Sales forecasts across most organizations lean high, and the bias repeats quarter after quarter. The causes are structural, which is why exhorting people to be realistic changes little.
The forecaster has an interest in the number
A representative reporting a weak pipeline invites scrutiny, coaching and pressure. Reporting a strong one buys time until the deals are due to close.
Managers aggregating those numbers face the same incentive one level up. A department forecasting a shortfall early attracts attention it would rather avoid.
The bias is therefore introduced at every level and compounds upward, without anyone deliberately misstating a specific deal.
Stage-based probabilities describe activity, not likelihood
Common practice assigns a fixed conversion probability to each pipeline stage. Advancing a deal to the next stage automatically raises its weighted value.
Since advancement is usually recorded by the representative and is often triggered by an internal action such as sending a proposal, the number rises on effort rather than on buyer commitment.
Probabilities derived from historical conversion by stage are more defensible, but only if the stage definitions describe verified buyer behavior rather than seller activity.
Losses are recognized late
Deals rarely end with a clear refusal. They go quiet, and a quiet deal is easier to leave open than to mark closed and lose from the pipeline.
Aging opportunities therefore accumulate, inflating totals with opportunities that ended months earlier without being recorded.
Rules requiring an opportunity to show recent buyer activity to remain open remove much of this, at the cost of a one-time drop in reported pipeline that has to be explained.
Timing is harder to predict than outcome
Representatives are often reasonably accurate about which deals will eventually close and badly wrong about when. Close dates cluster at quarter end because that is where they were placed, not where buyers were.
Buyer-side steps that cause slippage, including legal review, budget cycles, security assessment and signature authority, sit outside the seller's control and are usually underestimated.
Asking what specific event must occur next, and who owns it, produces better dates than asking for a confidence percentage.
Judgment and history are best used together
A forecast built only from the pipeline inherits every bias above. One built only from historical run rate ignores what is genuinely in play.
Comparing the two exposes the gap. Persistent divergence in one direction quantifies the organization's bias and can be corrected for explicitly.
The useful discipline is recording forecasts and comparing them to outcomes by individual and by segment. Without that record, the same optimism is repeated indefinitely because nobody can measure it.