Across roughly 90 venture-backed companies we have worked with, fewer than 1 in 5 could tell us their forecast accuracy on the day we started. Almost all of them had a forecast. Almost none of them measured how often it was right.

That gap matters because forecast accuracy is the one KPI that tells an investor whether every other number in your deck can be trusted. A burn multiple of 1.4x means little if last quarter's burn forecast missed by 30%. This post defines the forecast accuracy KPI precisely, shows the formula with a worked example, gives benchmarks by stage, and walks through how a partner reads the number during diligence.

What the forecast accuracy KPI measures

Forecast accuracy measures how close a forecast made at a specific point in time came to the actual result for a specific period. It is always expressed with three parameters:

1. The metric being forecast. Revenue, new ARR, net burn, ending cash, headcount, or opex.
2. The horizon. How far ahead the forecast was made. A one-month-ahead forecast and a four-quarter-ahead forecast are different KPIs.
3. The version. Which forecast you are grading. The board-approved annual plan, the most recent monthly reforecast, or a rolling forecast.

If any of these is missing, the number is not comparable across periods or companies. "We hit 92% forecast accuracy" means nothing until you add "on net burn, one quarter ahead, against the reforecast locked on the first business day of the quarter."

The forecast accuracy formula

There are three calculations finance teams use. Each answers a different question. Use all three.

1. Simple accuracy for a single period

This is the headline number most boards see.

Forecast Error % = |Actual - Forecast| / |Actual|
Forecast Accuracy % = 1 - Forecast Error %

Divide by actual, not forecast. If you forecast $1,000,000 of revenue and land at $900,000, the error is $100,000 divided by $900,000, or 11.1%, and accuracy is 88.9%. Dividing by the forecast gives 10% and flatters the miss. Investors who build their own models will divide by actual, so match them.

2. Mean absolute percentage error across periods

A single quarter can be lucky. MAPE averages the absolute error across several periods so one good month cannot hide three bad ones.

MAPE = (1 / n) x Sum of |Actual(t) - Forecast(t)| / |Actual(t)|
Forecast Accuracy (MAPE basis) = 1 - MAPE

Use the absolute value of each period's error before averaging. Without it, a 15% overforecast in January and a 15% underforecast in February average to zero error, which is false.

3. Forecast bias

Bias tells you whether misses lean one direction. It is the only one of the three that keeps its sign.

Bias % = Sum of (Actual(t) - Forecast(t)) / Sum of Actual(t)

A negative bias on revenue means you systematically overforecast. A negative bias on burn means you systematically spend less than planned, which sounds good until you realize the hiring plan slipped. Bias is the number that tells a partner whether your team is optimistic, sandbagging, or calibrated.

Worked example: one quarter, three line items

Here is a Series A company's Q3 reforecast against actuals. Figures are in thousands.

Three things a careful reader pulls from this table:

• Ending cash accuracy is misleading. 98.3% looks excellent, but ending cash is a large balance moved by a small flow. The burn line, at 88.8%, is the honest number. Always grade flows, not balances.
• Headcount and burn moved in opposite directions. The company hired 4 fewer people than planned yet burned $240k more. That means non-headcount spend ran well over plan, likely cloud or contractors. The KPI surfaced a problem that a single "we were 2% under on cash" line would have hidden.
• New ARR accuracy of 82% is the number the board will ask about. Recognized revenue lags bookings, so a bookings miss this quarter becomes a revenue miss next quarter. Forecast accuracy on bookings is the leading indicator.

Which metrics to grade

Grade the lines that drive decisions. For a venture-backed software company the set is short.

If you only have time for two, grade new ARR and net burn one quarter ahead. Those two lines answer the questions a partner actually asks: are you growing the way you said, and are you spending the way you said.

For the mechanics of building the burn and cash lines, our 13-week cash flow forecast template is the structure most of our clients run.

Forecast accuracy benchmarks by stage

These ranges come from our own client base across Seed to Series C plus conversations with investors on what they consider normal. They are not an industry survey. Published surveys like the KeyBanc and Sapphire Ventures SaaS survey report growth and efficiency metrics but do not publish forecast accuracy, which is part of why so few teams track it.

Accuracy is one quarter ahead, against the reforecast locked at the start of the quarter, on a MAPE basis over the trailing four quarters.

Two patterns show up consistently in this data:

• Burn accuracy should be high at every stage. Burn is mostly payroll, rent, and software, all of which are contracted. A Seed company that misses burn by 20% does not have a forecasting problem. It has a controls problem, usually around who can approve spend.
• Bookings accuracy improves with pipeline data, not with better spreadsheets. The jump from Series A to Series B accuracy happens when the CRM has stage-exit dates, weighted values, and a close-date hygiene rule. We cover the specific CRM fields in our guide to forecast accuracy KPIs for Series A companies.

