Greetings from Team Carnelian!
Carnelian strategies performance at a glance…
A life — or a portfolio — examined only in its successes is only half understood.
Over the past seven years, we have written to you every month about our investment philosophy, the big trends and themes we saw emerging, and our views on markets and the macro environment and implemented them in investing your money with full conviction. In this letter, we want to do something different: write about the other side of that conviction — our mistakes. People with high conviction can still be wrong in their judgement, and no professional investor can honestly claim never to have made one. We have got far more right than wrong over these seven years, but we still carry our own share of mistakes. We owe our investors an honest account of them, because it was your hard-earned money, not just ours, that was on the line.
So let’s talk about some of our mistakes and what we learned from them. The best thing one can do with a mistake is look at it dispassionately, and honestly. By sharing ours openly, we hold ourselves accountable and ensure that these lessons become institutional knowledge, not just personal memory.
Before we go further, it’s worth distinguishing between a mistake and an accident. A winning outcome doesn’t always mean a good decision, and a losing one doesn’t always mean a poor decision. Learning to distinguish between the two is the real skill.
How do we distinguish?
Our framework is simple. After every investment/event, we ask ourselves one question:
If this same trade or opportunity were presented to us again today, with the same information we had then, would we make the same decision? and why?
That single question is where real learning begins.
If the answer is YES, the poor outcome was more of an accident than a mistake. If the answer is NO, then it was a mistake. Understanding why the answer changed, whether because of flawed analysis, faulty assumptions, behavioural biases, or gaps in our process is where the real learnings worth billions lie.
Charlie Munger often argued that, edge in investing rarely comes from brilliance, it comes from removing blind spots, one at a time, before they cost you money. As he put it, “acknowledging what you don’t know is the dawning of wisdom.” Every mistake in this letter, in hindsight, is one such blind spot. Something we didn’t know, until the market made it visible.
The same is true well outside markets. Most errors in life aren’t repeated because they are hard to avoid, they are repeated because they are never honestly examined. A relationship, a career call, a business decision, a health habit: the pattern that hurts you the second time is rarely new, it is an old blind spot that was never corrected the first time.
Reflection is simply the discipline of looking directly at the thing you’d rather not look at, before life makes you look at it the hard way.
This letter is an attempt to practice that discipline.
When we first analyzed the business, we liked it. It was an integrated bagasse-to-paper player with a long history of delivery having a ten-year CAGR of 20% and an ROCE profile of 20%. What drew us even further was a professionally run team and genuinely institutionalized operations. At the plant, for instance, live production and profit numbers are displayed prominently for every employee to see, and the whole team is aligned and rewarded around a collective outcome. Management and ownership were clearly segregated, a separation rarely seen in a small-cap company of this size. We visited the plant and met people four levels down; twenty-year tenures were common, the company housed and fed its workers, and the tone throughout was one of quiet competence. Our two separate visits, by two separate teams, came back with the same conclusion.
The company was looking to raise capital for a ₹650 crore brownfield expansion at the same location. It also shared an innovative research breakthrough that could replace plastic packaging for chocolates and other premium products with paper-based alternatives, along with plans to expand into Guatemala, where financial closure with a PE fund was almost finalized.
Our thesis was that this expansion would move the company into a different league altogether, taking EBITDA from ₹85 crore to roughly ₹225 crore, a potential 4–5x story if it played out over three years. The existing base was steady. They were getting into a new product which was a complete import substitution opportunity and offered superior margins and ROCE profile which attracted us. We were skeptical of the scale of the Guatemala expansion specifically, but we hedged it by capping the company’s exposure there at ~₹90 crore, with no further corporate guarantees.
What went wrong: We invested in October 2024. In March 2025, tariff announcements hit, and the currency moved against the European machinery being imported for the capex, pushing the project cost up by ₹100 crore. At the same time, the core paper cycle turned, margins compressed and cash flow got stuck, which together delayed financial closure and pushed back commissioning. In August 2025, Hira resigned as CEO and promoter Ved Krishna took over as MD. The environment for small and mid-caps turned hostile on tariff-related sentiment, and the stock came under heavy pressure. Margins for the second quarter fell to 6%, from a usual range of 18–20% (though recovered later), and the stock lost roughly 60% of its value.
What we missed: It wasn’t management quality or business quality as both held up well. What we missed was properly sizing the risk of the ₹650 crore expansion itself, against a company whose total lifetime capex until then was under ₹200 crore. We didn’t assess the risk carefully enough of what it means for a company to take on capex disproportionately larger than anything in its history, what could go wrong, and how sensitive the project was to a cost or time overrun and business cycle. It’s basic diligence, in hindsight, but the confidence the management team inspired in us let it slip. The one thing that worked in our favor was that we had already hedged our exposure through how we structured that piece of the deal.
