Note11 May 2025

In which the forecast is a useful fiction everyone agrees to believe.

Forecasts Are Lies (But Useful Lies)

A field guide to pretending responsibly

The first time I built a "bulletproof" forecast, I was at my kitchen table in the middle of the pandemic. Two monitors glowed with toggles for every variable: revenue ramp, hiring plan, invoice-level billing models. I could zoom from the 10,000-foot view into a single customer's payment schedule, navigating the business like I was piloting through a digital telescope. When everything else felt chaotic, this felt different. I had data. I had formulas. Confidence intervals that looked scientific and safe.

Then the world ignored it entirely.

At that kitchen table, emails arrived with subject lines like "Effective immediately: all non-essential spend paused." Timelines stretched in ways no model could capture. The forecast's neat chain of dependencies, each cell feeding the next in perfect harmony, collapsed into suggestion rather than signal. Forecasts are polite fictions. Not malicious lies, but structured stories we agree to believe for a while. They break gently, almost apologetically, when reality barges in with complete indifference to what we thought might happen. Yet we keep building them because in the absence of certainty, a shared story remains one of the fastest ways to move forward.

A forecast is always a story about a future that won't happen. That is the point, not a flaw. The value of the exercise lies in alignment. You take a fog of uncertainty and give people something to point toward. Suddenly, decisions that felt impossible become possible: who to hire, where to invest, what to cut. Whether you land at 47 percent growth or 52 matters less than whether the process moved people out of paralysis and whether the decisions held up once the future arrived looking nothing like the plan.

Precision feels right. Decimals bring something different into a room than a round number. I learned this early, having built a table with multiple decimal places on churn probabilities: 14.327%, based on history and two gut-check questions to a single sales manager. No one questioned it. The decimals did the work of persuasion that the methodology couldn't. A round number invites argument. A precise one implies that someone already had the argument and won.

Behind every clean output sits a mess of judgment calls. Which customers count as "at risk" depends on who you ask and when. The sales team's version of pipeline coverage and finance's version diverge in ways that only surface when someone pulls the thread. I've watched a revenue projection change by fifteen percent based on whether one VP classified a deal as "likely" or "committed," a distinction that meant everything to the model and nothing to the customer, who hadn't returned an email in two weeks.

The shape of a forecast often reveals what someone hopes is true rather than what's likely. Conservative assumptions get pushed upward in hallway conversations before they reach the board deck. Aggressive ones survive because the hiring plan already depends on them. The fiction tightens around itself until the number on the screen feels inevitable, which is the moment it becomes most dangerous.

In a quarterly review, we presented a number everyone could defend and no one could prove. There were other versions sitting behind it (more conservative, more aggressive), but this was the one that fit the plan we were already building. By then, hiring plans were already drafted against it. Changing the number meant changing everything downstream.

A customer worth ten percent of our revenue had quietly paused spending the week before. The model on the screen still showed them at full run rate.

The discussion moved from headcount to marketing spend to product priorities.

“So we can afford three more heads?” someone asked. “The model says yes,” came the answer.

No one stops the meeting. The number stays on the screen, and the conversation reorganizes itself around it, moving from headcount to marketing spend to product timelines, each decision taking the number as a given, not because anyone declared it correct, but because it is already doing the work of something settled.

The customer who paused doesn’t quite enter the discussion. It’s not ignored exactly. It just sits off to the side, harder to reconcile with what the model is asserting, easier to leave where it is for now.

The caveats are still there if you go back, buried in the tabs, in the assumptions, in the way someone hedged a sentence and then moved on, but they don’t travel. What moves is simpler, more portable.

“The model says.”

It still says things like “assuming retention holds” and “given current trends.” No one argues with those either. The conditions stay attached to the sentence, but the decisions don’t.

Hiring accelerates on imaginary revenue. Roadmaps expand toward impossible futures. Marketing bets stack up like chips in a game no one can win. Admitting the forecast is wrong means questioning every decision it supported, so the questions never arrive. Capable teams walk straight off cliffs while still consulting the map. The best ones put the spreadsheet down often enough to check the ground beneath them.

A forecast should orient without binding you. Ranges beat single points: forty to fifty-five percent growth, with retention holding, hiring on track, and conversion rates steady. Each number tied to conditions that must hold for it to remain valid.

Scenarios matter: a stretch case, a base case, a worst case, and one labeled "Oh No." The point is to recognize the triggers for each before you're forced to decide under pressure. Forecasts must stay alive through interrogation. Something always changes mid-quarter: a lost customer, a missed hire, a market turn. Updating isn't maintenance. It is the job.

Even when nothing feels urgent, asking "What would have to be true for this not to work?" and treating the answer as an action item often keeps you upright. Trust the forecast enough to move, but question it enough to steer.

Forecasts will always be lies, but they can be useful lies if you treat them as tools instead of truths. Their worth is in helping you recognize a wrong turn in time to correct course.

You'll present numbers with more confidence than they deserve; and, make choices on elegant guesses at best. The real skill is holding that double vision: one eye on the map, the other on the horizon. When the gap between them starts to widen in ways that matter, you put the chart down and navigate by what you can actually see.

A forecast isn’t a prediction so much as a commitment. Once it lands, people start acting as if it’s real, and that’s what makes it matter.


Reply

I’d welcome your thoughts on this essay. Send me a note →

Related reading
Latest entries