Analyst price target herding: who moves first, and who follows

Analyst price target herding is easy to assert and hard to measure. So we measured it. Across 361,708 directional price target revisions in AnaChart’s records, more than half of them are not standalone calls at all. They sit inside a pile.

The question this study answers is not whether the pile exists. It is who starts it, how fast the rest arrive, and whether being early is a stable trait of an analyst or just a run of luck.

What a herd looks like in the data

Start with every recorded price target revision where we know both the old number and the new one, and where the two differ. That gives 361,708 revisions. 60.4% are raises and 39.6% are cuts.

Now sort them by stock and by direction, and chain together any revisions that land within ten days of each other. A chain of three or more is what we call a cluster. On that definition, 199,284 revisions, 55.1% of the total, belong to a cluster. There are 36,434 of them. The median cluster holds four revisions and runs its course in five days. The largest holds 64.

That largest one is worth sitting with. On 9 March 2023, Jason Helfstein lifted his target on META from $220 to $235. Over the following 50 days, 29 different analysts raised their Meta targets 64 times between them. The last revision in the run set $300, a level 27.7% above where the first mover had put it. Nobody in that sequence was reacting to Helfstein. They were all reacting to the same company. But the shape of the record is a single call followed by two dozen people arriving behind it, and that shape is what a subscriber sees.

The clock: how fast the followers arrive

Inside those 36,434 clusters there are 115,372 follow-on revisions. Timing them against the first mover gives the tightest result in the study.

The median follower is seven days behind. 16.0% publish inside 24 hours, which is fast enough that they are almost certainly responding to the same event rather than to each other. But the bulk of the distribution sits in the second week, long after the initial news has been priced.

One detail cuts against the obvious reading. Followers do not move timidly. The median first mover changes their target by 8.7%. The median follower changes theirs by 9.3%. Arriving late does not come with a smaller step.

Who moves first

Restrict the field to the 394 analysts with at least 150 clustered revisions, so the rate is measured on real volume. The median analyst in that group was the first mover 41.1% of the time. Here is the top of the distribution.

Lead rate is the share of an analyst’s clustered revisions where they were the earliest in the cluster. Ties on the same day all count as leaders. Minimum 150 clustered revisions.

John Walsh sits at 74.4%, close to double the median. Deane Dray is the most striking entry on volume: 438 clustered revisions is more than twice what most of this table carries, and he was still first 61.4% of the time.

Who arrives late

The other end of the same distribution is not a hall of shame. Following is a legitimate way to run coverage, and several names here run large, high-quality franchises. It is simply a different job.

Same population. Median days behind is measured from the first revision in each cluster.

Ramsey El-Assal led 12.8% of the time, roughly a third of the median rate, and typically published four days after the cluster opened. John Freeman has the longest median lag in the group at ten days, which puts him past the point where most of the cluster has already formed.

Is a lead rate real, or is it noise?

A ranking is only useful if it holds up out of sample. So we split the record in half at the end of 2021, kept the 153 analysts with at least 60 clustered revisions on each side, and compared them.

So being early is a durable trait, not a coin flip. It is also not a small effect: a 13 point spread in lead rate, sustained across a four year gap, is the difference between an analyst whose revision is usually news and one whose revision is usually confirmation.

Herding did not explode. Coverage did.

Here is the finding we nearly published the wrong way round. Plot the clustering rate by year across all stocks and it climbs from 18.4% in 2016 to 69.3% in 2025. That looks like an era of collapsing independence, and it would have made a better headline than the truth.

Clustering is mechanically easier when more analysts cover a name, and the number of well-covered names in the record grew enormously over the decade. Hold that constant and the picture flattens out.

Each row compares stocks with a similar number of directional revisions in the year. The 26 to 50 band, the most stable comparison available across both endpoints, moved 1.7 points in ten years.

One row does move. Thinly covered stocks, the ones carrying ten or fewer revisions in a year, went from 8.6% to 22.5%. The herd has not got denser where it was already dense. It has reached further down the market cap ladder.

Where the herd is thickest

Among stocks with at least 300 directional revisions, the most clustered names are large technology franchises with heavy, synchronised coverage.

Share of that stock’s directional price target revisions that sit inside a cluster of three or more. Minimum 300 revisions.

On NVDA, 93.4% of price target revisions arrive as part of a pile. A standalone, independent revision on Nvidia is the exception, not the rule. The practical consequence is that a single target change on a name like this carries almost no information on its own, because the base rate says more are coming.

