Forecast with ranges, not point estimates — the credibility upgrade
The instinct that costs the planning team credibility
Every planning conversation pulls towards a single number. “What will next month’s SL be?” “How many contacts will we close at?” “What’s next quarter’s headcount need?” The planner who answers “81%” or “9,400” or “217 FTE” sounds confident. They also set themselves up to fail. The actual will almost never land on that exact value, and over months the planning team builds a reputation for being “wrong all the time” — even when the central forecast was reasonable.
The alternative is a range. “Next month’s SL will be between 78% and 82%, most likely around 80%.” It sounds vaguer in the moment. It is more honest, more often hit, and over a year of conversations it quietly builds the credibility that point forecasts corrode. The maths supports it, the behavioural research supports it, and the operations that have switched to ranges keep them.
Why the point estimate corrodes trust
Three failure modes are predictable.
The point forecast is almost never hit exactly. If your forecast is 9,400 contacts and the actual lands at 9,547, the headline reaction is “the forecast was off by 147 — the planning team got it wrong.” The forecast was, in fact, well within reasonable accuracy — about 1.5% error. But the binary frame (right number / wrong number) doesn’t see that nuance. The team takes a credibility hit on a forecast that was perfectly defensible.
The point forecast invites false precision in everything downstream. “Staffing required: 217 FTE.” The number sounds precise. The recruitment team is briefed to land exactly 217. Finance budgets to the number. When demand drifts inside the natural variability, the operation finds itself making expensive corrections to a target it shouldn’t have committed to that precisely.
The point forecast hides uncertainty from the decision-maker. An operations director seeing “81% SL forecast” for next quarter sees a number to plan around. The same operations director seeing “78–82% SL forecast, with downside risk if absence trends continue” sees a decision to make. The range surfaces the decision-relevant information that the point estimate buries.
What a defensible range looks like
A range that earns its place has three properties: it’s grounded in the actual variability of the data, it’s narrow enough to be useful, and the reasoning behind the band is explicit.
The statistical component. The minimum range is the natural noise of the underlying process. For voice arrivals, that’s the Poisson floor (see why understanding Poisson matters for planners). The square root of the mean is the standard deviation; one SD either side captures roughly two-thirds of the natural outcomes. That’s the absolute floor of the range — the planning team cannot honestly promise a tighter band than the maths allows.
The forecasting-error component. On top of the natural noise, the forecasting model itself has error. The historical MAPE/WAPE for the operation gives you the size of that error. A model that runs at 8% WAPE on monthly contact volume produces an additional band of around ±8% on top of the natural noise.
The judgement-led adjustment. If there are known risks — pending IT changes, an active attrition spike, a marketing send not yet confirmed — widen the range on the affected side. Document the adjustment. The judgement layer makes the range honest rather than algorithmic.
The headline range for stakeholders is usually the ±1 SD band — the “most likely” range. The wider band (±2 SD or more) is the planning team’s contingency thinking; it shouldn’t be promised to finance unless asked.
The behavioural mechanism
The credibility shift isn’t about the maths — it’s about how outcomes are scored. A point estimate creates a binary “hit / miss” verdict in the stakeholder’s mind. The hit rate on a precise number is very low; the team takes hits monthly. A range creates a “inside / outside” verdict. The hit rate on a well-constructed range is 70–80%; the team is right most months. Over a year, the difference is enormous.
Behavioural research on expert forecasting (Tetlock’s Superforecasting work, the Good Judgement Project, IPCC climate reporting) consistently finds the same pattern: experts who provide calibrated ranges with explicit uncertainty are trusted more over time than experts who give confident point estimates, even when the underlying accuracy is similar. The mechanism is honesty. People are good at detecting false certainty; they reward honest uncertainty with trust.
How to present a range without losing the room
Five techniques that make ranges land in real meetings.
Lead with the centre of the range. “We’re expecting around 81% — most likely between 79 and 83.” The headline number is still there; the band sits next to it. Stakeholders who want a single number to write down get one; the band is available for those who care.
