Forecasting is the single skill that determines whether a VP Sales keeps their job. Not hiring. Not coaching. Not territory design. Forecasting. Miss your number by 20% two quarters in a row, and the board starts asking the CEO about your replacement. Nail it within 5% consistently, and you buy yourself the credibility to make every other change you need.

I've run forecasts at companies ranging from $2M ARR to $500M+. The methods change with scale, but the principles don't. You need a system that combines rep-level judgment with mathematical discipline, and you need to know which one to trust when they disagree.

This piece covers the five forecasting methods every sales leader should know, when to use each, and how to combine them into a forecast your CFO will stop second-guessing.

Method 1: Top-Down Forecasting

Top-down forecasting starts with the big numbers and works backward. Total addressable market, your current market share, growth assumptions, and capacity constraints. It's the method boards use when setting annual plans and the method most sales leaders ignore once the year starts.

That's a mistake. Top-down serves as a sanity check on everything else. If your bottom-up forecast says you'll hit $50M but your top-down analysis shows your market only supports $35M given your current win rate and deal velocity, someone is wrong. Usually it's the bottom-up forecast, inflated by optimistic reps and a pipeline full of stalled deals.

How to build a top-down forecast

  1. Addressable accounts: How many companies fit your ICP? Not TAM in dollars. Actual accounts you could sell to this year.
  2. Penetration rate: What % of addressable accounts can you reach given your current headcount and coverage model? Most companies can actively work 5-10% of their addressable market in any given quarter.
  3. Conversion rate: Of accounts you actively engage, what % become pipeline? Typically 10-20% for outbound, 20-40% for inbound.
  4. Win rate: Of pipeline created, what % closes? B2B SaaS averages 20-25% for mid-market, 15-20% for enterprise.
  5. Average deal size: Your trailing 4-quarter average, not what you hope to sell.

Multiply through: addressable accounts x penetration x conversion x win rate x deal size = your ceiling. If the plan asks for more than this number, the plan is wrong. Either you need more reps, a higher win rate, or larger deals. Pick which lever you're pulling before telling the board yes.

Method 2: Bottom-Up Forecasting

Bottom-up is where most weekly forecasting happens. Rep by rep, deal by deal, opportunity by opportunity. It's the most accurate method for near-term quarters (current and next) and the least accurate method for anything beyond that.

The three categories

  • Commit: Deals the rep stakes their credibility on. These should close barring something unexpected. Target accuracy: 90%+.
  • Best-case: Commit plus deals with a realistic path to closing this quarter. One or two things need to go right. Target accuracy: 60-70%.
  • Upside: Everything that could theoretically close. Early stage, stalled, pending budget approval. Target accuracy: 20-30%.

The formula most VP Sales use: forecast = commit + (best-case minus commit) x 0.5 + (upside minus best-case) x 0.1. This weights the categories by probability and gives you a number that's usually within 10% of actual, assuming honest rep input.

Honest rep input is the hard part. Reps sandbag when they don't trust their manager not to raise their quota. Reps over-forecast when they're afraid of being seen as underperformers. The only fix is consistency: if you punish a rep for a conservative forecast that they beat, you've trained the entire team to inflate their numbers forever.

Method 3: Weighted Pipeline

Weighted pipeline assigns a probability to each stage in your sales process and multiplies it by deal value. It's the most commonly used method in CRM-based forecasting and the most commonly wrong method in CRM-based forecasting.

Stage Typical Probability What Should Happen Here
Discovery 10% Qualified need confirmed, stakeholders identified
Evaluation 25% Demo completed, requirements documented
Proposal 50% Pricing sent, decision criteria agreed
Negotiation 75% Redline in progress, verbal commitment received
Verbal Close 90% Signature expected within 2 weeks

The problem: these probabilities are averages, and averages lie. A $500K enterprise deal at the proposal stage and a $25K SMB deal at the proposal stage do not have the same probability of closing. Deal size, buyer seniority, competitive dynamics, and quarter-end timing all affect probability. Using stage alone misses all of that context.

Fixing weighted pipeline

Adjust probabilities by segment. Enterprise deals should carry lower probabilities at every stage than SMB deals. Add time-based decay: a deal that's been in "negotiation" for 8 weeks should have its probability reduced by 10-15% regardless of what the stage says. Track your trailing conversion rates by stage and update the model quarterly. Most companies set these probabilities once and never validate them against actual outcomes.

Method 4: Historical/Run-Rate Forecasting

Historical forecasting looks at trailing performance and projects it forward. It's the simplest method and often the most accurate for stable businesses with predictable seasonality.

The basic formula: take last quarter's bookings, apply a growth rate (either the trailing growth rate or the plan's growth rate), and adjust for known seasonality. Q4 in B2B is typically 30-40% of annual bookings. Q1 is usually the weakest at 15-20%. If your Q4 was $12M and your annual plan is $40M, a Q1 run-rate of $7-$8M is realistic.

