
A sales forecast is a dated, numeric answer to one question: how much will we sign by the end of the period? There are five common ways to build one, and none is right alone. The forecast you can trust combines two: a deal-by-deal view based on what the buyer has done, for this quarter, and an activity-based view built from conversations and meetings, for the next one. Here is how each method works, where it breaks, and a weekly routine that keeps the number honest.
Forecasts rarely miss because of the formula. They miss for three human reasons. Reps keep deals they like in the pipeline after the buyer has gone quiet. Stages describe what the seller did ("demo done") instead of what the buyer did ("business case shared with finance"). And close dates are wishes that slide from month to month without anyone asking why.
A forecast is also only as good as the pipeline under it. If the top of the funnel was thin eight weeks ago, no method will save this quarter. That is why the fifth method below matters most to anyone running outbound.
Each rep calls a number, usually in three categories: commit, best case, pipeline. The manager adds them up and applies judgment.
Works when: the team is small, deals are few and large, and the manager knows every one of them. Breaks when: nobody checks the commit against evidence. Optimists over-call, sandbaggers under-call, and the total is an average of temperaments.
Each stage has a probability (say 20% after discovery, 60% after proposal) and the forecast is the sum of deal value times probability.
Works when: the stages have clear exit criteria and the probabilities come from your own history. Breaks when: a deal sits in "proposal" for ninety days and still counts for 60%. The method has no sense of time. We described stages with real exit criteria in SaaS pipeline stages.
Take what was signed in the same period before, adjust for growth, headcount and seasonality.
Works when: volumes are high, deals are small and the business is stable. Breaks when: anything changes: a new segment, a price increase, three new reps in ramp. In Europe, seasonality is strong enough to mislead it. August and the second half of December produce few signatures, and a run rate that ignores them will be wrong twice a year.
Each deal is scored on what the buyer has done, not on its stage. Is the economic buyer identified and met? Is there a written business case? Is there a next meeting in the calendar? Has procurement or legal been engaged? Is there a date the buyer needs the solution by, and a reason for it?
Works when: the cycle is complex and involves several people. This is what frameworks like MEDDIC formalise. Breaks when: the scoring becomes a form reps fill in to please the manager. The test is simple: every "yes" should point to an email, a meeting or a document.
Instead of asking what will close from today's pipeline, ask what today's activity will produce. Conversations become meetings, meetings become opportunities, opportunities become revenue, each at a rate you have measured, each with a delay you know.
Works when: you need to see next quarter before it starts. Breaks when: the activity data is unreliable, which is the case when calls are logged by hand.
The inputs exist. In our benchmark of 1,000,000 B2B cold calls, the average connect rate is 11.3%, and openers built on a trigger book a meeting on 7 to 10% of conversations, against 2 to 3% for pitch-first openers. With our north star of 20 conversations per SDR per day, a team's meetings for the month can be estimated in week one. We worked through the full chain in the math of outbound.
Use two, and compare them.
The rep's commit stays, as a conversation rather than a method. The useful question in a forecast call is not "what is your number?" but "what has the buyer done since last week?"
On tools: AI forecasting features in CRMs can help spot stalled deals and unrealistic close dates. They work from the data you log. If activity and next steps are missing from the CRM, the model predicts from noise.
We forecast from conversations first. Half of Sonik's pipeline comes from signals that were already in the CRM, and we know the rates from conversation to meeting and from meeting to opportunity well enough to see a weak quarter two months before it arrives.

That only works because the activity is captured without anyone typing it. In Sonik, every dial, conversation and meeting booked is logged by the dialer and synced to the CRM, so the top of the forecast is a fact rather than a declaration. When the number of conversations drops in week two, the manager sees it the same day and can act on the list or the calling slots instead of explaining a miss in week twelve.
Trust the forecast that rests on evidence: what buyers have done for this quarter, what reps are doing for the next. This week, take your five largest commit deals and write next to each one the last thing the buyer did, with its date. Any deal where that line is older than three weeks should leave the commit today.