THE SITUATION

In January 2016, a former Google executive named Adrian Aoun incorporated a company that would try to do to the doctor’s office what software had done to almost everything else. Aoun was not a physician. He had run special projects for Larry Page, built an AI startup Google acquired, and spent time at Sidewalk Labs. The thesis behind Forward was that primary care — slow, expensive, rationed, human — was a product waiting to be re-engineered.

For its first several years, Forward looked like it might be right. It opened a chain of membership clinics — eventually nineteen to twenty-five locations, staffed by more than a hundred primary care clinicians, charging $149 a month, cash-pay, no insurance. The experience was genuinely nicer than the system it competed with. Capital agreed. Across five rounds Forward raised roughly $657 million, and at its March 2021 Series D it crossed into unicorn territory: a $1 billion valuation, with SoftBank’s Vision Fund 2, Founders Fund, Khosla Ventures, and Marc Benioff on the cap table.

Then, in November 2023, Forward announced what it had really been building toward. Not more clinics — CarePods: AI-powered, self-service diagnostic kiosks you would step into without a doctor present, do your own blood draw, and let the hardware run the visit. The plan was to deploy 3,200 of them within a year. Aoun reached for the obvious comparison: the clinics had been Forward’s Model S, the expensive proof of concept; the pods were its Model 3, the version that scales to the planet.

Twelve months later the count was not 3,200. It was five.

On November 13, 2024, Forward shut down. The app went dark, every location closed at once, and about two hundred people lost their jobs. A company that had been worth a billion dollars three years earlier simply switched itself off.

The easy filing is “tech founder over-engineers healthcare, runs out of cash.” That filing is not wrong, exactly. It is just aimed at the wrong moment. Because the clinics — the part of Forward that actually worked, the part patients were still using — did not fail. Forward took them apart on purpose.

It didn’t abandon what worked because customers walked away. It abandoned it because its own valuation told it to.

WHAT ACTUALLY HAPPENED

The stated rationale, from roughly 2017 onward, was scale. Healthcare reached too few people too slowly; the answer was to migrate everything doctors and nurses did onto hardware and software and roll it out everywhere at once. Capture the category, ship the machine, fix the margins later.

What was actually happening underneath that rationale was narrower and more uncomfortable. Forward had built a real business — but a capped one. The clinics worked at $149 a month in dense, affluent cities. They held a billion-dollar mark for three years. By every public signal Aoun ever gave, they were not failing; he called them a “critical development layer,” a stepping stone, never a mistake. What they were was structurally incapable of producing the returns their own valuation now demanded. A premium urban clinic chain is a good business. It is not a venture-scale one.

That gap — between a good business and the business the cap table had been priced for — is the whole story. Forward never disclosed clinic-level economics, membership counts, or churn, so the honest description stops at working-but-capped; “profitable” is a word the record doesn’t support. The company was, in fact, almost silent on numbers of any kind — which is what makes the one concrete figure it did commit to, 3,200 pods in a year, the cleanest measure we have of the distance between Forward’s self-image and its reality. But working-but-capped was enough to set the trap. Because the moment a company carries a billion-dollar valuation funded by blitzscale-era investors, “we are a very good premium clinic business” stops being a success and starts being a write-down.

Three distortions did the damage. They are not three unrelated mistakes — each one made the next easier to miss — but the record doesn’t support pinning down a single clean chronology, so what follows is presented in the order that makes the mechanism clearest, not as a claim that one finished before the next began. And this time the first one to take hold was not in the boardroom or in the venture market. It was in the founder’s idea of what medicine is.

THE DISTORTION LAYER

First: Bias — the engineer’s view of medicine

Competence transference is the assumption that demonstrated mastery in one domain carries into another where it has no tools. It is one of the most reliable distortions in founder-led companies, and the mechanism is almost never arrogance. It is the sincere belief that first-principles thinking can stand in for the domain judgment other people spent thirty years acquiring.

Aoun stated the bias plainly and without embarrassment, which is what makes Forward such a clean specimen. He described himself as not a medical doctor but a computer scientist. The goal, he said as early as 2017, was to scale doctors the way “an engineer can scale through software.” By 2023 the worldview had hardened into a slogan: “Healthcare should just be a product.” Take everything clinicians do, migrate it to hardware and software. “We don’t even believe a doctor’s office should exist.”