When historical forecasting works

  • Stable team with low turnover (team changes make history unreliable)
  • Consistent deal mix (no major product launches or market shifts)
  • 3+ quarters of data at the current headcount level

When it fails

  • Post-layoff or rapid hiring (the team is different from the one that produced the data)
  • New market or product launch (no history to reference)
  • Pricing changes (historical conversion rates don't apply at a new price point)
  • Macro shifts (a recession makes trailing data misleadingly optimistic)

Method 5: AI-Assisted Forecasting

AI forecasting tools like Clari, BoostUp, and Aviso analyze CRM data, email activity, meeting patterns, and deal velocity to predict outcomes. They're getting better. They're also getting oversold.

The promise: AI looks at signals humans miss. A deal where email response times are increasing, meeting frequency is dropping, and the champion just changed their LinkedIn title to a role at another company. Those are signals that should lower the forecast. A human reviewing 200 deals per week can't catch all of them. AI can.

The reality: AI forecasting is only as good as your CRM data. Companies with disciplined stage management, consistent activity logging, and clean close dates see 10-15% accuracy improvement over rep-based forecasts. Companies where reps update Salesforce once a week from memory see minimal improvement because the model is training on garbage.

What AI forecasting needs to work

  • 12+ months of historical CRM data with accurate close dates and amounts
  • Email and calendar integration (activity data is where most predictive signal lives)
  • Consistent stage definitions that reps follow (not 5 different interpretations of "proposal sent")
  • Minimum deal volume: 100+ closed deals per quarter for the model to have statistical significance

If you have fewer than 50 deals per quarter, save your money. Statistical models need volume. An enterprise team closing 15 deals per quarter won't get meaningful AI predictions because the sample size is too small for pattern recognition.

Pipeline Coverage Ratios: The Number Your Board Cares About

Pipeline coverage measures how much pipeline you have relative to your quota. It's the leading indicator that predicts whether you'll hit your number weeks before the quarter ends.

Benchmarks by segment

  • SMB (velocity model): 2.5x-3x coverage
  • Mid-Market: 3x-3.5x coverage
  • Enterprise: 3.5x-4x coverage
  • Strategic/Global: 4x-5x coverage

How to calculate it: total qualified pipeline value divided by remaining quota for the period. A $5M quarter with $15M in pipeline = 3x coverage. The word "qualified" matters. Pipeline stuffed with deals that haven't had a discovery call or don't have a confirmed budget isn't pipeline. It's a wish list.

Track coverage weekly and trend it. If you start the quarter at 3.5x and it's dropping by week 4 to 2.8x without corresponding bookings, you're going to miss. That 0.7x gap represents deals that stalled, pushed, or were lost. Early detection gives you time to respond with pipeline generation sprints, deal acceleration tactics, or an honest conversation with the CFO.

Putting It Together: A Multi-Method Forecast

The best forecasts use multiple methods and reconcile the differences. Here's a framework that works from Series B through public company:

  1. Weekly: bottom-up forecast from rep deal reviews (commit / best-case / upside)
  2. Weekly: weighted pipeline calculation from CRM data
  3. Monthly: historical run-rate comparison (are we tracking ahead or behind last quarter/year?)
  4. Quarterly: top-down sanity check against capacity and market constraints
  5. Continuous: AI-assisted predictions as a fourth input (if deal volume supports it)

When bottom-up and weighted pipeline agree within 10%, you have a high-confidence forecast. When they diverge by 20%+, dig in. Either reps are sandbagging their commits (bottom-up too low), the weighted model has stale deals inflating the number (weighted too high), or there's a timing issue where deals will close but not in this quarter.

Forecasting Cadence and Communication

Your board wants a quarterly forecast. Your CEO wants a monthly update. Your team needs weekly accountability. Build a cadence that serves all three without creating a forecasting bureaucracy.

Weekly deal review (Monday)

30-minute session per manager. Review every deal in commit and best-case. Challenge assumptions. Update probabilities. This is where forecast accuracy is built or destroyed. A manager who rubber-stamps rep commits is a manager who will miss the quarter.

Weekly forecast submission (Tuesday)

Updated commit, best-case, and upside numbers to the VP/CRO. Include a 2-sentence narrative: "Commit up $200K from last week due to verbal on the Acme deal. Best-case down $150K, lost CompetitorX deal at negotiation stage."

Monthly board update

Forecast vs. plan, pipeline coverage, win rate trends, and risk/upside factors. Keep it to 3 slides. Don't explain every deal. Explain the patterns.

The Psychology of Forecasting

Forecasting is a social system, not just a mathematical one. Reps forecast based on what they think their manager wants to hear. Managers forecast based on what they think the VP wants to hear. The VP forecasts based on what the board expects. This compression of honesty up the chain is why most companies over-forecast.

Fix it with two practices. First, never punish a conservative forecast that beats. If a rep commits to $400K and closes $500K, that's an A+ quarter for forecasting and an A+ quarter for selling. Celebrate both. Second, create a "no-judgment" pipeline review where reps can flag at-risk deals without fear of consequences. The deal they're afraid to call out in week 6 is the one that blows up your forecast in week 12.

The goal isn't perfect accuracy. It's consistent accuracy. A VP Sales who calls the number within 5% every quarter, even if that number is below plan, maintains credibility. A VP Sales who swings between 80% and 120% of forecast, even if average attainment is fine, loses credibility because the business can't plan around unpredictability.