Explain the source of the band. “The natural variability in this queue is about 2pp; on top of that, we’re carrying some uncertainty about absence trends — so the realistic range is 79 to 83.” The reasoning makes the band defensible.
Distinguish the range from a worst-case. Stakeholders sometimes hear “79 to 83” and seize on 79 as a commitment. Be explicit: “79 isn’t a worst case — it’s the lower end of normal. Worst case, if absence trends keep going, is 75. We can plan for that if you want to know the contingency picture.”
Reference past accuracy. “Last quarter we forecast 79–83, the actual was 81. The quarter before, we forecast 76–80, the actual was 79.” A track record of inside-the-range outcomes makes the range itself credible.
Don’t apologise for the range. The most common mistake is treating the band as a confession of weakness — “we can’t be more precise.” The honest framing is “this is the best estimate the operation supports.” Stakeholders who notice the difference reward the planning team that owns the uncertainty.
Legitimate objections
Three objections come up and deserve straight answers.
“I need a single number for the board.” Fair. Give them the centre of the range, with the band as a footnote. Most board reports already have explicit risk and assumption sections; the band lives there. Telling a board that a forecast has uncertainty isn’t a weakness — it’s table stakes for any serious operating discussion.
“A range means you can’t be wrong.” Untrue. The discipline of reporting ranges is that most actuals fall inside them — not all of them. A range that’s never wrong is too wide. A well-calibrated range is wrong about 15–25% of the time; that’s the signal of an honest band. Report the historical hit rate of your ranges; the discipline is visible.
“Finance wants a budget commitment.” Budgets are different from forecasts. A budget is a commitment to operate within a particular cost envelope; a forecast is a best estimate of what the operation will do. The two should not be confused. Give finance a single budget number; give finance a forecast range that explains why the budget commits to that number.
When ranges hurt
Three situations where the range isn’t the right answer.
Binary decisions. “Should we open the new site?” needs a yes/no answer, not a range. The range belongs in the analysis that supports the decision; the decision itself is binary.
Regulatory commitments. SLAs in regulated industries are usually binary points: “90% of complaints resolved within 8 weeks.” Reporting a range against those obligations isn’t allowed. The forecast still has a band; the commitment is a point.
Trivial decisions. A range on whether to buy 12 vs 14 chairs is over-engineered. Match the precision of the answer to the consequence of the decision.
The operating-model shift
Adopting ranges as the default for material forecasts is a small cultural change with three operational implications.
The forecast template needs to carry the band. Spreadsheets, BI dashboards, MI packs all need columns for “forecast low” and “forecast high” alongside “forecast central.” The platform should make the range as visible as the centre.
The accuracy-reporting cycle needs to track hit-rate-in-range. The new metric: what percentage of actuals fell inside the reported range? A healthy band hits ~75–85%. A band that hits 100% is too wide; a band that hits 50% is too narrow. The reporting cycle reviews this and adjusts.
The leadership pack needs to model the range, not the point. Capacity scenarios run at both ends of the band. Hiring decisions stress-tested against the upside and downside. Finance impact at the central case and the bookends. The discipline of presenting the range to leadership trains leadership to use it.
The honest summary
The point forecast is a confidence trick. It sounds more sure than the data supports and it sets the team up for repeated, predictable disappointment. The range is more honest, more often achieved, and quietly more persuasive over months. The maths is the case; the credibility is the dividend. The operations that have made the shift don’t go back.
Conclusion
Provide ranges when projecting future performance. Lead with the centre, explain the band, distinguish from the worst case, reference your track record, and don’t apologise. Over a year of conversations, the planning team that does this is trusted more, second-guessed less, and given more latitude in the difficult conversations than the team that gives confident single numbers and misses most of them. The honest forecast is also the persuasive one.
Pair this with Poisson and natural noise, forecast accuracy metrics, composite metrics that hide the truth, the one-page MI pack, and the Excel paradox.