# Signal autopsy: Coherent fell 48%. The score stayed bullish the whole way down, and the price came back > Coherent lost nearly half its value between June and late July. The Quantral score never turned with it, holding in the 80s and low 90s on a handful of accounts with real track records. Then the price came back 46% to meet it. By Maya Koeva · 2026-08-06  We do not give buy tips, and this is not one. What we do is read the conversation around thousands of companies and score how strong and credible it is, in real time. Coherent, ticker COHR, is the optical components maker that sells the lasers and transceivers inside AI data centers. Between June and late July its stock lost 48%. Our [signal score](/learn/what-is-a-stock-signal) held in the 80s and low 90s through most of that fall, dipping only briefly at the very bottom, and it turned out to be reading the situation better than the price was. ## The score against the price Coherent closed at $426.89 on June 2 and at $222.05 on July 29. Here is what the 7-day score was saying on the way down, taken at each day's close. | Date | Score | Close | |---|---:|---:| | Jul 10 | 83 | $324.50 | | Jul 17 | 84 | $277.60 | | Jul 20 | 90 | $285.40 | | Jul 23 | 92 | $313.22 | | Jul 27 | 92 | $271.31 | | Jul 28 | 63 | $243.33 | | Jul 29 | 78 | $222.05 | | Jul 30 | 87 | $249.06 | | Aug 3 | 89 | $288.14 | | Aug 4 | 86 | $323.73 | The stock fell 31% between July 10 and July 29. Across those same weeks the score moved from 83 to 92 and back to the high 80s. It did not follow the price down, because it was never measuring the price. It was measuring whether the people with track records had changed their minds, and they had not. *[Chart: Coherent (COHR) daily closes, July 10 to August 4, 2026. The low is July 29 at $222.05.]* ## What held it up Between July 6 and August 5, 36 people across the accounts Quantral tracks made a directional call on Coherent. **22 of them, 61%, were trusted voices**, accounts with a measured [track record](/learn/how-a-track-record-is-graded) of being right, and the average [credibility](/learn/how-to-tell-if-a-finance-influencer-is-worth-following) of the group was 50%. Their calls split 54 bullish to 9 bearish. On most of those days the score was built from five to eight voices. That is the part worth sitting with: a score in the high 80s, on single-digit voice counts, while the stock was in free fall. A system that counted mentions would have had nothing to say about Coherent in July. Weighting by record instead of by volume is what let a handful of people carry a high reading. ## The calls at the bottom The dates matter here, so take them slowly. On July 28, with the stock at $243 and already down 43%, an account with an 81% credibility score argued the selloff had taken Coherent below the price Nvidia paid for its own stake, and read that as the buying opportunity. On July 30, the day after the $222 low, an 80% account laid out the bull case in specifics: datacenter and communications segment growth, plus the Nvidia relationship. On August 1, a 71% account made the narrowest version of the argument, that Coherent sits on a supply chokepoint for indium phosphide lasers. The score was not blind to the other side. On July 28 a 54% account posted the bear case, pointing at the decline and an earlier warning, and you can see the effect: the score dropped from 92 to 63 that day, before recovering to 78 and then 87 as the bullish voices came back in. Even at its low point it stayed on the bullish side of neutral. That is a score responding to evidence rather than sitting on a number. ## Volume was the last to know Mentions, as opposed to conviction, arrived at the very end. Coherent drew zero graded mentions on August 3, the day it rose 9.6%. On August 4 it drew 32, the loudest day in two months. *[Chart: Coherent (COHR) mentions on Quantral, July 6 to August 5 2026: quiet the whole way down, then a spike on August 4, after the move.]* So the two readings diverged completely. Volume put Coherent in front of you on August 4, at $323, after a 46% run. The score had it in the high 80s from July 10, while it was still falling. A [mention leaderboard](/blog/most-mentioned-stocks-week) is a record of what already happened. That gap is the whole reason we weigh a voice by its record instead of counting heads. ## Where it was early The score was right about direction and early about timing, and those are different things. Acting on July 10, when the score first hit 83, meant buying at $324 and watching it fall to $222 before it recovered to roughly where you started. The comfortable entries came later: the 90 on July 20 at $285 was worth 13% by August 4, and the 87 on July 30 at $249 was worth 30%. The score reads the conversation, not the calendar. A high reading says the credible people are leaning one way with reasons you can name. It does not say the market agrees this week. On Coherent it took about three weeks for the price to come around, and there was a 31% drawdown in the middle of being right. ## Why we score this way One stock does not prove a method, and we publish the reads that [go the other way](/blog/signal-autopsy-maxlinear) alongside the ones like this. But Coherent is a clean picture of what the credibility filter is built for. Five to eight people, most with records, holding a reasoned position while the price argued with them for three weeks, and a score that stayed with them instead of with the price. The [Coherent signal page](/stocks/COHR) has the tracked mention history and where the name sits in our coverage. The live score, the sentiment split, and the accounts behind it are in the app. --- *Quantral surfaces signals and context from public sources to support your own research. Nothing here is financial advice or a recommendation to buy or sell. Past signals are not indicative of future results. Mention counts and credibility figures cover the accounts Quantral tracks, bucketed by US Eastern trading day. Prices are closing prices via public markets through the August 4 close. Historical scores are reconstructed by replaying the app's own scoring function as of each date, with author credibility built only from predictions that had already resolved by that date; the same replay reproduces today's live score exactly. Positions attributed to accounts are their stated views, not Quantral's.*