Analyst price targets are a lagging indicator, and the lag is measurable. Across 426,782 consecutive revisions in AnaChart’s records, the correlation between how an analyst moved a target and how the stock had already moved since their previous call on that name is 0.705.
That number matters to anyone building a signal on revision data, because it means a large share of what looks like analyst opinion is trailing return wearing a different label.
82.6% of price target revisions move in the same direction the stock had already moved. Median gap between one call and the next on the same stock: 90 days. After a stock rose more than 30%, 94.5% of the next revisions were raises.
The pattern, measured
Take every case where an analyst set a target on a stock and later set another on that same stock. That yields 426,782 pairs across 5,165 analysts and 5,695 tickers. For each pair, compare the direction of the revision against the stock’s price change between the two call dates.
In the 357,537 pairs where both the target and the price moved, 82.6% went the same way. The relationship is not linear, and the shape is the useful part.
Read the middle bar first. With the stock flat, the follow rate falls to 62.0%. With no price signal to anchor on, revisions scatter. Read the ends and the picture inverts: after a fall of more than 30%, 93.7% of revisions were cuts, and after a rise of more than 30%, 94.5% were raises.
They follow, and they under-follow
Following the price would at least preserve the implied upside. It does not.
Across the 122,266 revisions that came after a gain of more than 10%, the median revision raised the target by 16.5% while the stock had already gained 21.5%. The headline number goes up. The distance between target and price gets smaller.
A raised price target frequently accompanies a contraction in implied upside, not an expansion. Any model reading revision direction as bullish without conditioning on trailing return is partly re-encoding momentum it already has.
Tenure does not condition it
The obvious control is experience. It does almost nothing.
| Analyst tenure at time of call | Revisions | Follows prior move | Median implied upside |
|---|---|---|---|
| Under 4 years | 73,414 | 81.8% | 15.5% |
| 4 to 12 years | 264,556 | 82.8% | 15.6% |
| 12 years or more | 63,555 | 83.0% | 15.9% |
Tenure measured as years between the analyst’s first recorded target anywhere in the dataset and the call in question. Implied upside computed across all 625,901 targets carrying a same-day reference price.
The spread across two decades of experience is 1.2 percentage points on follow rate and 0.4 points on conviction, and both run mildly the wrong way. Whatever drives this behaviour, seniority is not it.
What does condition it: coverage sector
The averages hide real dispersion, and the dispersion is usable.
| Analyst | Revisions | Follows prior move |
|---|---|---|
| Mathieu Robilliard | 155 | 56.8% |
| John Gerdes | 156 | 64.1% |
| Steven Seedhouse | 176 | 65.9% |
| Mani Foroohar | 165 | 66.1% |
| Steve Barger | 203 | 95.6% |
| Rupesh Parikh | 237 | 95.4% |
| Vincent Andrews | 165 | 94.5% |
Analysts with at least 150 consecutive revisions: the four least and three most tied to the prior price move. Lower means more independent of the tape.
The bottom of that table is biotech and energy coverage, where a trial readout or a reserve revision resets a model regardless of what the share price did. The top is consumer and industrial coverage, where a multiple on a forward earnings base tracks the price by construction. Neither group is doing anything improper. They carry different information content, and a model treating them identically discards that.
Even the most heavily covered names show it. Tesla at 77.9% and Nvidia at 78.0% are the least tape-following megacaps; Meta sits at 87.3%. Across every ticker with 500 or more revisions the range runs 73.9% to 91.0%. There is no meaningful population of names where revisions lead price.
Why most datasets cannot show you this
This is what determines whether you can run the analysis yourself, and it comes down to one field.
To classify a revision as following or leading you need three things joined correctly: the target value, the exact date it was issued, and the security’s price on that date. Most price target feeds ship the first two. Without the third, a revision from $180 to $210 is just a number changing. You cannot tell whether the analyst moved ahead of the market or after it, so momentum and opinion stay fused and neither can be isolated.
The same field is what makes an accuracy record computable at all. Whether a target was reached, how long it took and by what margin are all measured from the price at issue. A dataset without it can tell you what analysts said, never whether it was worth anything.