For broader context on how these targets fit alongside growth and efficiency expectations, SaaS Capital's research library publishes annual growth and retention benchmarks by ARR band, and the Bessemer Venture Partners Atlas has the clearest investor-side writing on what good planning looks like at each stage.

How investors interpret forecast accuracy

A partner looking at your forecast accuracy is not grading your spreadsheet. They are estimating three things.

1. How much to discount your next plan

If your last four quarterly bookings forecasts averaged 80% accuracy with negative bias, the partner will haircut the plan you are raising on by roughly 20% before building their return case. That haircut flows directly into valuation. A company with 92% accuracy and neutral bias gets underwritten at or near plan. The difference between those two haircuts at a Series B can be more dilution than the entire fundraising fee stack.

2. Whether the team understands its own business

Consistent misses in one direction mean the team has a blind spot. Overforecasting bookings every quarter means sales leadership is not being held to a weighted pipeline. Underforecasting burn every quarter usually means the hiring plan is aspirational. Either way, the partner concludes that the operating plan is not yet a management tool. The a16z writing on startup metrics makes the same point from the other side: metrics that are not used to run the business are not credible to the people funding it.

3. Whether the company will hit its next milestone before cash runs out

This is the one that decides whether a round happens. The partner takes your stated runway, applies your historical burn accuracy, and asks whether the downside case still reaches the milestone. A company with 18 months of stated runway and 85% burn accuracy has a realistic downside runway closer to 15 months. If the Series B milestone needs 16 months, the partner now sees a bridge round in your future, and prices accordingly.

Forecast accuracy, in other words, is the KPI that converts your other KPIs into an investor's confidence interval. That is why it belongs in the board deck alongside burn multiple and CAC payback, not buried in an FP&A appendix. Our Series A board deck KPI benchmarks show where it fits in the standard slide order.

Seven calculation mistakes that make the KPI lie

Most forecast accuracy numbers we see on day one are wrong. These are the errors we find most often.

1. Grading the latest forecast instead of a locked version. If you reforecast on the 25th and grade on the 31st, you will always look accurate. Lock the version on the first business day of the period and grade that one. Store the locked version somewhere nobody can edit.

2. Dividing by forecast instead of actual. Covered above. It consistently flatters misses when you overforecast, which is the direction most startups miss.

3. Netting positive and negative errors. A +12% month and a -12% month do not equal 0% error. Use absolute values for accuracy and keep the signed number separately as bias.

4. Grading balances instead of flows. Ending cash will almost always show 95%+ accuracy because the opening balance dominates. Grade net burn, the flow that actually changed.

5. Mixing horizons. Comparing a one-month-ahead burn forecast to a four-quarter-ahead bookings forecast and reporting a single "forecast accuracy" number is meaningless. Report each horizon separately.

6. Excluding "one-time" items after the fact. If a $300k legal bill was not in the forecast, it is a forecast miss. Reclassifying it as non-recurring after it hits does not improve your process. Track it as a miss, then decide whether to add a contingency line going forward.

7. Reporting a quarter-to-date accuracy mid-quarter. Accuracy through month two of a quarter is not a forecast accuracy number. It is a progress report. Only grade completed periods.

A simple test: if your forecast accuracy improved sharply the month you changed the calculation, the calculation was the problem, not the forecast.

How to improve forecast accuracy

Accuracy improves when the forecast is driven by operating data that updates automatically, not when someone gets better at guessing. The sequence we run with clients is consistent.

Start from a strategic plan, not a model. Most startup plans we see are a revenue curve with headcount attached. A forecast built on a plan with 1 or 2 clear objectives per horizon and an explicit list of things the company will not do is far easier to hold to, because fewer unplanned initiatives appear mid-quarter and blow up spend. The model should be a calculator that works backward from the target: what pipeline, how many new logos, at what ACV, with what headcount. David Skok's SaaS metrics guide remains the clearest description of how those drivers chain together.

Connect the forecast to source systems. Bookings accuracy jumps when the forecast reads directly from the CRM with enforced close dates and stage-exit timestamps. Burn accuracy jumps when the headcount plan is linked to the HRIS and the spend management tool so that every offer letter and every new vendor updates the forecast the day it happens. This is the engineering work most finance teams skip and most fractional CFO firms do not offer. We built an engineering team specifically to do it.

Route variances in real time. A variance discovered six weeks after month-end cannot be acted on. A variance pushed to the relevant budget owner in Slack the day it appears can be. Teams that get variance alerts within a week of the spend consistently land in the top accuracy band for burn.

Grade publicly, every month. Put the accuracy table in the monthly investor update and the board deck. Teams that know their forecast will be graded forecast more carefully. The Y Combinator Library has several founder-facing pieces on investor updates, and the single most effective change we see is adding a "forecast vs. actual" block that never gets skipped.

