Forecast accuracy measures how close a prediction lands to actual bookings. Across B2B SaaS it usually reaches about 90% on new and expansion revenue, but through manual effort that goes stale. ORM targets 95% without manual adjustment, held from day 1 through day 90 of the quarter.
Forecast accuracy is the single metric that tells you whether your revenue operation is working or performing theater. Every other sales metric feeds into it. Pipeline velocity, win rate, stage conversion, deal aging. They all exist to make this one number better.
And almost nobody gets it right.
87% of enterprises missed revenue targets in 2025 (Clari Labs, 2026). Across ORM's customers, forecast accuracy on new and expansion revenue usually reaches about 90%, but only through heavy manual effort that goes stale as conditions change. That gap is more than a calibration issue. It shows how revenue teams measure, manage and act on their pipeline data.
I have spent two decades building forecast models for B2B SaaS companies. The pattern is always the same. Teams measure the wrong things, update too infrequently, and confuse the forecast with a wish. This guide covers the formulas that matter, the benchmarks that set the bar, and the five changes that move teams from guessing to planning.
What Is Forecast Accuracy?
Forecast accuracy measures how close your revenue prediction was to actual results. It is expressed as a percentage, where 100% means you nailed it and anything below tells you how far off you were.
The concept is simple. The execution is where companies go sideways.
Most organizations treat the forecast as a single number produced once per quarter and judged at the end. That is like checking your speedometer only when you arrive at the destination. The value of forecast accuracy is not in the final grade. It is in the weekly signal that tells you whether you are on track, off track, and what to do about it.
Forecast Accuracy Formula
The standard formula:
Forecast Accuracy = 1 - (|Forecast - Actual| / Actual) x 100If you forecast $1M and close $900K, your accuracy is 90%. If you forecast $1M and close $1.2M, your accuracy is 80%. Note that over-forecasting and under-forecasting both count as misses. Beating the number by 20% is not a win from a forecasting perspective. It means your model did not capture what was happening in the pipeline.
Here is a quick reference for the variations you will encounter:
| Formula | What It Measures | When to Use | ||
|---|---|---|---|---|
| 1 - (\ | Forecast - Actual\ | / Actual) x 100 | Overall accuracy | Quarterly and annual reviews |
| (Forecast - Actual) / Actual x 100 | Directional bias (over vs under) | Identifying systematic optimism or pessimism | ||
| Mean Absolute Percentage Error (MAPE) | Average error across periods | Comparing accuracy across segments or time | ||
| Weighted Forecast Error | Error weighted by deal size | Enterprise pipelines where deal size varies widely |
The directional formula matters because the direction of the miss reveals the cause. Consistent over-forecasting means your pipeline is inflated with deals that will not close. Consistent under-forecasting means your model is not capturing deal momentum or you have a sandbagging problem.
Forecast Accuracy Calculator
You do not need a data science team to start measuring this. Drop your forecast and actual for the last few quarters into a forecast accuracy calculator and you get your absolute accuracy, your directional bias, and a benchmark against where B2B SaaS teams typically land. The point of running it quarter over quarter is not the grade. It is the trend line. A team climbing from 72% to 81% to 88% is building a system that learns. A team bouncing between 70% and 90% with no slope is guessing well some quarters and badly in others.
Over-Forecasting vs Under-Forecasting
The direction of the miss matters as much as the size. Over-forecasting and under-forecasting are both accuracy failures, but they have opposite causes and opposite fixes.
