The forecast that reads the calls
Adjust the weighted pipeline only where the CRM contradicts what a buyer said on a recorded call, and show the sentence. The gap is your pipeline review agenda.
Jared Hawkins-Young · Founder, Lost Star TechnologyOctober 8, 20265 min readPart 4 of 13Key takeaways
- A weighted forecast multiplies amounts by stage probabilities, so it is only as honest as the three fields a rep chose: amount, stage and close date.
- The leader's view shows the whole book filtered to the deals where the CRM contradicts what a buyer said on a recorded call, and every flag links to the sentence.
- Gartner reported in 2020 that only 45% of sales leaders and sellers had high confidence in their forecast accuracy; no sample size was published.
- We have not found a platform that forecasts primarily from conversation content. Gong comes closest, with its own claim of 20% more precision than CRM-based algorithms.
To forecast from call recordings, start from the weighted pipeline, then adjust every deal where the CRM contradicts what the buyer said on a recorded call, and show the sentence behind each adjustment. You end up with two numbers: the forecast the fields produce and the forecast the calls support. The gap between them is your pipeline review agenda.
Why is the weighted forecast built on the wrong input?
A weighted forecast multiplies each deal's amount by a probability for its stage and adds the results up. The arithmetic is fine. The inputs are three fields a rep chose: amount, stage and close date. The forecast is only as honest as the person filling them in, and that person has the most reason to round up.
Sales leaders know it. Gartner reported in February 2020, from its State of Sales Operations survey, that only 45% of sales leaders and sellers had high confidence in their organization's forecasting accuracy. Gartner did not publish a sample size, and the figure is from 2020, so treat it as a direction rather than a measurement.
What does a forecast built on the calls look like?
This is the leader's view, one level up from the single deal the rep sees in part 3. In October and November the two forecasts agree, because every deal due to close in those months is consistent with its calls. In December they part by $60,000. February loses $16,000. March gains $11,000, because a buyer said on a call that a second team wants the product and the rep never raised the amount. Two months the calls say miss, and one they say beat.
None of those adjustments is a model's opinion of a deal. Each one is a specific sentence that contradicts a specific field.
Which deals get flagged?
| Opportunity | Owner | Amount | Close | Stage | Signal check |
|---|---|---|---|---|---|
| Woodgrove · support rollout | Ana | $42,000 | Nov 21 | Proposal | consistent with 3 calls |
| Northwind · HubSpot rollout | Marcus | $30,000 | Dec 15 | Proposal | Close date + amount + stage "The review alone takes six weeks, and we haven't started it." REC 31:14 · Dana Reyes, VP Sales Jan 30 · $24,000 · Discovery · slips out of Q4 |
| Contoso · team plan | Jon | $36,000 | Dec 05 | Negotiation | consistent with 4 calls |
| Fabrikam · support and success | Marcus | $28,000 | Dec 12 | Proposal | Amount "Only the support team is moving this year." REC 09:15 · Head of Operations Lower to $18,000 · scope cut on the call |
| Adventure Works · 15 seats | Ana | $45,000 | Dec 19 | Commit | Stage "I can't sign this. Our CFO owns anything over twenty-five thousand." REC 44:51 · VP Operations Back to Discovery · economic buyer never on a call |
| Proseware · pilot to annual | Jon | $31,000 | Jan 16 | Proposal | consistent with 2 calls |
| Tailspin · CX team | Jon | $27,000 | Feb 27 | Proposal | Close date "Nothing new gets bought until the reorg settles in March." REC 21:30 · Head of CX Push to Apr 15 · slips out of Q1 |
| Litware · support team | Ana | $33,000 | Mar 20 | Proposal | Amount "If this works for support, the success team wants it too." REC 33:02 · Director of Support Raise to $51,000 · scope grew, never raised |
Northwind accounts for $18,000 of the December drop on its own: a $30,000 deal weighted at 60% in Proposal moves to January at $24,000 and back to Discovery. Adventure Works is the bigger hit. It sits in Commit at $45,000, and the VP Operations said on the call that the CFO owns anything over $25,000. The CFO has never been on a call. That is not a judgment about Ana, the rep. It is a dated sentence and an attendee list with a name missing.