Learning: Always pay close attention whenever a company’s capex is disproportionately large relative to its historical scale and build time and cost-overrun sensitivities into the model upfront. Most of the time, we would now rather not fund that scale of expansion at all as it creates a drag on both management’s time and the company’s balance sheet. Where the risk reward still justifies backing it, size the position only after building in those overruns, not before. It’s possible this would have worked out fine had the tariff shock not hit, but the environment is never fully within anyone’s control, and we have to be prepared for that regardless. Since building this into our process, we have passed on several ideas we would once have backed.
2. SpiceJet — grounded by the supply chain
We participated in SpiceJet’s fund raise in September 2024, backing the turnaround for two reasons: a strong sectoral tailwind of rising demand and low competitive intensity, and the fact that the same management had engineered a turnaround at this business once before, in 2018. Management’s case was that roughly 40 aircraft were grounded purely for want of repair and maintenance, against a flying base of only 20 planes. Once the dues were cleared and those planes would return to air, that alone represent enormous operating leverage. Management estimated that ₹2,500 crore would clear all debt, including dues to operational creditors, and leave enough to get the grounded fleet flying again. Investors, including several large, marquee names, committed to raising ₹3,000 crore in total, with some room built in to avoid another financial crunch down the line.
The thesis was simple and, at the time, well-supported: SpiceJet had just posted its second straight profitable quarter, a Supreme Court ruling had cleared the Kalanithi Maran overhang, and management planned to use the proceeds to clear statutory dues and settle creditors like Credit Suisse and Carlyle, alongside getting the grounded fleet back in the air. Various estimates suggested that with the full fleet flying, the company could generate a PAT of roughly ₹500 crore in FY25–26, which made the setup look very attractive.
What went wrong: The thesis was strong on paper, but it failed to underwrite one major risk: the supply chain. We leaned on management’s confidence that the fleet could be turned around quickly, and our own diligence on this front wasn’t rigorous enough. This cycle’s binding constraint turned out to be not only capital, but a supply chain bottleneck in aircraft and engine availability. Most of the fleet runs on engines from CFM International, the GE Aerospace–Safran JV, and 2024–25 turned out to be exactly when engine and MRO turnaround times blew out industry-wide, by as much as 150% in some cases. Grounded capacity and delayed inductions meant capital came in, but the fleet couldn’t scale fast enough to convert it into revenue, and the company slid into a spiral of operating and cashflow losses, compounded further by currency volatility around the tariff announcements.
The bigger mistake: We consider what happened next to be a bigger mistake than the original investment itself. Post our investment, in the quarters that followed, we didn’t get comforting answers from management in our reviews. We even interacted with the new COO brought in from IndiGo to turn operations around, and it became clear after that conversation that far more capital would be needed, and that the turnaround would not be easy. We waited for the December-quarter results before deciding to exit, and by then the situation had deteriorated further. Not cutting the position at a 40% loss was, in the end, a costlier mistake than the original call. We finally sold at a 60% loss.
Learning: A turnaround investment demands very detailed analysis, especially when there are these many moving parts. We violated one of our own core principles: the moment a mistake is recognised, act on it. Instead, we waited for the next quarter’s results instead, letting short-term temptation win out over our own principle. Our first principle exists precisely for moments like this: Once a mistake is recognized, correct it immediately, irrespective of implication. The one thing we can take some comfort from is that it was a member of our own team who first called this out and pushed us to cut the position.
3. Quick Heal — a proven CEO is not a proven product
Our investment thesis was that Quick Heal, a legacy consumer antivirus company, could successfully reinvent itself as an enterprise cybersecurity business. The company had already invested over ₹100 crore in building its enterprise platform and had appointed Vishal Salvi as CEO—a 29-year cybersecurity veteran and former Global CISO at Infosys, where he had led a business several times larger than Quick Heal. We viewed the promoters’ willingness to step aside in favour of professional management as a positive signal.
Our channel checks on Vishal were consistently encouraging. When we asked why he was joining a company of this size, his answer was straightforward: the platform had already been built, only a few pieces remained to be completed, and he believed the business could now be scaled. In his view, the missing ingredient was not technology but sales execution. He intended to build a stronger enterprise sales organisation, and early signs—including a large contract with a leading logistics company—appeared to support that thesis.
The broader backdrop was equally favourable. Enterprise cybersecurity, particularly a made-in-India offering, stood to benefit from rising geopolitical sensitivities, increasing digitalisation, and the implementation of the Digital Personal Data Protection Act. The company also had around ₹100 crore of cash on its balance sheet and continued to reinvest roughly 10% of revenue into R&D.