Over-forecasting means you predicted more than you closed. The usual cause is an inflated pipeline: deals that look real on the board but were never going to close this quarter. Chronic over-forecasting is a qualification and pipeline-hygiene problem. The fix is stricter stage criteria and confronting stale deals earlier. Under-forecasting means you closed more than you predicted. Beating the number feels good, but it is still a miss. It usually points to sandbagging, where reps hold deals back to protect themselves, or a model that does not capture late-arriving, fast-closing deals. The fix is rebuilding trust in the commit number and measuring intra-quarter contribution so the model accounts for deals that arrive and close inside the period. Forecast attainment is the related view your board will ask for: actual divided by the original forecast or plan, expressed as a percentage. Attainment above 100% means you under-forecasted, below 100% means you over-forecasted, and the goal is to land predictably close to 100% rather than to swing.Worth grounding this in ORM's own numbers. Across our customer base, forecast accuracy on new and expansion revenue, excluding renewals, usually lands around 90 percent, but teams reach it through heavy manual effort that goes stale the moment conditions shift. ORM targets 95 percent without manual adjustment, and holds it from day 1 through day 90 of the quarter, updating as the quarter progresses. Getting there takes a fully trained model built on your own historical sales performance, which is a 4 to 6 week exercise, not a switch you flip.
Why Forecast Accuracy Matters More Than Ever
Three market shifts have made forecast accuracy the defining RevOps metric of this era.
The Margin for Error Has Collapsed
The B2B new-logo win rate is about 19%, across 655,000 opportunities (Ebsta/Pavilion, 2025). Sales cycles have lengthened 22% since 2022 (Optifai, 2026), and the benchmarks for shortening that cycle have shifted with it. When fewer deals close and each one takes longer, the cost of a forecast miss goes up. You cannot recover from a bad Q1 forecast by accelerating Q2 deals when those deals take six months to close.
Board and Investor Expectations Have Tightened
In the era of efficient growth, revenue predictability has replaced revenue growth as the primary valuation driver. A company whose forecast swings widely from quarter to quarter gives investors a reason to discount its plan. The forecast is no longer an internal planning tool. It is a credibility metric with external stakeholders.
The Data to Get It Right Now Exists
Companies with weekly pipeline velocity tracking achieve 87% forecast accuracy versus 52% for teams that track irregularly (Digital Bloom, 2025). The gap is not about having better data. It is about using the data that already exists in your CRM with the right cadence and methodology.
How to Measure Forecast Accuracy
Measuring forecast accuracy requires three decisions: what to measure, at what level, and how often.
Start with Your Day 1 Forecast
The Day 1 forecast is the first forecast made for the quarter. It is the most important forecast to measure because it sets the expectation. The formula: (Day 1 Forecast - Actual Sales) / Actual Sales x 100%. A negative result means you under-forecasted. A positive result means you over-forecasted. If you are routinely within 10% on your Day 1 forecast, you are performing well.
On the first day of a quarter, your forecast based on existing pipeline is your inter-quarter forecast. Deals that are not in the pipeline on Day 1 but will arrive and close within the quarter form your intra-quarter forecast. Total Forecast = Inter-Quarter + Intra-Quarter + Already Won. The proportion depends on your sales cycle length. If you have a 45-day sales cycle, 50% of your forecast might be intra-quarter. If your sales cycle is 9 months, nearly all should come from existing pipeline. If this ratio shifts quarter to quarter, it signals a change in your business that deserves attention.
What to Measure
Track accuracy at three levels:
1. Total revenue. The number your CFO and board care about. This is the headline metric. 2. By segment. Enterprise, mid-market, and SMB behave differently. A blended accuracy of 80% can hide an enterprise forecast that is off by 40% and an SMB forecast that is nearly perfect. 3. By rep or team. Individual accuracy reveals coaching opportunities. A rep who consistently over-forecasts by 30% has a different problem than one who consistently under-forecasts by 10%.
At What Level of Granularity
The most useful accuracy measurement is at the deal level, rolled up to segment and total. This means comparing each deal's forecasted close date and amount against actual outcomes. Deal-level accuracy exposes patterns that aggregate numbers hide.
For example, a team might hit 85% total accuracy but have systematic errors in deal timing. They close the right amount of revenue, but not the deals they predicted. That looks fine on the scorecard. It is terrible for resource planning.