Read those statements as a thesis and they are merely ambitious. Read them as a lens — the thing through which every subsequent decision got interpreted — and you can see the damage they were about to do. Because if medicine is “just a product” awaiting migration to a machine, then a building full of human clinicians is not your most valuable asset. It is scaffolding. It is the thing you climb up and then kick away. The bias didn’t only produce the pod bet. It quietly recoded the working clinics as disposable — a “development layer,” in Aoun’s own framing — and in doing so it removed the one brake that should have stopped the company from cannibalizing the only part of itself that worked.

The Tesla analogy was where the bias put on its costume. Model S to Model 3 is a roadmap: a planned sequence, run from strength, in which nothing is failing and each step funds the next. That is a flattering story, and it was the wrong one. Forward’s clinics were not a springboard generating the cash to build the pods; they were a capped business being dismantled under financial pressure to fund an unproven replacement. The Tesla comparison wasn’t a description of what was happening. It was a vivid, available image standing in for an honest account of it — the availability heuristic doing exactly what it does, supplying a clean narrative precisely where a hard one was needed. And almost no one in the room was positioned to say so, because the people around the table were technologists and investors, not clinicians.

That’s the first distortion. The two that follow are where the company began starving the one part of itself that actually worked.

Read more

THE SITUATION

Years ago I worked with a woman who edited my speeches and columns. She was exceptional at it. I noticed something about the dynamic early: she didn’t capture what I said. She captured what I meant. The words were shaped by her craft. The thinking, the argument, the thing I was actually trying to say — that was mine. I never once felt like the substance had changed, only that it had been stated more precisely than I might have stated it myself. She was, in the best possible sense, an extension of how I think — with a better dictionary and a sharper ear for when a sentence wasn’t quite landing.

I’ve been asked how Bad Call is made. Given the current moment around AI, I want to answer that question directly. The answer is that what I just described — that relationship between thinking and expression — is the closest analogy I have for how this publication works. But the analogy only goes so far, and the part where it breaks down is the part worth explaining carefully.

WHAT ACTUALLY HAPPENS

Bad Call starts with the Decision Distortion Framework — four forces that systematically bend organizational judgment under pressure: noise, bias, accumulation, and incentive misalignment. That framework didn’t come from a single source. It came out of years of watching decisions go wrong in rooms I was in, refined against the thinking in Thinking, Fast and Slow and Noise (Daniel Kahneman), Subtract (Leidy Klotz), and The Tyranny of Metrics (Jerry Muller). Every case is selected, sequenced, and analyzed through that framework — the judgment about what a case is really about, where the distortion chain started, and what the clean version would have looked like, is mine throughout.

The research protocol I’ve developed for this is specific and demanding — two independent sources for every date and dollar figure, strict separation of contemporaneous accounts from retrospective post-mortems, and an explicit call on where the fraud line sits before any analytical frame is proposed. The research brief comes to me before any column is drafted. I review it, push back, ask questions, and challenge the proposed framework sequence until I’m satisfied that what we’ve identified is the real distortion chain, not just the surface narrative.

Then comes the conversation — and this is the part worth being precise about, because it’s the part that determines whose thinking you’re actually reading. My questions about what the research revealed. My read on which distortion entered the chain first and why. My pushback on how the framework applies to this specific case — whether what looks like bias is really noise, whether what looks like accumulation is really an incentive problem dressed as a building decision. That conversation goes deep. It accumulates into something closer to a pile of notes than a brief: my thinking about the case, expressed in real time through the exchange. By the time words are being arranged into sentences, the analytical work is already substantially done. In my voice. Through that back-and-forth.

The AI’s first draft draws directly on that pile. It is giving column form to thinking that is already mine — not generating analysis and handing it to me for approval, but transcribing a conversation into structured prose. I then work through multiple iterations, pushing back hard on anything that doesn’t reflect how I actually read the case, anything that sounds analytical rather than real, anything where the column has described a distortion rather than made you feel its logic from the inside. I don’t accept a draft until I feel like it says what I would say if I were sitting across from you explaining what went wrong and why it matters for your next consequential decision.

