We scored 121 earnings calls. Then the market voted.
This past results season, the 121 companies covered on this platform announced their January–March (Q4 FY26) numbers and held their earnings calls. An AI read all 121 transcripts and scored each call out of 10. Group the companies by that score, and here is what their stock prices did over the two weeks that followed:
Average move vs the Nifty over the 10 trading days after Q4 FY26 results, by score bucket. 121 companies. The dashed line is the market itself.
The four lines never cross. Companies whose call scored 8 or higher ran 20% ahead of the market within two weeks. Companies scoring below 6 fell about 4% behind in the first three days — and stayed there. The middle two sit in the middle, in order.
Nobody was trading on these scores. They were simply sitting on the website while all of this played out.
Two things about that chart deserve challenge — the Nifty is the wrong yardstick for companies this size, and some of these companies joined the platform after their results were public. Both get their own section below, with the numbers redone. Both shrink the effect. Neither removes it.
Wait — what score?
Every quarter, each listed company holds an earnings call. Management explains the results, analysts push back, and a transcript of the whole thing gets published — usually 20 to 40 pages of it. Almost nobody reads them.
We do. Every transcript, for every company we cover, goes to an AI that reads it the way a careful investor would: the reported numbers, what management promised, how they handled the uncomfortable questions. It prints one number out of 10 for how that quarter reads. That number sits on every company page here.
The obvious question a first-time visitor asks is whether that number means anything, or whether it's astrology with extra steps. Last time I tested it against companies' reported profits. The chart above is the harsher test: after results are announced, does the stock actually do what the score said the quarter read like?
How I ran it
No cherry-picking. I took every company on the platform that reported January–March 2026 results and had a results date on the exchange calendar — 121 of them. For each one:
- Note the day it announced results (from the NSE's own board-meeting calendar).
- Note the score our system had given that quarter's call.
- Measure the stock's move over the next two weeks, minus the market's move (if the market rose 1% and the stock rose 4%, the stock gets credit for 3%). That strips out good-market/bad-market luck.
Two things worth stating plainly. The score is computed only from the call documents — it never sees the stock price. And the scores weren't tuned to fit this result; the scoring logic was frozen months before I ran the test.
The full numbers
Averages can hide sins, so here is every bucket in full:
| Call score | Companies | Avg vs market, 2 wks | Median | Beat the market |
|---|---|---|---|---|
| 8 or higher | 11 | +20.1% | +21.0% | 9 of 11 |
| 7 – 8 | 50 | +5.4% | +2.8% | 31 of 50 |
| 6 – 7 | 35 | +2.4% | −0.1% | 17 of 35 |
| Below 6 | 25 | −4.1% | −3.4% | 10 of 25 |
Or as a picture — every dot is one company, coloured by whether it beat the market over the two weeks:
beat the market over the next two weeks didn't
Nine of the eleven calls that scored 8+ beat the market over the fortnight. For the statistically minded: the score-to-return correlation is +0.43 across all 121 companies, with p under 0.001 — this is very unlikely to be luck.
Does it last, or does it just bounce?
Two weeks is short. If the score is genuinely reading the business, the gap should hold or widen as that business plays out. If it's a knee-jerk reaction to a headline, it should close.
Before the answer, a correction I owe you. Every company we cover is a mid- or small-cap, and mid-caps as a group beat the Nifty by about 13 points over these months. Across two weeks that's worth roughly a point — irrelevant, which is why the chart at the top stands. Across two months it would be most of the story. So for the longer test I stopped measuring against the Nifty and used the Nifty Midcap 50: a fairer yardstick, and a harder one.
Average move vs the Nifty Midcap 50 at each horizon after Q4 FY26 results, by score bucket. All 121 companies at every point shown.
It widens at every step. Two months after results, the calls that scored 8 or higher were 32% ahead of the mid-cap market, while the below-6 group had gone precisely nowhere. The gap between best and worst grows from about 10 points at one week to 19 points at two months, and the four groups never swap order at any horizon.
That is the opposite of what a knee-jerk reaction looks like. A market that overreacted to a headline and then thought better of it would show these lines converging.
What I won't claim: three months. I can only measure that far for 50 of the 121 companies, because the rest reported too recently. Worse, 37 of those 50 had their next quarterly results land inside the three-month window — so any number there is partly measuring the following quarter's news rather than this one's call. The gap does keep widening. I'm not counting it as evidence, and neither should you.