How Often
Weekly. This is non-negotiable for B2B SaaS companies with sales cycles above 30 days. Companies that review forecast accuracy weekly can course-correct. Companies that review quarterly can only post-mortem.
The weekly review compares the current forecast to the prior week's forecast and to the actual run rate. If this week's forecast shifted by more than 10% from last week without a clear cause (a large deal closing or dying), the forecast is not data-driven. It is a guess that changes with rep sentiment.
Forecast Accuracy Benchmarks for B2B SaaS
These are the benchmarks we can stand behind, from ORM's own customer base and from published research we checked at the source:
| Benchmark | Figure | Source |
|---|---|---|
| Typical accuracy on new and expansion revenue, excluding renewals | About 90%, reached through heavy manual effort | ORM, across its customers |
| ORM's target | 95% without manual adjustment, from day one to day 90 | ORM |
| Enterprises that missed their 2025 revenue targets | 87% | Clari Labs, 2026, a competitor's research arm |
| Accuracy with weekly pipeline velocity tracking, against irregular tracking | 87% against 52% | Digital Bloom, 2025 |
Measure your own number before you compare it with anyone else's. Separate new and expansion revenue from renewals, since renewals are mostly known in advance and inflate any accuracy figure they are blended into. Then check accuracy at day one, mid-quarter and in the final weeks. A forecast that is only right in week 12 arrived too late to use.
Five Changes That Move Forecast Accuracy from 60% to 90%+
1. Replace Rep Judgment with Deal Signals
The single biggest source of forecast error is rep judgment. Reps are optimistic by nature and training. They overweight their most recent conversation with a prospect and underweight structural signals like deal age, stakeholder count, and competitive presence.
The fix is to layer objective deal signals on top of rep input. Deals with an 11x velocity delta between top and bottom performers in the same pipeline (Ebsta/Pavilion, 2025) are not being forecast correctly if you rely on rep calls alone.
Build a deal score based on measurable activity: number of stakeholder contacts, recency of last meeting, presence of a mutual action plan, and stage-appropriate milestones completed. Use that score to weight, not replace, the rep's call.
2. Track Leading Indicators, Not Lagging Ones
Most forecast models over-index on close dates and stage progression. Both are lagging indicators. By the time a deal slips its close date, the miss was baked in weeks earlier.
Leading indicators that predict outcomes:
- Stakeholder engagement depth. The typical B2B buying decision now involves 13 internal stakeholders and nine external influencers (Forrester, 2026). A deal that runs through one contact is exposed to everyone you have not met. - Activity recency. Deals with no activity in the past 14 days have a dramatically lower close probability. This is your early warning system. - Stage velocity. Is the deal moving through stages at or above the historical average? Deals that slow down rarely speed back up. - Champion activity level. Is the internal champion responding, scheduling follow-ups, and sharing materials internally? Or have they gone quiet?
If you shift 30% of your forecast inputs from lagging to leading indicators, accuracy improves immediately. Not because the model is smarter. Because you are looking at what predicts the outcome instead of what describes what already happened.
3. Implement Weekly Pipeline Hygiene
The 87% vs. 52% accuracy gap between weekly and irregular tracking (Digital Bloom, 2025) is the most compelling data point in all of revenue operations. It is also the easiest to act on.
Weekly pipeline hygiene means three things:
1. Every deal has a validated next step. If there is no next meeting or action scheduled, the deal is stale. Flag it. 2. Stage criteria are enforced. A deal does not move to Stage 3 because the rep feels good about it. It moves because it has met defined exit criteria for Stage 2. 3. Stale deals are confronted, not ignored. A deal that has been in the same stage for twice the historical average needs a decision. Recommit with a specific recovery plan or move it out.
This is not about more process. It is about replacing the fiction in your pipeline with facts. 76% of organizations say less than half their CRM data is accurate (Validity, 2025). Weekly hygiene is how you fix that, one pipeline review at a time.
4. Build Probabilistic Forecasts, Not Point Estimates
A forecast that says "we will close $2.3M this quarter" is a point estimate. It is guaranteed to be wrong. The question is how wrong.
