Which sales forecasting method should you use?
TL;DR
There's no best sales forecasting method — only the one that fits your scale, data, and stability. Weighted-stage for early teams, historical for stable mid-stage funnels, regression and AI once you have real deal volume. Every one assumes clean data underneath, so the method is the last call you make, not the first.
What choosing a method actually decides
Pick the wrong forecasting method for your stage and you don't get a rough answer — you get a precise, confident, wrong one. That's the worst kind, because it survives the meeting.
The method is downstream of two things you can't fake: how much deal volume you have, and how stable your market is. Get those honest and the choice mostly makes itself.
The data, explained
There are four methods worth knowing. Pipeline / weighted-stage multiplies deal value by stage probability — visual, cheap, fine for early teams. Historical / time-series projects forward from past periods, which works only on a stable funnel with real history behind it. Regression pulls win rate, deal size, and cycle length into one model and needs someone to maintain it. AI / predictive learns from deal signals and needs volume to be trusted.
That last point isn't a vibe. B2B pipelines are small and noisy compared with consumer datasets, so point a model at 40 deals a quarter and it learns your noise, not your business.
What that produces: the method is the last decision
Every method on the list assumes clean data and needs volume and stability. So the sequence matters more than the pick: fix the data, match the method to your stage, then run it on a cadence. We don't hunt for one universal method — industries change too fast for a single template to hold. Start from the data you have, weight the outliers, map the one-off events, build a baseline, then apply judgment. That's the playbook.
Which method by stage
| Method | Best for | Where it breaks |
|---|---|---|
| Pipeline / weighted-stage | 20–50 people, early stage | Useless without stage exit criteria |
| Historical / time-series | 50–100 people, ≥18 months clean data | Falls apart during fast growth or launches |
| Regression / multivariable | 100–150 people, dedicated RevOps | High upkeep; needs analytical resource |
| AI / predictive | 100+ people, real deal volume | Small, noisy B2B data |
Frequently asked questions
What's the simplest method that works?
Weighted-stage pipeline in your CRM, with real probabilities and stage exit criteria. Enough for most teams under 100 people.
When should we move to AI or predictive forecasting?
When you have enough clean deal volume to train a model — usually past 100 people. Before that, it's noise.
Can we combine methods?
Yes, and you should. Triangulate a weighted baseline with the rep commit and a historical sanity check.
Where we come in
We match the method to your stage and build the data foundation it needs to hold. Book a call →
Related reading
- Salesforce. “What Is Sales Velocity? (Formula and Tips).” Salesforce Blog. salesforce.com. Accessed July 16, 2026.
- Clari. “Pipeline Coverage: Best Practices for Sales Leaders.” Clari Blog. clari.com. Accessed July 16, 2026.
