Sales forecasting:why the forecast is always wrong in the same direction
By Enrico Giubertoni
Trusted by
Worldline
Alba Leasing
L'Erbolario
3M
This article is for those who sign a number at the start of the quarter and answer for it at the end: sales directors, CEOs, CFOs. Sales forecasting fails in companies less for lack of tools than for a bias that pushes the number always to the same side. By the end you will have a usable definition of a sales forecast, the four causes behind that bias and a first signal to observe on Monday morning, without buying anything.
What sales forecasting is: an estimate with an owner
A sales forecast is the estimate of the revenue a company will obtain in a defined period, built on open deals and their probability of closing. It serves to decide today how much to spend, how much to produce and how many people to hire. Read superficially, it is an accounting exercise; a deeper reading determines that it is the most expensive decision a company takes every quarter, because all the other decisions rest on it.
The budget says how much we would like to sell. The pipeline (the set of open deals, each with a value and a stage) says how much we could sell. The forecast says how much we will sell. Three numbers answer three different questions, and the first cause of error I meet in companies is that they are treated as one.
It is useful to read the three rows together, because the last column is the one that is usually missing.
Budget, pipeline and forecast answer three different questions
Number
Which question it answers
Who produces it
What it means for you
Budget
How much do we want to sell?
Management, once a year
It is a target: if the forecast copies it, you are hoping, not predicting
Pipeline
How much could we sell?
Salespeople, every day in the CRM
It is an inventory: it is worth the discipline with which it is updated
Forecast
How much will we sell?
The sales director, every month
It is a commitment: if it is always wrong on the same side, the defect is in the method
It follows that the forecast is a judgement on the weighted pipeline, and like every judgement it has an owner who answers for it.
Why sales forecasts are wrong: systematic optimism
In sales forecasting a random error falls above and below the target and compensates over time. An error that always falls on the same side is a characteristic of the system that produces the number. In the companies I have followed, the quarterly forecast sits above the actual result with a regularity that no chance explains. The right question is therefore “why are we always wrong in this way?”, more than “how much did we miss?”.
The reference data comes from Gartner, in its press release of 12 February 2020 on the State of Sales Operations survey: only 45% of sales leaders and sellers declare high confidence in the accuracy of their organisation’s forecast. More than half of those who sign the number, therefore, do not believe it. And those who do not believe the number decide by instinct, which is the consequence Gartner itself indicates.
Four causes explain almost all the bias in sales forecasts, and all four belong to the system, outside the competence of the salespeople.
The incentive to declare. A salesperson who reports a weak deal is put under pressure immediately; one who reports it as strong is put under pressure at the end of the quarter. The rational choice is to postpone the bad news, and the forecast inherits it.
Inflated stages in the CRM (the customer management system where deals live). A deal moves to the stage “offer sent” when the offer has been sent, before the customer has read it. The probability attached to the stage is statistical; the single case is particular.
The discount nobody records. The deal value stays at list price until signature, while during negotiation the price goes down. The forecast sees the list price, revenue sees the discount.
Short memory. The last month weighs more than the last year in the mind of the person estimating, and a good March raises the April forecast with nothing in the pipeline to justify it. It is recentism applied to sales numbers, and it acts even on those who know it.
A forecast that is always wrong on the same side is designed to be wrong that way.
Forecast accuracy is measured in one quarter, without software
Forecast accuracy is the distance between the number declared at the start of the period and the actual revenue at the end, expressed as a percentage of the actual. It shows whether the method improves or deteriorates over time. The calculation fits in one line; the real work is deciding which forecast to compare: the one from the start of the quarter, never the one from the last week, which has already seen most of the contracts.
The calculation itself is trivial. What decides whether the measure says something lies elsewhere: which number you compare, at which level of detail and for how many periods. A badly designed measure acquits the method it should judge, and for this reason we define it together in the first meeting. One thing is valid for everyone: look at the sign of the gap before its size.
Marketing and sales produce two different forecasts of the same quarter
The sales forecast is one of two forecasts circulating in the company. Marketing has its own, built on the contacts generated and on the expected conversion rate, and the two almost never coincide. When I measured the synchrony between the two functions in an industrial services company with recurring fees, business-to-business, over 50 million in revenue in its reference market, the defect was in one word: “lead” meant a name and a contact detail for marketing and a qualified contact for sales. Two definitions, two conversion rates, two forecasts, each correct within its own perimeter.