A probabilistic forecast gives you a range: "$1.9M at the floor, $2.3M at commit, $2.7M at best case." Each number is backed by a probability distribution derived from deal-level data.
The commit number should close 80% of the time. If your commit number only lands 50% of the time, it is not a commit. It is a hope.
To build this, you need historical close rates by stage, weighted by deal characteristics. A $200K deal in Stage 3 with two stakeholders and a 90-day age has a different close probability than a $200K deal in Stage 3 with five stakeholders and a 30-day age. Your model should reflect that.
5. Close the Feedback Loop
The difference between a forecast that improves and one that does not is the post-quarter review. After every quarter, answer three questions:
1. Which deals did we forecast to close that did not? Why? Was it a data problem, a judgment problem, or a timing problem? 2. Which deals closed that we did not forecast? Why? Are there patterns the model is missing? 3. What would we change in the methodology based on this quarter's results?
Most companies skip step three. They note the miss, adjust next quarter's targets, and run the same process. That is how you stay at 70% accuracy for five years.
The companies at 90%+ treat every quarter as a calibration event. They adjust stage probabilities, update velocity benchmarks, and refine the deal scoring model. The forecast gets better because they build a system that learns.
Two Levers Most Teams Miss
Weekly seasonality. Revenue does not arrive evenly across a quarter. ORM breaks every customer's quarter into 13 weekly weights. In one example curve, only 34% of the quarter has closed by the end of week six, against 46% on a straight line. A forecast that assumes a straight line reads a normal quarter as a miss in week six and a surprise in week 13. The full curve is in the 13-week quarter. Weights that fit the deal. Stage weights only work when every stage has strict entry and exit criteria. Most companies also apply one set of weights to new business, expansion and renewals, and to enterprise and commercial deals alike. Those deals close at very different rates. Separate them and the forecast stops averaging away the differences that matter.Common Forecast Accuracy Mistakes
Five patterns I see repeatedly:
Treating the forecast as a target, not a prediction. When the forecast becomes a political number that reps negotiate rather than a data-driven prediction, accuracy disappears. The forecast should describe reality, not aspirations. Ignoring pipeline age. A $5M pipeline with 40% of deals stale for 30+ days is not a $5M pipeline. It is a $3M pipeline with $2M of noise. Until you account for deal slippage, your forecast will over-predict. Forecasting once per month. Monthly forecasts miss the week-to-week shifts that determine the quarter. By the time you catch a problem in the second monthly review, you have lost four weeks of recovery time. Using a single model. No single forecasting method captures all the dynamics of a B2B pipeline. The best teams blend stage-weighted probability, velocity analysis, and deal-level scoring. For more on this, see the sales forecasting complete guide. Not segmenting. A blended forecast hides segment-level problems. Always break accuracy down by deal size, source, and team.Forecast Accuracy and Revenue Predictability
Forecast accuracy is the input. Revenue predictability is the output.
Revenue predictability means the company can commit a number to the board and hit it. It means the CFO can plan headcount, marketing spend, and cash flow without building in a 20% buffer for forecast error. It means the CEO can make commitments to investors that hold.
For companies between $100M and $1B ARR, the shift from 70% to 90% forecast accuracy is often the difference between a 3x and a 5x revenue multiple at exit. Acquirers and investors price predictability. A company that consistently hits within 5% of forecast commands a premium over one that swings 15-20% quarter to quarter, even if the total revenue is the same.
The path from a manually built 90% to a 95% target that holds all quarter is mostly a matter of discipline. The tools exist, as our survey of the best RevOps tools lays out. The data exists. What is missing is the discipline to use them weekly, the willingness to confront stale pipeline, and a forecast methodology that treats every quarter as a calibration event.
