Why Your Sales Forecast Is Unreliable — and How to Actually Fix It

The setup you have — RevOps analysts working with sales leaders to clean the data and build the forecast — is one of the most common configurations in mid-market B2B sales organisations. It is also one of the most structurally unreliable. Not because the people involved are incapable, but because the process has too many points where human judgement, optimism, and political pressure can enter and distort the output.

The result is a forecast that feels rigorous — there are spreadsheets, there are conversations, there are people whose job it is to do this — but consistently misses. Understanding why it misses is the prerequisite to fixing it.

Why Manual Forecast Assembly Breaks

Failure point 1 — The data being assembled is already wrongWhen RevOps analysts update CRM data in conjunction with sales leaders, they are correcting a symptom rather than fixing a root cause. The underlying problem is that reps are not maintaining their own deal records accurately in real time. Close dates are aspirational. Stage assignments reflect the rep's optimism rather than verifiable deal criteria. ARR figures may not reflect the most recent commercial conversation. Every adjustment a RevOps analyst makes to bring a record in line with reality is evidence that the system is not being used as a system of record — it is being used as a compliance form filled out before the forecast meeting.

Failure point 2 — The forecast inherits the bias of the conversationWhen a sales leader and a RevOps analyst discuss a deal to determine its forecast inclusion, the output of that conversation is shaped by the sales leader's confidence, communication style, and relationship with the analyst. A sales leader who is consistently bullish will consistently push deals into the forecast that do not close. A sales leader who manages upward effectively will frame weak deals compellingly. The analyst, who does not own the deal, has limited basis on which to push back. This is not a personnel problem — it is a structural one. Human conversation is a poor filter for forecast accuracy.

Failure point 3 — Structured and unstructured data are being mixed without distinctionA reliable forecast distinguishes between structured data — objective signals captured in the CRM such as time in stage, number of contacts engaged, days since last activity, contract value, stage probability — and unstructured data, which is qualitative information from rep conversations, emails, and calls. Both are useful, but they have very different reliability properties. Structured data is auditable and consistent. Unstructured data is rich but subject to interpretation and recall bias. Most manual forecast processes blend these without acknowledging the difference, producing a number that looks data-driven but is substantially based on rep narrative.

Failure point 4 — ARR timing is inconsistentWhen does a deal count in the forecast? When the rep says it's going to close this quarter? When the contract is sent? When it is signed? When the first invoice is issued? When revenue recognition begins? Different organisations — and sometimes different people within the same organisation — answer this question differently. When ARR timing is not defined with precision and enforced consistently in the CRM, the forecast number means different things to different stakeholders. A deal that closes on the last day of the quarter may be in the CFO's forecast but not the finance team's recognised revenue. These are not the same number and conflating them produces planning errors.

Structured vs Unstructured Data: Why the Distinction Matters

Most forecast processes rely heavily on unstructured data — rep calls, manager conversations, deal narrative — because it feels more informative. A rep who says "I spoke to the economic buyer last week and they're ready to move" is providing context that no CRM field captures. That context is genuinely valuable. The problem is that it is also the easiest category of information to get wrong, to misremember, and to selectively emphasise.

Structured CRM data — stage, time in stage, days since last contact, number of stakeholders engaged, contract value, close date history — is less rich but far more reliable. It is auditable. It does not change based on who is telling the story. And crucially, it can be modelled: organisations with sufficient deal history can identify which combinations of structured signals are predictive of actual close, and weight their forecasts accordingly.

The practical implication is this: your forecast should be built primarily on structured data, with unstructured qualitative context used to adjust individual deals at the margin — not to construct the baseline. If your RevOps team is spending most of its forecast cycle collecting qualitative updates from sales leaders, the process is inverted. The baseline should come from the system. The human conversation should challenge and refine it, not construct it.

ARR Recognition Timing: Getting the Definition Right

Before you can measure forecast accuracy, you need an unambiguous definition of what counts as ARR and when. The four most common definitions — and the problems each creates if used inconsistently:

Commit date: The quarter the rep commits the deal will close. This is the most optimistic definition and consistently overstates achievable ARR. It is useful for pipeline management but not for financial planning.

Signature date: When the contract is signed by both parties. More reliable than commit date, but still subject to last-day-of-quarter compression and commercial terms that may delay revenue recognition.

Billing start date: When the customer is first invoiced. More conservative and more directly linked to cash flow, but may differ significantly from signature date on enterprise deals with complex commercial structures.

Revenue recognition date: When revenue is recognised under your accounting policy — typically when the service is delivered or when a performance obligation is satisfied. This is the number that matters to finance and to your P&L. It may be substantially later than signature date on multi-year or implementation-heavy deals.

The recommendation is to maintain both a sales forecast (signature date based) and a revenue forecast (recognition based), make both visible in the CRM, and ensure the CFO and CRO are aligned on which number each report is using. The most common source of forecast distrust between finance and revenue leadership is that they are looking at the same deal with different timing assumptions and producing different numbers — both of which are technically correct.

How to Measure Forecast Accuracy Honestly

What a Trustworthy Forecast Process Looks Like

The target state is a forecast that is generated primarily by the system, reviewed and refined by humans, not assembled by humans from scratch. In practice that means: the CRM generates a baseline forecast from structured signals at the start of each week. Sales leaders review deals where the system forecast diverges significantly from their own view and document the specific reason for the difference. RevOps monitors the accuracy of those human overrides — are the deals the sales leader upgraded actually closing more often than the system predicted? This creates accountability for qualitative adjustments rather than treating them as free additions to the forecast.

Your RevOps analysts should be spending their time on this monitoring and analysis work — understanding where the system is systematically wrong and improving the model — not on manually updating deal records that reps should be maintaining themselves.

The data cleaning work your analysts are doing right now is valuable in the short term. In the medium term it is masking an adoption problem that needs to be fixed at the source: reps maintaining their own records accurately, in real time, because the system gives them something useful in return for doing so.