We participated in a block deal at around ₹200 per share in September 2023. The stock subsequently re-rated by nearly 3.5x, reinforcing our conviction that the transformation was underway.
What went wrong: We tracked the business closely. The first few quarters showed little improvement, but we believed management deserved time to execute. With each subsequent meeting, however, we came back with less conviction. We did not receive clear or reassuring answers on strategy or execution, yet we failed to treat those interactions as the warning signs they were. Instead, we allowed the rising stock price to reinforce our confidence.
Vishal resigned in July 2025. By then, the more important mistake had already been made: we had ignored the accumulating evidence that challenged our original thesis. His resignation should have been the final trigger to exit, but we delayed once again. The broader correction in small-cap stocks only amplified the outcome. In hindsight, we should have read his resignation itself as a signal — that he may have lost confidence in the company’s ability to deliver on the product’s potential. We may never know the full story, but from the numbers since, our own reading is that the business’s challenges have simply overpowered that potential.
Learning: Our mistake was never the original investment. It was not sensing the early warning signs of the challenges the CEO was facing and not acting quickly enough after he left. We took too much comfort in the rising price and didn’t notice the widening gap between that price and actual delivery. We didn’t lose money on our original entry price but lost money on what we bought afterwards. A deteriorating market made the fall worse, but that doesn’t change the core lesson: once a mistake is recognized, correct it, not wait for the environment to improve first.
There was a second mistake sitting underneath the first: anchoring. At the time of investment, we described this internally as “the next Persistent Systems.” Once you build an anchor like that, it quietly colors every piece of data you look at afterward, however objective you intend to be. (For context: when digitization was still an early-stage theme in 2012–13, we backed Persistent Systems as one of the pioneers of that shift, at a market cap of around ₹3,500 crore; it is worth roughly ₹85,000 crore today. A 20-bagger in what was otherwise a slow-growth industry seems like a minor miracle at the time, but it came from correctly identifying both the theme and the right company early. We thought we had found the same setup in cybersecurity, and that anchor, a classic case of anchoring bias colored our judgement more than we realized at the time.
We are still not writing Quick Heal off. Persistent went through several rough patches of its own before compounding into a 20-bagger, and Quick Heal may yet follow a similar path, the core business is still intact. What cost us our patience wasn’t the price movement alone; it was the absence of urgency from management to fix what was broken.
4. Polycab India — the concern that got fixed, and the position that didn’t
We invested in Polycab at the IPO. The thesis was straightforward: a first-generation promoter with a strong market reputation, expanding into FMEG and exports on top of an already dominant wire and cables business. Our channel checks with distributors and retailers came back positive on both the company and its products.
What changed our view was this: reading the company’s annual report closely, we found it was classifying acceptances as other liabilities rather than as debt, which we read as a corporate governance concern. Alongside that, an export order it had signed wasn’t ramping up and the margins promised on it weren’t showing up, and the FMEG business was losing money. We sold, having doubled our money on the position.
We later learned that the accounting treatment was not specific to Polycab at all, it was a convention followed across the industry along with the export disruption turning out to be temporary. The company’s CFO, in fact, listened to our concern on the acceptances directly, and the very next quarter began disclosing it clearly in the investor presentation. Yet we never revisited our original conclusion or rebuilt the position. From our exit, the stock compounded by roughly another 5x, while continuing to outperform earnings expectations quarter after quarter.
Learning: This was one of the first investments we made after founding Carnelian, built on a strong positioning in the industry and a focus on forensic research capability, something that still remains core to how we operate today. At the time, though, we were carrying the weight of applying that framework at 100%, without the practical context that experience later gave us. Not every forensic red flag carries the same weight, and treating a minor, industry-wide accounting convention with the same suspicion as a company-specific governance issue is itself an error. But the sharper failure wasn’t the original assessment. It was that, having doubled our money, we treated the exit as a closed, correct decision and never revisited the company again, not even when the CFO directly addressed the exact concern we had raised. We have written before about our risk framework (link) and about Type C risk, the risk of opportunity lost through inaction as another important risk to keep in mind alongside the risk of permanent capital loss. Polycab is the cleanest example we have, of paying that price twice over: once for the faulty conclusion, and a second time for not revisiting the name after the company had addressed our concern and gone on to deliver stupendous growth.