From this it follows that part of the systematic error originates upstream of the sales force, where marketing numbers enter the pipeline without being translated. It is the question of data that marketing and sales read differently along the customer journey, and when the two departments write the same definition on the same line, the forecast improves before anyone touches the CRM.
Average value per contract grew by 75%, with the qualified contacts generated practically unchanged.
Measurement of 16 July 2026, first half 2026 against first half 2025, in an industrial services company with recurring fees, business-to-business, over 50 million in revenue in its reference market. Perimeter: core business line, paid search network. The merit is distributed across several causes, and I do not declare a link that the data does not prove.
The number is here for a precise reason: in the same company, before the intervention, the forecast built on marketing contacts overestimated the quarter exactly because it counted contacts that sales would never call. Aligning the definition made the declared pipeline smaller and the forecast more reliable.
The forecast is corrected in the system, not in the number
When the forecast has been too high for four quarters, the temptation is to apply a coefficient: take the sales force’s number and cut it by 20%. It works once. From the second time, the sales force knows and declares 20% more. You correct the number and worsen the system.
Unreliable sales forecasts almost always show the same symptoms, and whoever signs the forecast recognises them before measuring them.
Deals advance a stage for what the salesperson did, while the customer did nothing: “offer sent” instead of a customer who has confirmed the budget in writing.
The deal value stays at list price until signature, even after three price concessions.
Nobody remembers precisely what was declared at the start of the quarter, because the number is revised every week.
Marketing and sales call two different things a “qualified contact” and have never noticed.
When these symptoms coexist, the forecast is a system problem, and no sales course will solve it. It is a problem of cash that arrives later than the number promised, and the first to pay is the CFO, who has already spent on the basis of that promise. How much each symptom weighs, and in which order to intervene, depends on the company: it is the reading I do with management in the work of Organizational Synchrony (the synchrony between the functions that feed the number).
New software is not the answer. Gartner, in a survey of 303 sales leaders conducted in July 2023 and published on 6 February 2024, reports that 84% of them acknowledge that sales analytics has had less influence on results than management expected. The tool is there. What is missing is the discipline that makes the tool credible, and discipline is a management decision, not a purchase.
What to do on Monday morning
One single action, of observation, that one person can carry out in less than an hour: retrieve the forecast declared at the start of each of the last four quarters and write the actual result next to it. Look only at the sign of the gap.
What that sign means in your case, where it originates and what to correct first is another question, and the answer changes from one company to another. The rest of the verification, with the questions to ask the pipeline and the two functions, is in the Forecast Check-up, the first step of consulting for the optimisation of marketing and sales performance.
The budget is the sales target that management sets at the start of the year, and it does not change with the progress of negotiations. The forecast is the estimate of what will actually be sold in a period, updated on the real pipeline. When the two numbers always coincide, the forecast is not predicting: it is repeating the budget.
How does a forecast differ from a pipeline?
The pipeline is the inventory of open deals, each with a value and a stage. The forecast is a judgement on that pipeline: which deals will close in the period and for what amount. Salespeople update the pipeline; the sales director signs the forecast and is accountable for it.
How do you calculate forecast accuracy?
It is the distance between the forecast declared at the start of the period and the actual revenue at the end, expressed as a percentage of the actual. The sign of the gap matters more than its size: an error that always falls on the same side indicates a defect of method, and reading it in your own case is the work of the diagnosis.
Why are sales forecasts wrong almost always on the high side?
Because the people who feed the forecast have an incentive to postpone bad news, because CRM stages advance on the salesperson’s actions and not on evidence from the customer, because the discount granted during negotiation is not recorded before signature, and because the last month weighs more than the last year in the estimate. Four causes of the system, none of competence.
Do you need AI software for accurate sales forecasting?
Not as a first step. An algorithm learns from the pipeline it receives, and if stages are inflated and values are out of date it produces an optimistic forecast with more decimals. First you define verifiable stage criteria and a shared definition of a qualified contact; then, on clean data, a predictive tool can add precision.
