What an analyst consensus is — why 40 forecasts set the share price
A consensus estimate is the average of the forecasts that securities analysts publish for a company on revenue, operating profit or earnings per share. It is not the figure the company issues itself — that is guidance — and it is not a single official number either, since each data provider polls a different set of analysts and the same company can carry two or three different consensus values. Beating it by roughly 10 percent is called an earnings surprise and missing by roughly 10 percent an earnings shock, but the 10 percent line is convention, not regulation
The three lines
- Definition — the average of analysts' forecasts. Distinct from guidance, which the company issues itself
- Divergence — providers poll different analysts, so one company can carry $91.85bn and $92.2bn at once
- Consequence — prices react to the gap between results and consensus, not to results alone
Key questions
- What does consensus mean in stock market terms?
- It is the **average of the forecasts individual securities analysts publish** for a company's results — typically **revenue, operating profit, net profit and earnings per share**, for a quarter or a year. In Korea, providers such as FnGuide aggregate domestic brokerage research; internationally, Visible Alpha, LSEG and FactSet do the same job. The distinction that matters most is this: **the consensus is not the company's number.** When a company says what it expects next quarter, that is **guidance**. Consensus is an estimate assembled **outside** the company, from guidance plus industry data plus each analyst's own work. The two are usually different, and the gap between them carries information of its own.
- How big does a beat have to be to count as an earnings surprise?
- **Around 10 percent, by convention.** Results more than roughly **10 percent above consensus** are commonly called an **earnings surprise**, and more than roughly **10 percent below** an **earnings shock**. But **no exchange or regulator sets that line.** In practice one outlet can call the same result a surprise while another calls it in line, and the disagreement is usually traceable to three choices: **which line item** is being measured (revenue, operating profit or EPS), **whose consensus** is being used, and **what threshold** is being applied. Coverage of SK Hynix's April 2026 results split exactly this way, with 'missed consensus' and 'earnings surprise' both appearing for the same quarter.
- Why does a stock fall after strong results?
- Because **prices respond to the difference between results and expectations, not to results**. The market knows the consensus before the release and has already priced that outcome. So the only genuinely new information on the day is **how far the actual figures diverged from it** — and a record result that merely matches consensus contains no new information at all. A second factor often matters more: releases usually come with **guidance for the next quarter**, and if that guidance falls below consensus, the shares can fall no matter how good the quarter just reported was. The market is pricing the next quarter, not the last one. Margin figures work the same way: a gross margin that slips even a point gets read as weakening pricing power.
A consensus estimate is the average of analysts' forecasts.
It is not the company's number, and there is no single official version of it.
Yet on results day, what sets the share price is not the results — it is the gap against this number.
1. Who produces what
| Term | Produced by | Nature |
|---|---|---|
| Guidance | The company | Management's own forecast. Disclosure |
| Consensus | Securities analysts | An average of outside estimates |
| Whisper number | Informal market expectation | No aggregator |
Aggregation is done by data providers. FnGuide compiles Korean brokerage research into per-stock consensus figures. Visible Alpha, LSEG and FactSet do the equivalent internationally.
Four line items are usually covered.
| Item | Meaning |
|---|---|
| Revenue | Total sales |
| Operating profit | Profit from the core business |
| Net profit | After interest and tax |
| EPS | Net profit divided by shares outstanding |
US coverage tends to lead on EPS, Korean coverage on operating profit. The same result can beat on one and miss on another.
2. There is no single consensus
This is the most common misunderstanding. Saying "the consensus is X" is incomplete without naming whose.
A live example: the reference points for Nvidia's revenue ahead of its August 26, 2026 release.
| Basis | Revenue |
|---|---|
| Company guidance | $91bn ±2% |
| Consensus (Visible Alpha) | $92.2bn |
| Consensus (40-analyst poll) | $91.85bn |
Three different numbers. If the actual print lands at $92bn, the same result is a miss on one basis and a beat on another.
The divergence has mundane causes.
- Different analyst panels — one provider polls 40, another 60
- Different refresh timing — whether last-minute revisions are captured
- Different outlier handling — some providers trim extreme estimates
Which is why "beat consensus" and "missed consensus" headlines sometimes appear for the same company on the same day.
3. The 10 percent convention
| Outcome | Term | Usual threshold |
|---|---|---|
| Well above consensus | Earnings surprise | about +10% or more |
| Close to it | In line | between |
| Well below | Earnings shock | about -10% or less |
This is convention, not regulation. No exchange or supervisor set it, and institutions apply it differently.
Coverage of SK Hynix's April 2026 results split along exactly these lines, with "missed consensus" and "earnings surprise" both published for the same quarter. The three variables above — line item, provider, threshold — were all in play at once.
When you read "earnings surprise," it is worth checking all three.
4. Why strong results can sink a stock
Laying the day out in order explains it.
| Point | What the market knows | What is in the price |
|---|---|---|
| Before | Consensus | Results equal to consensus |
| The release | Actual results | Results minus consensus = new information |
| Just after | Next-quarter guidance | The revision to future expectations |
The second row is the mechanism. Results identical to consensus contain zero new information. A record quarter that was expected to be a record quarter is not a reason for the price to move.
The third row flips more prices than the second. Guidance below consensus can sink a stock regardless of how good the reported quarter was, because the market is buying the next quarter, not the last one.
Add a quality measure — gross margin — and a fourth factor, and results-day pricing runs on at least four comparisons.
| Comparison | What it settles |
|---|---|
| ① Results vs consensus | The quarter just ended |
| ② Guidance vs consensus | The quarter ahead |
| ③ Margin vs prior quarter | The quality of the business |
| ④ Share price into the print | Whether expectations were already cut |
Row ④ is easiest to see with a live case. Nvidia entered its August 26 release on a seven-session losing streak. When expectations have already been marked down, the same number produces a different reaction. This brief covers that setup in "Nvidia reports on August 26 — the bar is $92 billion."
5. Reading consensus carefully
- It is an average, not a prediction. There is no reason the mean of many wrong estimates should be right. Its significance is that it is what the market has already priced.
- Analysts revise close to the release. Consensus a week before a print may differ from consensus on the day.
- Thin coverage means unstable consensus. An average of three analysts is not comparable to an average of forty.
- Some items have no consensus at all. New business segments and one-off gains often go unmodelled, so whatever they contain arrives as entirely new information.
6. What is not settled
- Origin of the 10 percent line — widely used, not formally set, and its provenance was not established.
- Provider methodology — panel selection, refresh cadence and outlier handling were not checked against published documentation.
- Whisper numbers — not aggregated anywhere, so not verifiable.
- Frequency of post-results declines — varies by company; no generalised figure was found.
- Related briefs — "What program trading is" covers how large post-results flows are executed; "What GPU depreciation is" covers how accounting choices move margin figures.
Sources
- FnGuide Company Guide — Earnings surprise and consensus screening
- Stockplus — What earnings surprise and earnings shock mean
- Sisa Prime — Economic terms explained: earnings shock and earnings surprise
- MTN — SK Hynix: consensus miss or earnings surprise? Why the readings diverged
- Seeking Alpha — Nvidia Earnings Preview: Q2 2027 (consensus example)
- Regards of Wallstreet — When Does Nvidia Report Earnings? The Bar Is $91 Billion (guidance example)