Failure Examples: How Convoy Scaled on a Data Mirage
Convoy was not a fragile company. By 2022 it had raised roughly $900 million, carried a reported $3.8 billion valuation, and counted Jeff Bezos, Bill Gates, Marc Benioff and Reid Hoffman among its backers. It had built genuine technology, automating the match between shippers and truckers at a scale most of the freight industry had not attempted, and it was frequently described as one of the most credible attempts to modernise a vast and fragmented market. This was not a company operating at the margins of plausibility. It was, by most external measures, a category leader.
In October 2023, it cancelled every shipment on its board and shut down within a matter of days. Staff were laid off with no severance and told their stock options were worthless. Two weeks later, Flexport acquired what remained of the technology for $16 million, a fraction of a percent of the company's peak valuation.
The CEO's explanation pointed outward, to "a massive freight recession and a contraction in the capital markets." Both were real, and both mattered. But a market downturn is not, on its own, a full account of why a business fails. Downturns are a feature of freight, not a surprise, and the more useful question is why Convoy was so exposed to one.
That question is this: if the demand Convoy scaled on was real, and the volumes moving across its platform were genuine, why could the company not survive that demand returning to normal?
The answer sits in the gap between two things that are easy to conflate. There is demand, and there is durable, monetisable demand at a margin that works. A demand mirage is demand that is real but borrowed from temporary conditions, and it clears when those conditions revert. Convoy had a great deal of the first. What it needed, and what the freight boom and the cheap-capital era obscured for several years, was the second. This is a strategic review of how a company can scale convincingly on a demand signal that looks like proof of a business and turns out to be proof of a cycle.
What Was Convoy Actually Building?
Convoy was founded in 2015 by Dan Lewis and Grant Goodale, both former Amazon employees, to fix a market that looked ready for it. Freight brokerage, the business of connecting companies that need goods moved with the truckers who move them, was enormous, essential and strikingly analogue. Much of it still ran on phone calls, spreadsheets and personal relationships between brokers and carriers. Loads went unfilled, trucks ran empty on return journeys, and margins disappeared into a chain of intermediaries. The inefficiency was real, and the opportunity to remove it was easy to describe.
Convoy's answer was a digital freight network. Rather than a broker working the phones, the platform used software to match a shipper's load with an available carrier automatically, pricing the job, handling the paperwork and aiming to keep trucks full on both legs of a trip. The company reported automating the large majority of its load matching, a level of automation the traditional brokerage model could not approach. The pitch was often shortened to "Uber for trucking," and while that undersold the complexity, it captured the ambition: take a fragmented, manual market and run it through a single, intelligent layer.
What matters for this review is the shape of the business underneath the technology. Convoy was building a two-sided marketplace, and a two-sided marketplace only works when both sides show up in the right proportion. It needed enough shippers to attract carriers, and enough carriers to serve shippers, and it needed to hold both while taking a margin thin enough to stay competitive with the incumbents it was displacing. Brokerage is not a high-margin business. Brokers have long operated on gross margins in the region of 13 to 15 percent, the slice they keep between what the shipper pays and what the carrier is paid.
That is the structure to hold in mind. Convoy was not selling a premium product into a captive market. It was intermediating an enormous flow of freight at a slim margin, and its entire model depended on that flow being large, steady and dense enough to make the economics work. When freight was abundant, the machine looked impressive. The question the next section takes up is why that abundance was so easy to mistake for something more durable than it was.
Why Did the Demand Look So Convincing?
For roughly two years, everything Convoy needed to be true appeared to be true. The pandemic reshaped how people spent, and it did so in Convoy's favour. Locked-down households redirected money from services to goods, government stimulus added spending power, and e-commerce volumes surged. All of that had to be moved, and the freight market ran hot as a result.
The pricing tells the story plainly. Spot rates, the price to move a load booked on short notice, climbed through 2020 and 2021 until the national truckload rate sat above $3.00 per mile by the end of 2021. Capacity was tight, shippers were competing to secure trucks, and any platform sitting between the two had more volume flowing through it than it could easily handle. For a business whose economics depend on density of freight, this was close to ideal conditions.
Seen from inside the company, and from the outside by investors, the signals were emphatic. Volumes were rising, the platform was busy, the automation was being tested against real freight at real scale, and the growth curves pointed in one direction. Each of those is exactly what product-market fit is supposed to look like. A marketplace that is filling loads, adding shippers and keeping carriers active is, on the face of it, a marketplace that works.
The difficulty is that a rising tide flatters every boat in the harbour, and it is very hard, in the moment, to tell how much of your progress is you and how much is the water. Convoy's growth was genuine in the sense that the loads were real and the volume was real. What was not visible in the numbers was the source of that demand. A surge driven by stimulus cheques, a one-off shift from services to goods, and a temporary capacity squeeze is not the same as a permanent increase in the structural need for the product, even though both produce the same encouraging chart.