The woman who edited my speeches would recognize this process immediately. The instrument is different. The relationship is the same.

CARRYOVER

There is a lot of conversation right now about AI and authorship, and most of it is asking the wrong question. The right question isn’t whether AI touched the words. It’s whether the thinking belongs to the person whose name is on the piece. The case selection is mine. The framework is mine. The judgment about where a distortion chain started and what a clean decision would have looked like is mine. The research protocol that disciplines the brief is mine. The conversation that turns research into analysis is mine. What the AI contributes is what a great editor always contributed — precision, structure, the sentence that lands where my draft of it didn’t quite.

By that standard, what you’re reading in every Bad Call column is mine — expressed more precisely than I might have expressed it on my own, but mine.

The woman who edited my speeches understood that distinction without ever needing to articulate it. She wasn’t the author. She was the instrument through which I said what I meant.

Every word may not be mine. Every thought is.

— RC

Because it always seemed fine at the time.

LFB Holdings offers a free Decision Distortion assessment — 20 questions, immediate results, tailored to your dominant force. Self Assessment

Bad Call publishes every other Tuesday. One decision sequence, dissected without narrative cover. Full analysis for paid subscribers at badcall.ai

THE SITUATION

In June 2019, a 25-year-old named Johnny Boufarhat incorporated a virtual-events company in London. He had built the idea while housebound by an autoimmune condition — a platform that recreated the texture of a real conference inside a browser: a stage, breakout rooms, networking, sponsor booths. It was a reasonable product for a niche problem. For most of its first nine months, it stayed niche.

Then, in March 2020, every conference on earth was cancelled at once.

What followed is the fastest financing sequence in European startup history. A $6.5 million seed round in February 2020. A $40 million Series A in June. A $125 million Series B in November, at a $2.1 billion valuation. A $400 million Series C the following March, at $5.65 billion. And a $450 million Series D in August 2021, at $7.75 billion. Five rounds in eighteen months. More than a billion dollars raised. By the final round, a two-year-old company was worth more than storied public names like Manchester United or Marks & Spencer, and its founder — now 27 — was, on paper, the youngest self-made billionaire in Britain.

The cap table read like a complete roll call of the era’s most sophisticated capital: Andreessen Horowitz, Tiger Global, Coatue, General Catalyst, IVP, Salesforce Ventures, and the sovereign wealth funds of Singapore and Abu Dhabi. By the Series D there were at least sixty institutional investors on it.

One fact sat quietly underneath all of it. Every dollar of that valuation rested on demand that existed because in-person events were, temporarily, illegal.

Then the world reopened. In February 2022, Hopin laid off 138 people — 12% of staff. In July, it cut another 242 — 29%. In August 2023, it sold its original events platform, the product the entire company was named for, to RingCentral for $15 million up front against a peak valuation of $7.75 billion. In February 2024 it wound up its UK parent and moved what was left to Delaware. By April 2024, the remainder had been sold to an Italian acquirer for an undisclosed sum.

The temptation is to file this under “pandemic darling implodes.” That filing is wrong, and the way it is wrong is the whole point. Hopin was not vaporware. The product worked. The market was real. The people who used it during lockdown were not the victims of a hoax. The failure was subtler and far more instructive: investors priced emergency demand as if it were structural demand, and then financed the company into a cost structure that only made sense if the emergency never ended.

This is not a story about a fake company. It is a story about a real signal, misread — and then built upon, eighteen months deep, until the misreading could no longer be undone.

They priced a temporary emergency as a permanent business.

WHAT ACTUALLY HAPPENED

The stated rationale, from the Series A onward, was that the world had permanently changed — that virtual and hybrid events were not a lockdown workaround but the future of how humanity would gather. The strategic logic followed from the premise: capture the category now, before anyone else, and sort out the economics later.

What was actually happening was narrower and more fragile. Customers were not demonstrating that they preferred virtual conferences. They were demonstrating that they needed a substitute while the real thing was prohibited. Those are different facts, and the difference is everything.

