What Is a Consensus Estimate? Meaning, Method and Real Accuracy

What is a consensus estimate?

Consensus means agreed opinion. In markets it refers to a representative forecast pooled from the analysts covering a company — most commonly for earnings per share and revenue. Reported EPS above the consensus is a beat; below it is a miss. The benchmark is neither last year's result nor the company's own target: it approximates what the market currently expects.

A consensus estimate is not a correct answer — it is the level the market currently expects, expressed as a number.

How the estimate is built

Each analyst models revenue, costs, margins and tax to arrive at an EPS estimate. Data vendors pool those submissions into a representative figure. Vendors differ in method — some use the mean, others a median or their own filters — so the same company can carry slightly different consensus figures depending on the source. The number of contributors matters too. Megacaps are covered by dozens of analysts; small caps sometimes by two or three, where one revision can move the whole figure.

How accurate is the consensus?

Across US earnings reports between 1 September 2025 and 31 August 2026, the relative gap between actual and consensus EPS had a median of 9.3%. 31.7% landed within 5%, 51.8% within 10%, and 23.0% were off by more than 25%. The consensus is a useful benchmark; it is not a precise prediction of any individual company's EPS.

9.3%

Median relative gap between actual and consensus EPS

31.7%

Reports landing within 5% of the estimate

23.0%

Reports off by more than 25%

Period: 1 Sep 2025 – 31 Aug 2026 · Sample: 4,926 US earnings reports with both consensus and actual EPS available · Source: AFTERNS earnings database

Why is the mean error 92.1% when the median is 9.3%?

The same dataset produces a mean error of 92.1% against a median of 9.3%. Companies with EPS close to zero are the reason. An estimate of $0.02 against an actual of $0.10 is an eight-cent difference but a 400% relative error. A handful of those values lifts the mean dramatically. Medians and the distribution of errors are the more stable way to read forecast accuracy. Every figure in this article is a median.

Mean error 92.1%, median error 9.3% — from the same dataset. Statistics about consensus accuracy depend entirely on how they are calculated.

Are estimates more accurate for larger companies?

In this sample, larger market-cap bands generally showed a smaller median error: 12.3% below $10bn against 6.3% in the $50–200bn band. Above $200bn the error rose slightly again to 7.8%. Rather than concluding that larger always means more accurate, the safer reading is that small caps showed relatively larger errors. One possible explanation is differing analyst coverage and information flow — but this data alone cannot establish the cause.

Market cap is as of today, not as of the release date. Error is the absolute difference between actual and consensus EPS (median).

How common is a beat?

In the same sample, 77.7% of reports came in above consensus. For context, FactSet puts the S&P 500's positive EPS surprise rate at a 78% five-year average and a 76% ten-year average, with 86% in Q2 2026. The populations differ — AFTERNS covers the broad US market — but the direction is the same: beating consensus is not unusual. That is why the bare fact of a beat carries little information.

Why the consensus keeps moving

A consensus estimate is not fixed. Analysts revise whenever new information arrives:

So the estimate a month before a release and the estimate the day before can differ. Continuous revision is frequently cited as one reason beat rates run high, though guidance policy and sector characteristics also play a part.

1. 1Company guidance
2. 2Competitor results
3. 3Industry indicators
4. 4Commodity prices
5. 5Exchange rates
6. 6Interest rates
7. 7Management commentary

What is a whisper number?

Market participants sometimes expect a figure above the published consensus — informally called the whisper number. Because it is unofficial, it cannot be measured as a single figure. It is inferred from pre-earnings price action and the move implied by the options market. It explains the familiar case where a company beats and the stock still falls: a stock that has already rallied was probably expected to clear more than the published number.

Never mix GAAP and adjusted EPS

US consensus figures are frequently provided on an adjusted (non-GAAP) basis, excluding one-off items. The EPS a company files with the SEC is GAAP. The two can differ substantially for the same quarter. Always compare like with like. Measuring GAAP EPS against an adjusted-EPS consensus can flip a beat into a miss.

Four things to check alongside the consensus

Beat or miss alone explains price direction barely better than a coin flip — 45.4% of beats closed lower the next session. A more useful reading:

1. 1The size of the gap, not just its signa +1% beat and a +20% beat are both beats, and they do not mean the same thing.
2. 2EPS and revenue togetheran EPS beat with a revenue miss may reflect cost control rather than growth.
3. 3Next quarter's guidanceprices often respond more to the change in forward expectations than to the quarter just reported.
4. 4How the consensus itself moved recentlyif estimates were climbing into the release, simply clearing them may not satisfy the market.

Data and method

Every figure here was computed directly from the AFTERNS US earnings database.

1. 1Periodearnings released between 1 September 2025 and 31 August 2026.
2. 2Sample4,926 reports with both a consensus and an actual EPS. Of 5,005 total releases in the period, 79 (1.6%) had no consensus and were excluded.
3. 3Duplicateseach quarterly release counts as one report.
4. 4EPS basisconsensus and actual are compared as adjusted (non-GAAP) figures on the same basis, never mixed with GAAP EPS from SEC filings.
5. 5Error definition|actual EPS − consensus EPS| ÷ |consensus EPS| × 100. Because the denominator can approach zero, medians are used as the representative value.
6. 6Market cap is as of today rather than the release date, so the banded comparison is indicative.
7. 7Limitationanalyst coverage counts are not collected, so coverage depth was not verified directly. The market-cap relationship is correlational and should not be read as causal.

Frequently asked questions

Related guides

• What Is an Earnings Surprise? Definition, Calculation and Price Data
• What Is an Earnings Shock? Definition, Threshold and Crash Data
• US Earnings Calendar Subscription · Auto Alerts for Report Dates
• What Time Do US Companies Report Earnings? BMO vs AMC Explained
• Earnings Analysis: How to Read Post-Earnings Price Moves