This is the heart of how a demand mirage forms. The demand is not fake. It is simply borrowed from conditions that will not last, and it arrives looking identical to the durable kind. Distinguishing the two while the boom is still running is the genuinely hard part, and it is where the next section turns.
What Was the Mirage?
The mirage was not that demand existed. It was that three separate tailwinds were arriving at once, all pushing in the same direction, and the combined effect was easy to read as a single durable signal about the business. Pulled apart, each one was temporary.
The freight itself was cyclical. The volume moving across the platform was inflated by pandemic spending patterns that were always going to normalise. Goods consumption pulled forward during lockdowns would eventually revert toward services, and stimulus-driven demand would fade as the stimulus did. The loads were real, but the level was borrowed from a moment.
The pricing was inflated. Convoy was intermediating that freight at spot rates near historic highs. A thin percentage margin looks much healthier when the underlying rate is above $3.00 per mile than when it falls back toward $2.00. The same volume of business generates less revenue, and the same margin percentage returns fewer dollars, the moment prices normalise.
The capital was cheap. A long era of low interest rates made it possible to fund growth ahead of profitability, to subsidise both sides of the marketplace in pursuit of scale, and to treat the question of durable unit economics as something that could be solved later, once scale had been reached.
Individually, each of these is a known feature of markets. Freight is cyclical, spot rates move, and capital conditions loosen and tighten. The mirage was the way they overlapped. Together they produced a period in which volume was high, revenue per load was high, and the cost of funding the gap was low, all at the same time. That combination made the model look not just viable but proven.
The tell, in hindsight, is that all three tailwinds were external to the product. None of them was evidence that Convoy had solved the hard structural problem of running a two-sided freight marketplace at a profit. They were evidence that the environment was, for a while, unusually forgiving. A business can grow impressively on a forgiving environment, and the growth is not an illusion. What is illusory is the inference that the growth proves the environment no longer matters.
This is what separates a demand mirage from ordinary demand. Ordinary demand persists when conditions return to normal. A mirage clears. The critical question a company in Convoy's position has to ask is not "is demand strong," because in a boom the answer is always yes. It is "which part of this demand survives if the cycle turns, the prices fall and the capital gets expensive," and that is a far less comfortable question to sit with while the charts are still climbing.
Why Are Brokerage Economics So Unforgiving?
A demand mirage is more dangerous in some businesses than others, and freight brokerage is close to the worst place to encounter one. The reason is the margin. Brokers typically keep somewhere in the region of 13 to 15 percent of the shipper's spend, paying the rest to the carrier. That slim spread has to cover the technology, the staff, the customer acquisition, the support and every other cost of running the business, and it leaves very little room for error.
Convoy's model added a second pressure on top of the thin margin. A two-sided marketplace has to attract and hold both sides, and in a competitive market that usually means subsidising them. Shippers are offered keen pricing to move their freight onto the platform; carriers are offered attractive rates and steady work to keep their trucks in the network. Every dollar spent winning one side is a pound not falling to the bottom line, and in the land-grab phase of building a marketplace, that spending is treated as an investment in eventual scale rather than a cost to be recovered now.
The assumption underneath that approach is that scale eventually fixes the economics. Get large enough, the logic runs, and density improves, matching gets more efficient, empty miles fall, and the margin per load widens until the business becomes profitable through sheer volume. It is a coherent theory, and in some marketplaces it holds. The trouble is that in a thin-margin, cyclical, commoditised business like freight, scale amplifies whatever the underlying economics happen to be. If each load is marginally profitable, more loads help. If each load is subsidised to win share, more loads simply mean more subsidy, and growth makes the hole larger rather than smaller.
This is the structural trap. During the boom, high rates and abundant freight made the per-load economics look workable, which appeared to validate the scale-first strategy. The volume was there, the rates were there, and the path to profitability seemed to be a matter of continuing to grow into a favourable market. What that reading missed is that the favourable market was doing the heavy lifting. The margin was not being manufactured by the model. It was being supplied by the cycle, and a model that depends on the cycle to make its unit economics work has not yet proven it has unit economics at all.
A business with fat margins can absorb a downturn by giving some of them back. A business intermediating a commodity at thirteen cents on the dollar, while subsidising both sides to hold share, has almost nothing to give back before it is losing money on every load it moves. That is the position Convoy was scaling into, and it is why the turn, when it came, was so unforgiving.
What Happened When the Cycle Turned?
The turn began in 2022, and it came from both directions at once. On the demand side, the pandemic spending pattern reversed. Households shifted back toward services, the inventory that retailers had over-ordered during the shortages piled up in warehouses, and the volume of goods needing to move fell away. Analysts date the start of the freight recession to around April 2022, and it went on to become one of the longest downturns the sector had seen.
On the supply side, the boom had done lasting damage by pulling in capacity that did not leave when demand did. High rates during 2020 and 2021 had encouraged huge numbers of drivers to obtain their own operating authority and enter the market. The number of for-hire carriers in the United States rose from roughly 241,000 in June 2020 to more than 475,000 by July 2023, an increase of about 96 percent. Twice as many carriers were now competing for less freight.