The growth was not fake. This is the part that makes the case hard, and the part worth studying. Every number was real. Organizations on the platform went from roughly 1,800 to more than 100,000. Registered users went from a few thousand to seventeen million. Annual recurring revenue went from $20 million to $100 million in under a year. Headcount went from eight people to roughly eight hundred, and then past a thousand. None of it was invented.

But all of it was contaminated — not by dishonesty, by circumstance. The curve was real and the cause was temporary, and the financing treated the curve as if the cause were permanent. Four distortions produced that treatment. They did not arrive at random. They arrived in sequence, and the first one was not inside Hopin at all. It was in the market that surrounded it.

THE DISTORTION LAYER

First: Incentives — the year speed became the product

In 2021, global venture funding hit an all-time record: roughly $643 billion, nearly double the prior year, with more than fifteen hundred individual rounds of $100 million or more. Hopin’s $400 million and $450 million raises were not anomalies in that environment. They were the environment.

And in that environment, the relationship between diligence and winning inverted. When capital is scarce, slow and careful questions are how an investor protects itself. When capital is abundant and the best deals are oversubscribed, those same questions become the reason you lose the allocation. The clearest expression of the new logic was Tiger Global — a Hopin investor across the Series B, C, and D — which made 335 separate investments in 2021, roughly one every business day, on a model that was openly, almost proudly, fast: a single meeting, a review of the numbers, the highest valuation on the table, and no board seat to slow anyone down. Speed had become a deal-winning product. In a market like that, asking whether pandemic demand would survive the pandemic was not prudence. It was a way to come second.

You can see the consequence on Hopin’s own cap table. Sixty-plus institutional investors, and — in the words of one early backer, speaking to Sifted — “a bunch of sous chefs, but no chef running the kitchen.” Nobody owned the diligence. Nobody injected governance. Each new marquee name made the next one’s independent scrutiny feel less necessary, because surely someone with that logo had already done the work.

This is incentives operating as an institutional force, which is the most important thing to understand about them: nobody had to decide to skip the hard question. The reward structure decided in advance which questions would get asked, and “is this demand permanent?” was not one that paid. It set the table for everything the cognitive distortions did next.

That’s the first distortion. The three that follow are where the signal that this was a window — not a new world — kept arriving, and kept getting read as proof of the opposite.

Read more

The Situation

It’s April 2022. Convoy — the Seattle-based digital freight network founded by two former Amazon executives — closes a $260 million capital package. A $160 million Series E led by Baillie Gifford and T. Rowe Price. A $100 million venture-debt facility from Hercules Capital. A $150 million line of credit from J.P. Morgan. The valuation: $3.8 billion.

Dan Lewis, Convoy’s co-founder and CEO, tells the press that Baillie Gifford and T. Rowe Price are “long-term-oriented investors” who will give the company “the time and the space to deploy the capital in smart ways.” The raise is widely framed as a prudent move to shore up the balance sheet as the freight market turns south.

Eighteen months later, on October 19, 2023, Convoy cancels every shipment on its marketplace, lays off the majority of its remaining 500 employees with no severance, and shuts down. Approximately $920 million in equity and venture debt is wiped out. Bezos, Gates, Benioff, Bono, T. Rowe Price, Fidelity, CapitalG, Generation, Greylock, Y Combinator Continuity. Forty companies that spun out of Convoy alumni now exist. Convoy itself doesn’t.

In his wind-down memo to employees, Lewis describes what the previous eighteen months had felt like from the inside: We were running up the down escalator… and it kept speeding up.

What Actually Happened

The stated rationale, from founding through the final raise, was that Convoy was building the technology platform for a fragmented, antiquated $800 billion industry. Machine learning would match loads to trucks. Automation would eliminate the broker phone-and-fax model. Network effects would compound. The company would become to trucking what Uber became to taxis, what Amazon became to retail.

What was actually happening was a freight brokerage with a sophisticated technology layer, operating in a commodity industry with low switching costs, fungible capacity, and fragmented demand — being valued and operated as if it were a consumer technology platform with the unit economics of one.