5. Ola — a mistake even though we made money
We invested at the IPO and added more on listing, at around ₹70. Our thesis was a rare combination: an early mover in a fast-growing EV category with a large addressable market, the only player with a fully integrated model from cell to battery to vehicle, in-house R&D, manufacturing, and wide marketing reach, backed by a passionate promoter and a genuinely strong R&D team. The business was executing well at the time: market share was rising, growth was visible, and the story was working.
We usually avoid loss-making companies, but here we saw significant operating leverage building over the next three years and expected the price to double over that period.
As it happened, the stock rallied to ₹140 within a few days of listing. We exited, on the reasoning that we were getting in a week what we had expected to take three years, and that these prices wouldn’t hold once the listing euphoria settled. This was August 2024, a genuinely euphoric market, and we have always been skeptical of loss-making new-age names in that kind of environment. Post our exit, the stock fell as low as ₹25, on concerns about customer experience and an aggressive, cash-burning store expansion, as the euphoria wore off and the broader market turned poor.
By the normal scorecard, this shouldn’t be on a list of mistakes. But it is, deliberately so. On honest introspection, we hadn’t done enough channel-check work on the customer experience side before investing. Had the price not appreciated post listing, we might well have continued holding the stock through the drawdown that followed. We had acted in a rush to meet the IPO timeline, not with the diligence the position deserved.
Learning: The real mistake surfaced only after the trade was done. Some of our team members began describing it as a smart, well-timed move, buying and selling at exactly the right moments. That is where the actual lesson lives, more than in the trade itself: a profitable trade is not always a good decision, just as a losing trade is not always a poor one.
We ran it through our own golden question: If we were presented with the same opportunity again, would we make the same decision? The answer is no.
So, the real mistake wasn’t the investment; it was letting a good outcome get mistaken for good judgement. It was a sheer accident dressed up as a decision, and it left us with a lesson.
What ties these together
Look at them together, and the pattern has less to do with the companies than with recurring gaps in our own process.
First, we didn’t see or articulate some risks we were underwriting at entry, the risk that ended up deciding the outcome. In Pakka, it was the risk of a disproportionately large capex programme, not management or business quality. In SpiceJet, it was the supply chain, aircraft and engine availability and not capital or promoter intent. Call it a blind spot.
Second, as the situation changed, we failed to revisit the original hypothesis with the same rigour we had used to build it in the first place. Call it anchoring bias.
Third, most of these mistakes were born in the middle of 2024, when markets were Euphoric, activity was high and our own risk guardrails were lower than they should have been. Ola is the clearest example of this. We were fortunate not to make many material mistakes during that period, but the learning still has to be drawn from the smaller ones. Cycles matter, pay close attention.
None of these are mistakes of intelligence. Every one of them was made by people who had done real work and had a genuinely defensible thesis at the time. That is Munger’s whole point, and it’s precisely why it is hard to act on: the money is not lost to bad ideas, it is lost to good ideas that were missing one honest question.
If we had to summarise what these mistakes changed, it is this: today we ask better questions. What is the single most important risk we are underwriting? What evidence would make us change our mind? How much of this story belongs to the company, and how much belongs to the cycle? And what, specifically, would make us sell?
Closing & disclaimer:
We usually avoid naming companies in our communications, especially when the context isn’t a positive one. But if we hadn’t done so in this letter, many of our investors would have been left without the context needed to follow along.
We want to be very clear: nothing in this letter is meant to pass judgement on any of these companies, or to put them in a poor light. These were our mistakes, not theirs. What reads as a mistake for us may well have simply been an accident for the company itself. Pakka, for instance, would likely be at a very different stage today had the tariff shock never happened.
Most companies do learn from their own mistakes/setbacks and change, and we continue to track each of these names closely. We may well buy some of them again in the future. We met Bhavish from Ola recently, who is very passionate about the business and working hard to fix the customer experience. Pakka may yet prove to be a multi-bagger once the expansion is behind it, since the company’s core capability remains intact. We have seen this pattern play out many times in our decades of experience.
Our aim in sharing all this with our investors and partners was simple: any experience, or any mistake, examined without reflection is wasted. Reflection doesn’t just stop us from repeating the same faulty pattern but improves the way we see the problem in the first place. It pushes us back to first principles and forces us to trace the second-order effects we skipped the first time around. Done honestly, it is also what removes our blind spots and sharpens our awareness of the next one already forming.
We would simply request our readers to try applying this same Mistake-vs-Accident framework to their own investing, business, and personal decisions. This can truly help one become a better investor, leader, entrepreneur, and human being. We are certainly at it and constantly trying.
Thank you for reading this patiently. This is the longest letter we have ever written. We could have gone on for many more pages, but we wanted to be mindful of your time.
Thank you, as always, for the trust you continue to place in us, despite all our mistakes.