How Accuracy in Forecasting Can Be Measured
Accuracy in forecasting can be measured by four formulas, each answering a different question about how well the prediction held up. Treating any one of them as the single metric is the mistake that causes most forecast post-mortems to miss the point.- Absolute accuracy: `1 - (|Forecast - Actual| / Actual) × 100`. The headline number used by the board. - Directional bias: `(Forecast - Actual) / Actual × 100`. A positive number means systematic over-forecasting, a negative number means systematic under-forecasting. Direction is more diagnostic than magnitude. - Mean Absolute Percentage Error (MAPE): average of `|Forecast - Actual| / Actual` across periods. Useful for comparing teams or segments over time. - Weighted Forecast Error: error weighted by deal size. Critical for enterprise pipelines where one $5M deal swings the absolute number more than five $1M deals.
The right way to use these is in combination. A team running 85% absolute accuracy with a consistent positive directional bias has a sandbagging culture. A team at 85% with a consistent negative bias has a discipline problem at the commit stage. A team with high MAPE but acceptable absolute accuracy has compensating errors that average out by quarter-end but make in-quarter resource decisions impossible. See forecast variance for how the related concept of variance from plan is used at the board level, and the commit vs. best case glossary entry for the standards each forecast category should meet.
Measuring Forecast Performance Over Time
Forecast performance, measured across quarters, tells a different story than any single quarter's accuracy number. A quarter at 92% accuracy looks excellent in isolation. Eight quarters bouncing between 70% and 95% with no improvement curve is a system that is not learning.Three views to maintain quarter over quarter:
- Day 1 forecast vs. actual. This is the cleanest performance signal because Day 1 is the least influenced by in-quarter signals. A Day 1 forecast that lands within 10% of actual is a forecast methodology that works. - Inter-quarter (Day 1 pipeline) vs. intra-quarter (deals that arrived and closed in-quarter) contribution. A shift in this ratio is one of the earliest signals of a change in the business. If intra-quarter contribution grows from 30% to 60% over four quarters, something has changed in either the sales cycle length or in how the team is being incentivized to backload bookings. - Manager-by-manager and rep-by-rep trailing accuracy. Aggregate accuracy hides the manager whose commit is reliable from the manager whose commit is fictional. Trailing four-quarter accuracy by leader is the cleanest way to allocate where coaching attention belongs.
Tracking these three views turns forecast accuracy from a single quarterly grade into an operating system that compounds. Each quarter the methodology adjusts in response to what the prior quarter exposed, and the absolute number floor rises. For a structural view of how forecast performance fits into the broader revenue operations function, see revenue operations KPIs and the best sales forecasting tools for how the right tooling supports the discipline.
Frequently Asked Questions
How do you calculate forecast accuracy?
Forecast accuracy = 1 - (|forecast - actual| / actual), shown as a percentage. If you forecast $1 million and close $900,000, accuracy is 90%. Track it every quarter and by segment.
What is a good forecast accuracy rate?
Across B2B SaaS, accuracy on new and expansion revenue, excluding renewals, usually reaches about 90%. Most teams get there through heavy manual work that goes stale fast. ORM targets 95% without manual adjustment, from day one to day 90.
What causes forecast inaccuracy?
Most often, a model built on old assumptions after the market moved. Deal sizes shrink, win rates drop or cycles stretch, and the model does not notice. Leaning on rep judgment and counting stale deals make it worse.
What is the best early signal that a forecast will miss?
A rep changing a close date. A deal that slips from one quarter to the next is less likely to close, even in commit. Earlier still is silence: no activity, no changing data and no notes on a deal.
When in the quarter should a forecast be accurate?
From day one. A model that only gets it right in the final week has described the quarter rather than forecast it. Snapshot the forecast at days 1, 30, 60 and 90, and compare each with the actual.
How long does it take to build an accurate forecast model?
About four to six weeks for a model fully trained on your own sales history. It trains on the imperfect data you already have, so a big cleanup first is not required.
Five free days of implementation
Start with ORM and your first five days of implementation are free. We build your forecast model on your live pipeline, then you decide.
Claim your five days