The gap between what Convoy was and what Convoy was being valued as wasn’t a forecasting error. It was structural. It was built into the operating philosophy from the first board meeting. And by April 2022, the architecture of capital commitments around that philosophy had become impossible to dismantle in the time the market gave the company.

Four distortions operated here. They did not arrive simultaneously. They arrived in sequence — and the first one was on the board.

The Distortion Layer

First: Incentives — the operating philosophy as institutional doctrine

In 2018, Reid Hoffman — co-founder of LinkedIn, partner at Greylock, and one of Convoy’s first board members — published Blitzscaling: The Lightning-Fast Path to Building Massively Valuable Companies. The book formalized a philosophy Hoffman had refined through PayPal and LinkedIn. The core claim was direct: in the internet era, markets are built on winner-take-all outcomes, and the right strategy is to prioritize speed over efficiency in an environment of uncertainty. Funding growth before the revenue model is proven is not a bug. It is a feature. Capital is the rocket fuel that buys market position before competitors can respond.

Hoffman is explicit about the conditions under which blitzscaling works. The market has to have network effects. Switching costs have to be high. Economies of scale have to be real. Barriers to entry have to exist.

Freight brokerage has none of these. A carrier representative consults a dozen brokers in a day. Shippers compare rates across the market constantly. Capacity is fungible — the same truck can move a load for any broker. The industry has roughly sixteen thousand brokers and no meaningful barriers to entry beyond a bond and a phone. Leonard Sherman of Columbia Business School, asked about Convoy after the collapse, put the point in one sentence: freight brokerage is decidedly lacking in every condition blitzscaling requires.

But the philosophy was on the board, written by the partner who sat in the room, applied to an industry where the philosophy’s own author had specified it wouldn’t work. This is what incentives operating as an institutional force looks like. The reward structure didn’t just push Convoy’s executives toward certain decisions. It chose, in advance, which decisions would be visible and which would be invisible — and the table was set before any of the cognitive distortions engaged.

That’s the first distortion. The three that follow are where the philosophy stopped being an investment thesis and became the architecture of a company that could not be honestly re-examined from the inside.

Read more

The Situation

There’s a term that circulated in banking circles for years to describe a particular kind of technology solution. RFOP. Room Full Of People. The product worked — it just worked because of the humans operating it, not because of the technology being sold. The pitch was automation. The reality was sneakers.

Olive AI was the most expensive RFOP in healthcare technology history.

It didn’t start that way. In 2012, Sean Lane founded a company called CrossChx in Columbus, Ohio — a patient identity management platform built around fingerprint recognition. The original idea was legitimate: link patient medical records to a single verified identity to improve care, reduce costs, and prevent fraud. CrossChx raised tens of millions across several rounds and built a genuine if modest business in healthcare administration.

Then in 2017, with growth stalling and the original revenue model failing to scale, the company rebranded. CrossChx became Olive AI. The fingerprint business became an AI and RPA platform for healthcare automation. The pivot was not driven primarily by a technological breakthrough. It was driven by the recognition that AI commanded a fundamentally different class of investor attention and a fundamentally higher valuation multiple than identity management software. The narrative gap between what the product was and what it was being sold as wasn’t a mistake that emerged from optimism. It was architectural. It was built into the rebrand.

Olive raised roughly $902 million in venture capital, achieved a $4 billion valuation, and deployed its platform across more than 900 hospitals in over 40 US states. At its peak it employed around 1,200 people. It was the kind of company that made the cover of healthcare trade publications and the shortlist of every hospital system’s digital transformation agenda.

On October 31, 2023, Olive announced it was shutting down entirely. Waystar paid roughly $10 million for the clearinghouse and patient access units — what the company itself had described as the heart of its business. Roughly $902 million raised and a $4 billion valuation had resolved, two years later, into a $10 million sale of the core and a wind-down of the rest.

From $4 billion to immaterial in two years. The question worth asking isn’t what went wrong. It’s when the sequence of decisions that made this outcome nearly inevitable actually began.

The answer, it turns out, is 2017. The moment the narrative and the product formally separated.

