The forecast wasn't wrong. It was negotiated.
Uniphore, a $110M enterprise AI company, built its revenue forecast the way most companies do: bottoms-up, in a spreadsheet, once a week, from whatever its eighty sales reps chose to type in. Reps sandbagged to protect themselves. Leaders padded to protect the board number. Fewer than 60% of reps submitted on time at all, and about a fifth of the pipeline slipped in the final days of every quarter. I replaced the spreadsheet with an evidence-based forecast (deal confidence derived from what buyers actually did, not what reps reported) and rebuilt the planning and board-reporting layer underneath it. Forecast accuracy went from 68% to over 80% in a single quarter. Slippage fell below 8%.
The problem
The forecast was 68% accurate at a company doing $110M a year. It was assembled every week, bottoms-up, in a spreadsheet, from what eighty reps across three regions chose to enter, and fewer than 60% of them entered anything on time.
So the number that went to the executive team (and, four times a year, to the board) was built from a partial set of inputs, submitted late, by people with an obvious interest in what those inputs said. Planning was worse: the annual plan lived in Excel, and plan, forecast and actuals were reconciled by whoever had the patience. Reporting ran on a delay that made it decorative.
The most important number in the business was a guess. Everyone senior knew it was a guess, and nobody could prove by how much.
A forecast nobody could trust. At a company doing $110M a year, the weekly forecast was 68% accurate, assembled bottoms-up in a spreadsheet from what eighty reps across three regions chose to enter.
Built from partial inputs. Fewer than 60% of reps submitted on time. The number going to the executive team (and quarterly to the board) was assembled from whoever had entered something, late, by people with an obvious interest in what it said.
Planning by hand. The annual plan lived in Excel: historical pulls, hand-built models, version files, quarterly rebalancing by hand. Plan, forecast and actuals lived in different places and reconciled by whoever had the patience.
Reporting arrived stale. The quarterly board pack took three weeks to build, the monthly pack a week. By the time either landed, it described a company that had already moved on.
The diagnosis
The forecast wasn't inaccurate. It was negotiated.
I was brought in to build an offshore team and fix the forecasting and analytics programmes. The technical work was obvious enough; the thing that mattered took a fortnight to see: a bottoms-up weekly spreadsheet forecast is not a prediction. It's a bargaining process wearing a spreadsheet's clothes. Reps under-commit because a missed commit is punished; leaders over-commit because the board expects a number. You don't fix that with a better spreadsheet, a stricter deadline, or a more forceful pipeline review. You fix it by taking the human out of the input.
Get the signal from what the buyer does, not what the rep reports
Deal confidence should come from evidence: emails answered, calls held and what was said in them, quotes generated, stakeholders engaged, activity accelerating or going quiet. All of it already existed, in the CRM, in email, in Gong, in the CPQ. None was being read. Once the forecast came from behaviour, sandbagging and padding stopped being possible. There was nothing left to sandbag.
You cannot fix a forecast by asking for it more often
The instinct when submission compliance is at 60% is more cadence: another review, another deadline, another dashboard nobody fills in. They already had a weekly cadence. Adding meetings would have added meetings. Compliance went to 98% because the system filled in most of it and the rep only had to confirm.
Moving the plan into a system is a diagnostic, not a migration
We set out to replace Excel with a planning platform. What that exposed was that plan, forecast and actuals had never reconciled. The inconsistencies had simply been absorbed, quarter after quarter, by people rebuilding the sheet. It forced a mid-cycle rebuild of the annual operating plan. That wasn't scope creep; that was the finding.
What I built
An evidence-based forecast
Revenue intelligence wired into Salesforce, email, Gong call recordings, the DealHub CPQ and HubSpot signals, producing deal-level confidence scores and rep-level forecasts from observed buyer behaviour rather than reported opinion. The rep's job changed from constructing the forecast to confirming it.
A planning system, live in two months
Drivetrain replaced the Excel model, wired into Salesforce for real-time plan-versus-actual across marketing, sales, retention and finance. The mid-cycle AOP rebuild it triggered produced Uniphore's first genuine single source of truth for global pipeline, coverage, bookings and channel performance.
Board and executive reporting that arrives while it's still true
Tableau on top of Drivetrain and the CRM. The quarterly board pack went from three weeks to one, the monthly leadership pack from a week to two days, the weekly pod review to real-time.
A central deal desk
120 to 140 high-value deals a month, with contracts up to $1.5M, running through one desk: deal inspection, quote generation and tiered discount approvals. Legacy ad-hoc pricing was consolidated into three pricing families.
And the team that still runs all of it
A four-pod offshore RevOps function built from zero (systems, deal desk, analytics and forecasting, and programmes), including an ops onboarding curriculum and Mindtickle certification that ramped new hires in two months against a five-month benchmark.
The outcome
Eight-month engagement. The team it produced scaled to 14 and still runs Uniphore's global RevOps function, years later.
Our forecast was a weekly negotiation and everybody knew it. Shashank didn't argue with the number. He changed where it came from. The team he built still runs our entire revenue operations function.
What this engagement didn't solve
An accurate forecast is not a bigger forecast, and I want to be clear about which one I delivered. Accuracy went from 68% to over 80%. Whether Uniphore then hit its number is a different question, driven by things a forecasting system doesn't touch: the product, the market, the reps. A forecast tells you the truth earlier. It doesn't sell anything for you, and I have no attainment number to claim.
Compensation and quota design were never in scope. A real gap, because the incentives are precisely what were distorting the forecast in the first place. I removed the rep's ability to game the number. I didn't fix the reason they wanted to.
And eight months is not long enough to know whether a culture changed or merely complied. The systems held. The team held. It's still running the function years later, which is the outcome I actually care about. But the honest position is that I built something durable and left, and durability is easier to claim than to prove.
Why this matters for you
This is your problem if:
Your forecast is assembled bottoms-up in a spreadsheet, from what your reps choose to type
Deals slip in the last fortnight of the quarter and nobody saw it coming
Reps submit the forecast late, or not at all, and your answer so far has been to ask harder
Sales gives one number, finance gives another, and the board gets a third
Your quarterly board pack takes weeks to build and is stale before it lands
You are considering a revenue intelligence tool and aren't sure what it would actually change
A forecast built from what reps type is a negotiation, not a prediction
Reps under-commit and leaders over-commit, and both are behaving rationally. Get your deal confidence from what the buyer did (emails answered, calls held, quotes opened, stakeholders engaged) and the gaming stops, because there's nothing left to game.
You cannot fix a forecast by asking for it more often
If submission compliance is at 60%, another weekly review buys you another weekly review. Make the system fill in the answer and the rep confirm it, and compliance stops being a discipline problem.
Moving your plan into a system is a diagnostic, not a migration
It will tell you whether your plan, your forecast and your actuals have ever actually reconciled. In my experience they usually haven't, and finding that out is worth more than the software.