Latest / Tech IPO Conversations with Fexingo: Public Software Companies and Capital Markets / How Post-IPO Companies Use Earnings Call Tone Analysis
Transcript
- Lucas: Luna, I've been digging into something that's quietly become a big deal in the post ipo world — companies analyzing the tone of their own earnings calls. Not just what they say, but how they say it, and what that does to their stock. Luna: You mean like, running their own transcripts through sentiment analysis to see if they sound too bullish or too cautious? Lucas: Exactly. And it's not just internal curiosity. There's a whole ecosystem of software now — think NLP models trained on thousands of earnings call transcripts — that can flag a single sentence that might spook the market. We're talking about companies literally rehearsing their tone before they go live. Luna: I remember reading that one study found the tone of the CEO's opening remarks explains something like 30% of the stock's move in the two days after the call. Is that the kind of data they're using? Lucas: That's exactly the stat. It's from a paper a few years back, and it's become gospel in investor relations circles. So naturally, post ipo companies — especially ones that are still volatile, still building credibility — are taking this very seriously. Look at Meta just last week. Their Q2 call had this moment where Mark Zuckerberg said AI spending would 'ramp significantly' — and the stock dropped 3% intraday before they clarified. Luna: Right, because 'ramp significantly' sounds open-ended, like costs could balloon. If they'd run that phrase through a tone model beforehand, they might have softened it to something like 'we expect measured increases.' Lucas: Exactly. And some companies are doing exactly that. I spoke with a startup called ToneSignal — they work with about 60 post ipo companies. They take the draft script, run it through their model, and flag sentences that score high on 'uncertainty' or 'negative sentiment.' The company then rewrites those lines. It's like a spell-check for investor confidence. Luna: But isn't there a risk of over-optimizing? If everyone sounds perfectly confident all the time, doesn't that become suspicious? I mean, markets can smell a scripted answer. Lucas: That's the tension. The data shows that calls that score too positive — like, in the top 5% of positivity — actually underperform the next month. Investors start wondering, what are they hiding? So the goal isn't to be relentlessly upbeat. It's to match tone to reality. If you beat estimates by 10%, you should sound confident. But if you missed, a cautious tone is actually more credible. Luna: So it's about calibration, not manipulation. That makes sense. And I imagine the AI models themselves are getting pretty nuanced — they can detect sarcasm, hedging, even vocal tone if you feed them audio. Lucas: Absolutely. Some firms are now doing real-time analysis. They have a dashboard in the room where the IR team can see a sentiment score tick up or down as the CEO speaks. If it dips, they can signal the CEO to clarify or add a reassuring sentence. It's like a live teleprompter for tone. Luna: That feels a little Big Brother, but I get why it's useful. And it's not just the call itself — they're also analyzing the Q&A. Because that's where the real risk is. A offhand comment can tank a stock. Lucas: Right. And companies are using the same tools to prep for analyst questions. They feed the model expected questions, generate answer options, and score each for tone. So by the time they're on the call, they've already rehearsed the safest phrasing. It's a huge shift from even five years ago, when most IR teams just read the script cold. Luna: I have to ask — does this actually move the needle? Is there a study showing that companies using tone analysis have better post-call stock performance? Lucas: There's preliminary evidence. a 2024 paper from Wharton looked at a sample of 200 post ipo companies — half used some form of tone analysis, half didn't. They found that the ones using it had, on average, a 1.2% smaller absolute stock move — up or down — in the hour after the call. That's a significant reduction in volatility. And for a company with a $10 billion market cap, that's $120 million in avoided value swings. Luna: Wow. So it's not just about sounding good — it's about reducing uncertainty. And less uncertainty usually means a lower cost of capital. Lucas: Exactly. And that's the real prize for post ipo companies. They're still trying to establish a stable investor base. Every volatile earnings call pushes that further away. So if tone analysis can smooth out those bumps, it's worth the investment. The tools themselves aren't cheap — a subscription to a service like ToneSignal runs around $150,000 a year — but against a potential $120 million swing? That's a no-brainer. Luna: But there's also a risk of groupthink, right? If every company starts using the same models, all earnings calls start to sound the same. And that might reduce the information content of the call overall. Lucas: That's a real concern. Some academics argue that if everyone optimizes for the same 'safe' tone, the market loses a valuable signal. The idiosyncratic, honest moments — even if they cause short-term volatility — can be informative for long-term investors. So there's a trade-off. You don't want to sanitize the call into uselessness. Luna: It reminds me of the movie 'The Invention of Lying' — where everyone just says exactly what they think, and it's chaos, but also kind of honest. Then someone invents lying, and suddenly everyone's polite but you can't trust anything. Lucas: Ha. That's the risk. But I think the smart companies are using these tools to flag trouble spots, not to rewrite the entire script. They keep their natural voice, but they avoid the landmines. And honestly, if you're a shareholder, you'd rather the CEO not say something that tanks the stock on a poorly worded phrase. Luna: Agreed. And it's not just the CEO — the CFO, the IR head, anyone who speaks on the call. I've seen cases where a CFO's offhand 'we're seeing some headwinds in Q3' sent the stock down 5%, when the actual numbers were fine. A tone check could have caught that. Lucas: Exactly. And this is especially important for post ipo companies because they're under a microscope. Every word is parsed by analysts and algorithms. So they need to be deliberate. I think we'll see this become standard practice within the next two years — like having a legal team review the 10-K before filing. Luna: Speaking of standard practice, if today's conversation was useful to you — maybe you're at a post ipo company thinking about your next earnings call — consider supporting the show. We keep it ad-free thanks to listener support at buy me a coffee dot com slash fexingo. Every bit helps us keep digging into topics like this. Lucas: Yeah, and it's genuinely what allows us to stay independent. No sponsors, no pressure to soften our takes. So if you find value in these episodes, that's the way to keep them coming. Luna: Alright, back to tone analysis. One more thing — are there any public companies that have openly talked about using these tools? Or is it all happening behind the scenes? Lucas: Mostly behind the scenes. IR teams don't love advertising that they're 'optimizing' their calls — it sounds like they're being inauthentic. But I have seen a few mentions in earnings call transcripts themselves. For example, on Nvidia's Q2 call, Jensen Huang joked that their IR team 'ran my remarks through the sentiment model and told me to smile more.' That was a rare public nod. Luna: And I bet Nvidia's stock didn't exactly suffer from that call. So maybe there's a middle ground — use the tools, but keep a human touch. Let the CEO's personality shine through, just without the accidental bombs. Lucas: That's the sweet spot. And I think that's where the industry is heading. The technology is a safety net, not a straitjacket. If companies can maintain authenticity while avoiding costly missteps, everyone wins — investors, management, and the market overall. Luna: Well, and on that note, I think we've given our listeners a lot to think about before the next earnings season. Lucas, great conversation. Lucas: Thanks, Luna. And to our listeners — go run your own sentiment analysis on an earnings call transcript. See what you find. It's eye-opening.