Spotlight: every top call we had on the board in advance
Averages are easy to hide behind, so here are the individual names — and to get there I have to admit something about the 121.
We add companies to the platform continuously, and a company that has just posted a spectacular quarter is more likely to catch our attention. Of the 121, thirty-seven were added after their own results were already public. For those, our score went up after the fact. They are still a fair test of whether the score reads a call correctly, but they are not a test of whether it called anything in advance.
So the stricter question: among the 84 companies that were already on the board before the call happened, which ones scored 8 or higher — and what did they do?
There are exactly four. Not the best three of four. All of them:
| Company | Score | 2 weeks | 1 month | 2 months |
|---|---|---|---|---|
| Sansera Engineering | 8.2 | +22.9% | +21.1% | +31.3% |
| Laurus Labs | 8.0 | +17.0% | +20.4% | +32.8% |
| TD Power Systems | 8.0 | +9.7% | +0.8% | −7.5% |
| Sagility | 8.2 | −8.7% | −9.2% | −8.7% |
Move vs the Nifty Midcap 50 after each company's own results date. Sansera reported 20 May, Laurus 30 April, TD Power 14 May, Sagility 12 May.
Two of the four ran away from the market and kept going. One drifted up and then gave it all back. One was simply wrong — Sagility's call read as excellent to the machine and the stock lost ground for two straight months.
That is a hit rate of two out of four at the very top of the scale, and it is the number I would want to know before trusting any of this. A score of 8 is not a buy signal. It says one quarter read unusually well, and roughly half the time over the following two months, the market disagreed or lost interest.
The same correction applies to the whole study. Rerun everything on those 84 pre-existing companies and the signal survives but roughly halves: the gap between the strong and weak groups is 6 points at two weeks and 10 points at two months, against 12 and 19 for all 121. The correlation drops from +0.43 to +0.30, which is still very unlikely to be chance. The direction holds everywhere. The magnitude was flattered by which companies we happened to add.
What the machine actually read
The score isn't magic. It reads what management said and how they said it. The extremes from this quarter show exactly what that means.
The lowest score of the season was Kaynes Technologies at 3.2. The scorer's written reasons: the company had missed its own revenue guidance by roughly ₹900 crore, withdrew numerical guidance entirely and replaced it with a vague promise to grow at "double market growth," and gave long, non-committal answers when analysts pressed on a working-capital blowout. The stock fell 22% behind the market over the next two weeks.
The highest was Fineotex Chemical at 8.8: revenue up 162%, an acquisition target moved forward by two years, debt-free with cash on the books. The stock went 54% ahead of the market in a fortnight.
My favourite detail is smaller. Schneider Electric Infrastructure's management, asked about the coming quarter, said "fingers crossed." The scorer flagged the outlook as uncertain and scored the call 4.9. Two weeks later the stock was 18% behind the market. Sometimes it isn't deep analysis — someone just has to actually read the whole transcript. Every quarter. For a hundred companies. That's the part we automated.
Before you get excited
An honest list, because one good chart can make anyone reckless:
- This is one season. 121 companies, January–March 2026 results. I ran the same test on the quarter reporting right now and it's noisier so far — the low-score bucket still lags, but the high scorers aren't popping the same way. When those windows close, I'll publish it, whichever way it lands.
- The two 8+ scorers that lagged are instructive. A great quarter everyone already expected is often already in the price. The score reads the business, not the crowd's positioning — it can't tell you what's priced in.
- The headline numbers are flattered by hindsight, as the spotlight section explains. Thirty-seven of the 121 were added to the platform after their own results were public. On the 84 that were already covered beforehand, the effect is about half as large — real and statistically significant, but half. Take the 84-company version as the honest one.
- The top bucket is thin. Among those 84, only four calls scored 8 or higher. Every average for that group rests on four companies, and two of them didn't work.
- Averages aren't promises. Within every bucket there are companies that did the opposite. 40% of the weakest bucket still beat the market.
- This is a research note about whether our scoring works, not investment advice.
Where this leaves us
The score set out to answer a modest question — how did the quarter read? — and it turns out the market's verdict usually matches, not just on the day but for months afterward. I find the low end the most useful part: a bad call, scored honestly on the day it happened, kept underperforming. The market did not shrug bad quarters off.
The platform scores these calls the day the transcript appears, often the day of the call itself. Next season's scores are going up on company pages right now — the same numbers this study just tested, printed before anyone knows how the market will vote.