The flags come from a short list of contradictions, and each one is a fact a stage field has no place to hold:
- A paper process named and not started, longer than the time left before the close date.
- An economic buyer who has never been on a call or a thread, on a deal in Commit.
- Scope cut on a call but still priced in the contract.
- A buyer who named a reason nothing will be bought before a given date.
- Scope that grew on a call while the amount stayed put.
| Weighted forecast | Adjusted to what buyers said | |
|---|---|---|
| Input | Amount, stage and close date, as the rep set them | The same fields, plus every typed signal from calls, threads and notes |
| Who chose the input | The rep | The buyer said it; the rep confirms it |
| What it cannot see | A review not started, a signer never met, a scope cut | Deals with no recorded call, which show as no evidence |
| How you check a number | Ask the rep | Press the flag and hear the sentence |
Does any forecasting tool already read the calls?
The common approach in forecasting tools is to learn from CRM history: stage movements, amounts, close dates and activity counts. We have not found a platform that forecasts primarily from what buyers said. Gong comes closest. Its forecast page says it uses 300+ unique signals to predict deal outcomes with 20% more precision than algorithms based on CRM data, and VentureBeat reported a version of the same claim with a product update in November 2023. That is a vendor claim against its own baseline, not an independent audit.
The point is not the model. It is the input. A paper process named and not started, a buyer never met, a scope cut still priced: each is a dated, quotable fact, and none of them is something a stage-field model was built to weigh.
How do you run the pipeline review with it?
Start the review with the flagged rows, not the top of the list. For each one the rep presses the flag, hears the sentence, and either accepts the adjustment or says why not. Accepted adjustments move the forecast you take upstairs. Dismissed ones stay visible with the reason, so you can see later who was right. The deals that are consistent with their calls do not need discussing, which is most of the list.
Then keep score. Track the adjusted forecast against the actual result each quarter, next to the weighted one. Either the gold line lands closer or it does not, and after two quarters you will know. Part 12 covers what else to measure, and the published numbers worth knowing before you start.
Who this is not for: if most of your deals close on one call or through self-serve checkout, there is not enough conversation per deal to adjust. And a deal with no recorded call is not consistent with its calls. It shows as no evidence, which is a flag of its own.
Next in the series
Behind the Northwind flag is a promise nobody has kept yet: Dana's introduction to her security lead. Part 5, Deals do not die on price, is about catching every promise at the moment it is spoken, on both sides of the deal. The series page lists every part in order.
Common questions
- How do I forecast sales from call recordings?
- Keep the weighted forecast as the baseline, then flag every deal where a typed signal from a recorded call contradicts its amount, stage or close date: a security review not started, an economic buyer never met, a scope cut still priced. Adjust those deals only after the rep has heard the sentence, and track the adjusted forecast against actual results each quarter.
- Does Gong already forecast from conversations?
- Gong says its forecast uses 300+ signals and predicts deal outcomes with 20% more precision than algorithms based on CRM data. That is Gong's own claim against its own baseline, not an independent audit. We have not found a platform that forecasts primarily from what buyers said.
- What happens to deals with no recorded calls?
- They show as no evidence, never as consistent. A deal nobody has recorded a conversation on is not confirmed by anything, and the pipeline review should treat it that way.
Sources
- Gartner, press release, February 12, 2020: less than 50% of sales leaders and sellers have high confidence in forecasting accuracy (State of Sales Operations survey; no sample size published)
- Gong, Forecast product page: 300+ signals, 20% more precision than CRM-based algorithms (vendor claim)
- VentureBeat, Gong Forecast gets AI upgrade, November 9, 2023
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.