What Actually Happened

The stated rationale at every stage was growth. More hospitals. More use cases. More capital deployed against a massive market opportunity. Lane was building the operating system for healthcare administration and the window was open. Move fast, capture the market, figure out the margins later.

What was actually happening beneath that rationale was a gap — between what Olive was selling and what Olive was delivering — that widened with every new contract signed and every new dollar raised.

Despite the AI branding, Olive’s platform was primarily built on brittle RPA scripts that failed to scale or integrate seamlessly with healthcare systems. RPA — robotic process automation — is a legitimate technology. It is also a considerably less sophisticated and less defensible product than artificial intelligence. The gap between what RPA actually does and what the phrase “healthcare AI” implies in a sales conversation is significant. Olive lived in that gap for years.

When the scripts failed — and they failed regularly in the complex, non-standardized environment of hospital IT systems — human workers fixed the errors behind the scenes. The automation wasn’t autonomous. It was a room full of people in sneakers, running between screens, manually completing what the pitch deck called AI-driven workflow automation. The product worked. Just not the way it was sold.

That gap was the original distortion. Every decision that followed was made in its shadow.

The Distortion Layer

Four distortions operated here, arriving in sequence and compounding with each new round of capital.

First: Incentives — the structure that rewarded the narrative over the product

The decision to sell AI when you have RPA isn’t primarily a technical decision. It was a commercial one — and by 2018 it had become a structural one. The AI label wasn’t just positioning. It was the business model. Hospital systems evaluating vendors weren’t comparing RPA businesses at $4 billion valuations. They were evaluating AI businesses. The label mattered for procurement decisions, for board presentations, for digital transformation narratives that hospital leadership needed to tell their own stakeholders.

The incentive to maintain the AI framing ran in both directions. Olive needed the label to command the valuation. The hospital systems needed the label to justify the investment internally. Every sales conversation that closed on that basis signed a contract that the product would struggle to fulfill. Every contract created an obligation that required more people — the sneaker brigade — to honor manually. Every new hire added to the cost structure that the revenue model couldn’t support.

The incentive to sell the roadmap rather than the product created a compounding operational problem that grew faster than the company’s ability to solve it. Nobody decided to build a fraudulent company. The incentive structure made selling the gap the path of least resistance at every stage — and organizations, like water, find that path without being directed toward it.

The first distortion is the one you can see from outside. The three that follow are the ones that made it impossible to stop — and they are the reason a $4 billion company sold its core for $10 million. They are below, for subscribers.

Read more

BAD CALL BRIEF · FREE ISSUE

THE SITUATION

Early in my career I worked for one of the smartest leaders I’ve encountered. She ran a senior team that was genuinely talented — experienced, opinionated, and not shy about saying so. She could have walked into any meeting, laid out her thinking, and had the room aligned within twenty minutes. She almost never did.

Instead she did something I’ve never forgotten and have recommended ever since. Before any significant decision, she would have a team member frame the issue in neutral terms — no position, no recommendation, just the situation and the relevant facts. Then she would ask the most junior person in the room to speak first. She worked her way up the seniority ladder deliberately and methodically. She was almost always the last person to talk.

The quality of thinking in those meetings was unlike anything I experienced before or (too often) since. Ideas emerged that the senior people hadn’t considered. Assumptions got challenged before they calcified into conclusions. The most experienced voices in the room — including hers — were regularly surprised by what they heard.

She had no formal name for what she was doing. Behavioral economists call it counteracting the availability cascade. What she understood intuitively was something most leaders never fully grasp: the order of conversation is itself a decision. And most of the time that decision gets made by default, in favor of the heaviest voice in the room, before the meeting formally begins.

WHAT ACTUALLY HAPPENS

The stated purpose of most group decision processes is to surface the best available thinking from the people in the room. Diverse perspectives, collective intelligence, the wisdom of a well-assembled team. That’s the intent. It’s a reasonable intent. It almost never fully delivers.

What actually happens is that group decision processes are systematically shaped by a set of behavioral dynamics that operate below the level of anyone’s awareness — including the most experienced people in the room. These dynamics don’t announce themselves. They feel like normal conversation. They produce conclusions that feel like consensus. And they consistently deliver outcomes that reflect the heaviest voice rather than the best available thinking.