What this changes about reading a price target

Three things follow from the numbers above, and none of them require you to think badly of anyone.

A revision on a crowded name is usually not new information. On the most-covered stocks, nine in ten revisions are part of a cluster. Treating each headline as an independent signal double counts the same underlying event over and over.

Position in the cluster is a measurable property of the analyst. It persists across years, it survives an out-of-sample split, and it varies by a factor of nearly six between the extremes of the table above. That is a fact about the person, available before you read their note.

The first fortnight is the whole game. 92.9% of a cluster lands within 21 days. If you want to know whether a revision is the start of something, you are looking at a two week window, not a quarter.

How we measured this, and what it cannot tell you

The population is every recorded price target revision in AnaChart’s dataset where the prior target, the new target, the analyst and the date are all present, and the two targets differ. That is 361,708 revisions drawn from records reaching back to 2004, with dense coverage from 2013 onward. Analyst identities are consolidated to a canonical name, so a person who changes firm keeps one record rather than splitting into two.

A cluster is built per stock and per direction. Revisions are sorted by date, and a new cluster begins whenever the gap to the previous revision exceeds ten calendar days. Clusters of three or more are kept. Two consequences of that rule are worth stating plainly. First, because the chain is built link by link, a cluster can run longer than ten days in total, and the median one runs five. Second, the ten day window is calendar days, not trading days, so a cluster spanning a long weekend gets slightly less room than one that does not.

What this study cannot separate. The dataset does not carry earnings dates or news timestamps. When 29 analysts raise their Meta targets inside 50 days, this method cannot tell you how much of that is analysts responding to each other and how much is 29 people independently reading the same quarterly report. Both produce identical patterns in the record. Everything above is a description of timing and sequence. It is not a claim about causation, and it is not a judgement of anyone’s work.

One further caveat on the leaderboards. Lead rate depends on the company an analyst keeps. An analyst covering a sector where the whole street publishes on the morning of an earnings release will show a different rate from one covering names with staggered reporting, regardless of how independently either of them thinks. The split-half test above shows the measure is stable, not that it is free of that influence.

Related reading: price targets follow the stock rather than lead it, and why the analyst covering your stock probably will not be there in ten years.

Frequently asked questions

What is analyst price target herding?

It is the pattern where a price target revision on a stock is quickly followed by more revisions in the same direction from other analysts covering that stock. In AnaChart’s records, 55.1% of the 361,708 directional price target revisions sit inside a cluster of at least three same-direction revisions on the same ticker, where each one lands within ten days of the one before it. Only 26.2% of revisions stand completely alone.

How fast do other analysts follow a price target change?

The median follower publishes seven days after the first mover in the cluster. 16.0% arrive inside a single day, 41.1% within five days, and 72.3% within ten. By day 21, 92.9% of the cluster has landed. The window that matters is the first two weeks.

Which analysts move price targets first?

Among the 394 analysts with at least 150 clustered revisions, John Walsh leads the field: he was the first mover in 74.4% of his 164 clustered revisions. James Fish follows at 72.8% of 246, then Mike Matson at 68.6% and Gary Tenner at 68.5%. The median analyst in this group led 41.1% of the time, so these are roughly 1.7 times the norm.

Is a high lead rate repeatable or is it just luck?

It repeats. Splitting the record at the end of 2021 and keeping the 153 analysts with at least 60 clustered revisions in each half, an analyst’s lead rate before 2022 correlates 0.583 with their lead rate after it. Analysts in the top quartile early on went on to lead 47.1% of the time, against 34.3% for the bottom quartile.

Has price target herding got worse over time?

The headline number says yes and the headline number is misleading. Across all stocks the clustering rate rose from 18.4% in 2016 to 69.3% in 2025. But hold coverage density constant and it flattens: among stocks carrying 26 to 50 revisions in a year, clustering was 73.8% in 2016 and 72.1% in 2025. What grew is the number of stocks with enough analysts on them to form a herd. The one real change is at the thin end, where stocks with ten or fewer revisions went from 8.6% to 22.5%.

Which stocks have the most clustered price target revisions?

Large, heavily covered technology names. Nvidia tops the list of stocks with at least 300 directional revisions, with 93.4% of them sitting inside a cluster, followed by CrowdStrike at 92.5%, Salesforce at 91.1% and Zscaler at 90.3%. Meta Platforms sits at 90.1% across 1,245 revisions, including one run in March 2023 where 29 analysts raised their targets 64 times in 50 days.