What the AnaChart dataset carries
Every one of the 833,361 records carries the analyst, the broker, the ticker, rating before and after, target before and after, the issue date, and the reference close on that date, resolved to a canonical analyst identity so a person’s history survives a change of firm. Coverage runs from 2004 to 2026, with deep continuous coverage from 2013, across 7,191 analysts and 9,686 tickers, with no survivorship filtering.
How to use this in a model
Orthogonalise before you use it. Regress revision magnitude on trailing return over the interval since that analyst’s previous call on the name. The residual is the part not explained by price, and it is a far cleaner proxy for a change of view than the raw revision.
Condition on the analyst, not the firm. The 56.8% to 95.6% spread above is stable at 150-plus observations and usable as a weight. An analyst whose revisions historically lead the tape carries different information from one whose revisions track it.
Treat a flat tape as the high-information regime. The 62.0% bucket is where revisions carry the most independent content, and where the fewest people look.
Method and limits
Construction. For each analyst and ticker, targets were ordered by date and consecutive pairs formed. Each pair gives the percentage change in the target and in the reference close between the two dates. A revision follows if both moved the same direction.
Filters. Pairs more than 400 days apart were dropped. Changes beyond plus or minus 200% were dropped as artifacts. The 82.6% figure uses the 357,537 pairs where both values moved by more than 0.1%; the 0.705 correlation uses all 426,782 pairs.
What this does not measure. It says nothing about whether targets were reached, how fast, or how any individual analyst performed. Those are separate fields computed from full price history. This measures revision direction against the price move that preceded it. Sector labels above describe the named analysts’ coverage, not a formal classification.
Frequently asked questions
Do analyst price targets lead or lag the stock price?
In aggregate they lag. Across 426,782 consecutive price target revisions covering 5,165 analysts and 5,695 tickers, the correlation between a revision and the stock’s prior move is 0.705, and 82.6% of revisions move in the direction the stock had already travelled since that analyst’s previous call on the same name. The effect strengthens with the size of the move: after a stock rose more than 30%, 94.5% of the next revisions were raises.
Why can’t I measure this with a standard price target feed?
Because the calculation needs the reference price at the moment each target was issued, and most feeds do not carry it. You need three fields joined correctly: the target, the timestamp, and the security’s price on that date. Without the third you can observe that a target changed but not whether it moved with or against the market, so momentum and analyst opinion stay entangled. AnaChart carries the issue-date close on every one of its 833,361 records.
How much of a price target revision is just momentum?
Enough to matter for a factor model. After a stock had risen more than 10%, the median revision raised the target by 16.5% while the stock had already gained 21.5%. The target rises but the implied upside contracts, so a raise can coincide with a weaker forward-looking signal than the target it replaced. Treating revision direction as an independent alpha source double counts trailing return.
Does analyst seniority change the pattern?
Barely, and not in the direction most people expect. Splitting every revision by how long the analyst had been publishing, the share that follow the prior move is 81.8% for analysts under four years, 82.8% for four to twelve years and 83.0% for twelve years or more. Conviction is flat too: median implied upside is 15.5%, 15.6% and 15.9% across the same buckets on 625,901 targets. Tenure is not a useful conditioning variable here.
Which analysts are least tied to the prior price move?
They concentrate in sectors where value is reset by events rather than by price. Among analysts with 150 or more revisions, Mathieu Robilliard follows the prior move 56.8% of the time, John Gerdes 64.1%, Steven Seedhouse 65.9% and Mani Foroohar 66.1%, all biotech or energy coverage. At the other end, consumer and industrial analysts such as Steve Barger at 95.6% and Rupesh Parikh at 95.4% track the tape almost mechanically. That spread is itself a usable conditioning variable.
What delivery options does the dataset support?
Snowflake and BigQuery shares, or flat file delivery for teams that prefer their own pipeline. Coverage runs back to 2004 with deep continuous coverage from 2013, across 7,191 analysts and 9,686 tickers, with no survivorship filtering. A free Apple sample with the full field set is available so you can reproduce the calculations in this article first.
The free Apple sample carries the full field set, including the issue-date reference price these calculations depend on. Request the free sample, or see the dataset specification and delivery options.
Data as of April 2026. Published 27 July 2026.