Four dynamics operate here. They don’t arrive randomly. They compound.

THE DISTORTION LAYER

Here is the structure of what goes wrong, in the order it goes wrong.

First: Anchoring and the availability cascade

The first substantive contribution to any group discussion carries disproportionate weight. Not because it’s the best idea. Because it anchors everything that follows. As the discussion proceeds, a second dynamic activates: the availability cascade. The anchored position starts getting referenced, built upon, and reinforced. Each repetition makes it more socially credible — not because the underlying argument has strengthened but because the idea has been in circulation longer.

The two dynamics together mean that the range of thinking available to the group narrows steadily from the moment the first substantive voice speaks. This dynamic intensifies sharply when the first voice belongs to the most senior person in the room.

Second: Confirmation bias — and its most dangerous form

Once an early position has weight in the room, each participant begins — unconsciously — to weight incoming information in a way that confirms rather than challenges the emerging direction. Information that supports the leading view gets absorbed easily. Information that challenges it faces a higher burden of proof.

But the most commonly cited version of this dynamic — that we ignore or dismiss information that clearly conflicts with our position — is actually the less dangerous one. Obvious contradictions are at least visible. They can be named. They create friction that sometimes forces honest examination.

The more insidious form operates on neutral information. Data that carries no inherent directional signal. A market trend that could mean several things. An ambiguous result. A piece of feedback that is genuinely open to interpretation. In a room where the anchor has already formed, that neutral information doesn’t stay neutral. It gets processed through the frame of the existing position and emerges as confirmation — not because anyone is being dishonest, but because the brain is completing a pattern it has already started drawing.

This is where the damage compounds quietly. The group isn’t disregarding contradictory evidence — it’s actively building a case from material that shouldn’t be in the case at all. Every ambiguous data point becomes one more brick. The position doesn’t just survive the new information. It grows stronger on information that should have been weightless. By the time a genuinely contradictory signal arrives, it’s entering a room that has spent an hour constructing a wall out of neutral materials it mistook for mortar.

Third: Noise — the silence that looks like agreement

Around any significant decision, there are people in the room who have a different read. Most of them say nothing. Not because they lack conviction. Because the social geometry of the room makes speaking costly in ways that staying quiet isn’t. Each person who disagrees assumes that others must have additional information justifying the emerging conclusion. They defer to a consensus that isn’t actually there — a phenomenon behavioral economists call pluralistic ignorance.

Fourth: Sunk cost closing the exit

Even when the dynamics above produce a clearly flawed direction, correction becomes exponentially harder as time passes. Once resources have been committed — capital, time, reputation, public statements — the cost of changing course stops being purely analytical and becomes personal. The exit that was available early has quietly closed.

WHAT A CLEAN PROCESS LOOKS LIKE

My former boss was doing something structurally important — not just culturally. She wasn’t asking people to be braver or more honest. She was removing the conditions that made honesty costly before the meeting began. Four structural interventions that work:

Reverse the speaking order. Always. On any decision that matters, the most junior voice goes first and the most senior goes last. This isn’t about being democratic. It’s about preventing the availability cascade from forming before the discussion has a chance to develop.

Separate information from advocacy. The person who frames the issue should have no position on the outcome. Framing and advocating are different cognitive tasks and mixing them poisons the information environment before the discussion begins.

Make the private calculations visible. Before any significant commitment, every member of the decision-making group writes down their honest assessment independently — before hearing anyone else’s. Those assessments get shared simultaneously. The divergence between them is the most important data point in the room.

Build a designated challenge role with explicit authority. Not a devil’s advocate who performs skepticism and gets noted. A specific person, assigned before the meeting, whose job is to find the strongest case against the emerging direction and present it with the same rigor as the case for it. Rotate the role — the same person playing it every time becomes the designated pessimist and gets discounted.

CARRYOVER

The leader I worked for early in my career didn’t have a framework. She had an instinct — that the heaviest voice in the room was also the most dangerous one if it arrived too early — and she built a simple structural response to it that she applied consistently.

The instinct was right. The structure was the point.

