. A promised document is attached from the approved folder. A revised quote is built from the scope stated on the call, so the six-seat line comes out and onboarding stays in. A question the approved material does not answer goes to the person who can answer it, with the thread. A commitment the buyer owes gets a short reminder drafted for its due date.
| Issue | What gets built | Where it stops |
|---|---|---|
| SOC 2 report | Attached from the approved folder | In the follow-up draft |
| Revised quote, four seats | Priced from the stated scope: $24,000, onboarding included | Attached to the same draft |
| Written HubSpot scope note | Nothing. It is not in approved material | Routed to a named person |
| Security lead introduction (theirs) | A two-line reminder for the due date | In the rep's drafts |
Minutes after the call, Marcus has one email to Dana with two attachments waiting in his drafts: the recap that part 6 takes apart. Nothing was sent. A person presses send, every time, even when the stakes look small. The rule for what runs without a person is the same one the outbound series sets out in outbound part 13: a known event with an approved action runs, and anything else goes to someone who can decide.
What changes when every promise is tracked?
- Said out loud
- Maybe remembered
- Maybe written down
- Maybe chased
- Deal stalls
- Said out loud
- Captured at 38:40
- Owner and date set
- Drafted after the call
- A person sends it
- Closed out
On the top path nobody did anything wrong. The rep meant to send the report. The buyer meant to make the introduction. Each assumed the other side was moving. On the bottom path the promise is a row with a timestamp, and when it closes, it closes because the email went out with the attachment in it, not because someone ticked a box. It reopens if the buyer says it never arrived.
The ledger also feeds everything after it. The open security introduction is why the close-date argument in part 3 holds, and it is one of the flags behind the December drop in part 4. When a rep leaves, the next one inherits the ledger, not a predecessor's memory. And when Northwind goes quiet, the open row is what the follow-up opens on, instead of "just checking in."
Who this is not for: deals that close on one call with nothing owed in between. And the ledger only knows what was said somewhere it can read, which means a recorded call, a thread or a rep note. A promise made in a hallway needs a one-line note from the rep, or it stays where promises have always stayed.
Next in the series
Part 6, Every meeting ends with a recap, puts the ledger to work: the recap drafted minutes after the call, every inbound question answered or routed to an owner, and a quiet deal revived on its own rhythm. The series page lists every part in order.
Common questions
- How do I keep track of what was promised on sales calls?
- Treat every commitment as an issue the moment it is spoken: what was promised, by whom, when it was said and when it is due, with the timestamp from the recording. Track both sides, yours and the buyer's, and close an issue only when there is evidence it happened, such as the attachment in a sent email.
- Does the agent send the follow-up emails itself?
- No. It attaches documents, builds the quote and drafts the email, then stops. The rep reads it and presses send. A question the approved material does not answer is routed to a named person instead of being guessed.
- What about promises made off the recording?
- The ledger reads recorded calls, email threads and rep notes. A promise made in a hallway or on an unrecorded call gets onto the ledger when the rep writes a one-line note, and from then on it is tracked like any other.
Sources
- The JOLT Effect, official site: a study of 2.5 million sales conversations
- Scale Venture Partners, The JOLT Effect with Matt Dixon (webinar): 40 to 60 percent of the average rep's pipeline lost to no decision
- Matthew Dixon and Ted McKenna, The JOLT Effect, Portfolio, September 2022 (publisher listing)
Written by Jared Hawkins-Young from the system he builds and runs. Diagrams are drawn from the same specs as the software. Published October 8, 2026. Illustrative numbers are labelled as such; every other figure carries its source.