Add contingency explicitly, not implicitly. Padding every line by 10% destroys the bias signal. Instead, forecast each line honestly and add a visible contingency line sized to historical error. Investors respect an explicit contingency. They distrust a forecast that always comes in exactly on plan.

For a full worked model with these drivers pre-built, our free SaaS financial model template includes a forecast-vs-actual tab that calculates all three versions of the KPI automatically.

Reporting forecast accuracy to the board

One slide is enough. It should contain:

• A trailing four-quarter table with accuracy and bias for new ARR, net burn, and headcount, one quarter ahead.
• A one-line root cause for any line below the stage benchmark.
• The specific process change made in response, with an owner.

Do not report a blended "overall forecast accuracy" number. Boards see through it, and it hides the one line that needs attention. If you are building the full deck, the investor-ready board deck template shows the slide in context, and the Series B data room checklist covers how to present historical forecast-vs-actual in diligence.

The OnlyCFO newsletter has written about how finance leaders lose credibility with boards, and unreported misses sit near the top of that list. A miss you report with a root cause costs you one hard conversation. A miss the board discovers costs you the next plan's credibility.

When the KPI says you need help

If your Series A company's bookings accuracy is below 75% or burn accuracy is below 90% for two consecutive quarters, the problem is almost never the person building the spreadsheet. It is the lack of a strategic plan to forecast against, the lack of system connections that keep the forecast current, or the lack of a routing process that gets variances to the people who can act on them.

That is the work a finance architect does, and it is the work we built CFO Advisors around. We start with the strategic plan, build the model as a calculator investors can underwrite, connect the forecast to your CRM, HRIS, and spend tools through our own data pipeline, and push variances to budget owners in Slack at whatever cadence the business needs. Our clients have raised over $1.2 billion, and forecast accuracy is one of the first numbers their investors ask us about. If your next raise depends on a plan a partner will believe, talk to a fractional CFO who measures the forecast as carefully as the forecast measures the business. You can also see how we evaluate accuracy when assessing a finance function in our 7 criteria for hitting 95% forecast accuracy.

FAQ

What is a good forecast accuracy KPI for a startup?

It depends on the metric, the horizon, and the stage. For a Series A company forecasting one quarter ahead, 75% to 85% on new ARR and 90% to 95% on net burn is normal. Series B companies should be at 85% to 92% on bookings and 93% to 97% on burn. Burn accuracy should be high at every stage because most spend is contracted; a large burn miss indicates a controls problem rather than a forecasting problem.

How do you calculate forecast accuracy percentage?

Take the absolute difference between actual and forecast, divide by actual, and subtract from 1. For $900k actual against a $1M forecast, the error is $100k divided by $900k, or 11.1%, so accuracy is 88.9%. Across multiple periods, average the absolute errors first (MAPE) and then subtract from 1. Track bias, the signed error, separately so you can see whether misses lean one direction.

Should forecast accuracy use the original budget or the latest reforecast?

Both, reported separately. Accuracy against the annual plan tells the board how well the company understood the year at the outset. Accuracy against the quarterly reforecast tells them how well the team is reading the business in real time. Investors care most about the reforecast number because it predicts how reliable your next plan will be. Never grade against a reforecast made inside the period being graded.

Why is forecast bias important if I already track accuracy?

Accuracy tells you the size of the miss. Bias tells you the direction. A team that is 85% accurate with neutral bias is calibrated and just noisy. A team that is 85% accurate with consistent negative revenue bias is systematically optimistic, and an investor will discount every forward plan by that bias before underwriting the round. Bias is also the earlier warning signal; it often shifts a quarter or two before accuracy degrades.

How does forecast accuracy affect startup valuation?

Indirectly but significantly. Investors underwrite rounds on a forward plan. If historical accuracy suggests the plan is 20% optimistic, they haircut the plan, which lowers the forward revenue they are pricing against. At a Series B, the gap between being underwritten at plan versus 80% of plan can be worth more dilution than the entire cost of the raise. High accuracy with explained misses lets a company raise closer to its own plan.

How often should a startup measure forecast accuracy?

Monthly for net burn and headcount, quarterly for bookings and revenue, and annually against the original plan. Report the trailing four quarters on one board slide each quarter, with root causes for any line below the stage benchmark. Only grade completed periods; quarter-to-date numbers are progress reports, not accuracy measurements.

Sources

1. KeyBanc Capital Markets and Sapphire Ventures - 2024 SaaS Survey
2. SaaS Capital - Research library of private SaaS growth and retention benchmarks
3. Bessemer Venture Partners - Atlas, investor guidance on planning and metrics by stage
4. a16z - Writing on startup metrics and operating discipline
5. David Skok, For Entrepreneurs - SaaS Metrics 2.0 and driver-based forecasting
6. Y Combinator Library - Founder guidance on investor updates
7. OnlyCFO - Newsletter on finance leadership and board credibility