Your next significant group decision will be shaped by dynamics that are already operating before anyone sits down. None of that is inevitable. All of it is structural. Which means all of it can be interrupted — but only before the heaviest voice speaks, not after.

The question worth asking before your next consequential meeting isn’t who should be in the room.

It’s who speaks first.

This is a Bad Call Brief — a free condensed issue.

The full paid issues deliver the complete distortion layer anatomy, clean decision analysis,

and a Carryover you can bring into the room on Monday.

If this is the lens you’ve been looking for, you’ll know within two issues.

Bad Call

Bad Call publishes every other Tuesday. One decision sequence, dissected without narrative cover.

Full analysis for paid subscribers at badcall.ai

— RC

BAD CALL

Decision autopsies for people who know better.

Most organizations don’t fail because of one catastrophic mistake made by a fool.

They fail because of a sequence of reasonable-seeming decisions made by smart, experienced, well-resourced people — each one individually defensible, each one making the next harder to question, until the outcome that seemed unthinkable becomes the one that was nearly inevitable.

That pattern has a name. It has a structure. And once you can see it clearly, you see it everywhere.

What Bad Call Is

Bad Call is a decision intelligence brief that reconstructs those sequences. Not to assign blame. Not to celebrate hindsight. To show precisely where the thinking went wrong, what was driving it beneath the stated rationale, and what a cleaner version of the same decision would have looked like.

Each issue takes one organization — a company, a board, a leadership team, an investment group — and runs it through a structured analytical framework built around the forces that distort consequential decisions. The framework isn’t theoretical. It’s been built and tested against some of the most expensive decision failures of the last thirty years, and refined through decades of sitting inside the kinds of decisions this publication dissects.

The structure of every issue is fixed and deliberate: the situation as it appeared at the time, what was actually driving it, where the distortion entered and why, what a clean version would have looked like, and one idea the reader can carry into the room on Monday.

No long introductions. No storytelling for its own sake. No frameworks you already know dressed up as new insight. The structure is the product.

Who It’s For

Bad Call is written for people who make consequential calls for a living. Founders. Board members. Senior operators. Advisors. People who can’t afford to let bias, noise, accumulation, and misaligned incentives quietly run their judgment off the road.

Bad Call doesn’t work for everyone. It’s not supposed to.

If you’re the right reader you’ll know within two issues.

Who Writes It

Bad Call is written by RC — an operator and advisor with decades of experience inside the kinds of organizations and decisions this publication dissects. The framework came from watching smart people make predictable mistakes in real time, trying to name what was happening before it compounded, and occasionally failing to make the naming stick.

The analysis comes from someone who has been in the room. The point of view is direct. The writing is attributed.

What the Framework Looks Like in Practice

In banking circles, there’s a term for a particular kind of technology solution. RFOP. Rooms Full Of People. The product works — it just works because of the humans operating it, not because of the technology being sold. The pitch is automation. The reality is sneakers.

Olive AI raised $850 million, reached a $4 billion valuation, and deployed across 900 hospitals. In October 2023 it shut down completely. The gap between what was being sold and what was being delivered had been visible in the renewal data for years. Every function read the same signal differently. Nobody had a framework that forced a common answer.

Nobody decided to build a fraudulent company. The incentive structure made selling the gap the path of least resistance at every stage — and organizations, like water, find that path without being directed toward it.

Three cognitive biases compounded in sequence, each one making the last harder to reverse. By the time the data became unambiguous, the accumulation of misdirected commitments had made honest correction nearly impossible.

That’s what Bad Call does. Not the headline. The mechanism. Not what happened. Why it was nearly inevitable — and what a cleaner version would have looked like.

On Price

Bad Call is a paid publication, priced to reflect the value of what it delivers rather than the volume of what it produces. One issue per week. No filler, no padding, no content for content’s sake.

A 30-day satisfaction guarantee applies to all new subscriptions. If it isn’t what you expected, contact RC directly within 30 days.

Founding memberships — In The Room — are available in limited quantity for those who want permanent access and a direct line to the thinking behind the analysis. There are 25. That number won’t change.

Bad Call publishes every other Tuesday.

Because it always seemed fine at the time.

— RC