The effect on pricing was exactly what that imbalance implies. Spot rates fell steadily from their late-2021 peak, giving back the pandemic gains and then some, with the spot market dropping more than 20 percent while operating costs stayed high. For a broker, this is the worst of both worlds. There is less freight to intermediate, and each load that does move is worth less, because the rate the margin is calculated on has shrunk.
For a business with fat margins, this would have been a painful but survivable squeeze. For Convoy, it landed directly on the vulnerability described in the previous section. The thin brokerage margin, already stretched by subsidising both sides of the marketplace, had almost no cushion to absorb a fall of this size. The volume that had made the model look like it was working drained away, the rates that had made the per-load economics look healthy collapsed, and the scale the company had built now worked against it, because every load in a loss-making network adds to the loss.
The mirage did not fail gradually. It cleared. The demand that had appeared to prove the business was revealed as demand that had been on loan from the cycle, and when the cycle called it back, what remained was a low-margin, capital-hungry operation moving a shrinking volume of freight at falling prices. The next section turns to why the second pillar holding the company up, cheap capital, gave way at precisely the same moment.
Why Did Cheap Capital Hide the Problem for So Long?
Convoy raised money across an era in which capital was unusually cheap and unusually patient. Between 2018 and 2022 it moved from a $1 billion valuation to a reported $3.8 billion, gathering roughly $900 million along the way from investors who were, for most of that period, rewarding growth far more than profitability. In a low-interest-rate environment, a company burning money to build scale is not an anomaly. It is the strategy the market is actively funding.
That funding did something subtle and important. It removed the pressure that would otherwise have forced the question of unit economics to the surface. A business that has to fund itself from its own margins finds out quickly whether those margins exist. A business that can raise the next round to cover the gap can defer that discovery, sometimes for years. The losses on each subsidised load were real throughout, but they were survivable as long as fresh capital kept arriving to replace them, and as long as the growth story justified the next cheque.
So the same cheap capital that funded Convoy's expansion also concealed the flaw in it. As long as money was available on favourable terms, the company did not need its marketplace to be profitable yet. It needed the boom to continue and the funding to continue, and for a while both did. The structural problem was not hidden because anyone was hiding it. It was hidden because the conditions that would expose it, expensive money and a weak market, were both absent at the same time.
Then they arrived together. Central banks raised rates sharply through 2022 and into 2023, and the appetite for funding unprofitable late-stage companies contracted with them. Convoy's own CEO named this directly, pointing to "a contraction in the capital markets" alongside the freight recession. The subsidy that had kept the model afloat was withdrawn at the very moment the freight cycle stopped supplying the margins. The company needed either profitability or patient capital to survive the downturn, and monetary tightening removed the second while the freight recession removed any near-term route to the first.
This is the part of the pattern that recurs well beyond freight. Cheap capital does not just fund growth. It funds the postponement of hard questions, and it allows a business to look proven for as long as the money and the market both cooperate. When they stop cooperating together, the postponed question arrives all at once, and usually at the worst possible time. The next section asks whether, given all of this was visible in principle, Convoy could have seen it coming.
Could Convoy Have Seen It Coming?
It would be too easy to say the collapse was obvious, and it was not. Freight is cyclical, and everyone in the industry knows it, but the timing, depth and length of any given downturn are genuinely hard to call. Convoy was also not run by naive people. The founders came from Amazon, the backers were among the most sophisticated investors in the world, and the strategy of scaling first and monetising later was, at the time, the strategy the entire venture market was endorsing. Judged against the prevailing wisdom of its era, Convoy was executing that playbook well.
The more honest framing is not that the company missed an obvious signal, but that it was answering the wrong question with great success. The question it was optimising for was how to grow, how to add shippers, hold carriers, automate more of the match and take more volume, and it made real progress on all of it. The question that would have changed its exposure was different and less comfortable: what does this business look like if freight volumes normalise, spot rates fall back toward $2.00 and the next funding round is either expensive or unavailable.
That is a stress test rather than a forecast, and the distinction matters. Convoy did not need to predict that the downturn would begin in April 2022, which nobody could reliably do. It needed to know what the model would do under a downturn of ordinary severity, because in freight a downturn of ordinary severity is not a tail risk. It is a certainty on a timescale that matters. A business that intermediates a cyclical commodity at a thin margin has to assume the cycle will turn within its planning horizon, and design for that, rather than treating the good years as the baseline.
Read that way, the failure was less about foresight and more about framing. Reading a market's direction is about where things are heading, and Convoy read that reasonably well: the freight market was digitising, and it still is. Stress-testing durability is about what survives when the favourable conditions reverse, and that is the analysis the boom made it easy to skip. The two are often confused, because in good times they give the same answer. It is only when conditions turn that the difference between them becomes the difference between a company that bends and one that shuts down in a week.
What Does This Reveal About Reading Demand Signals?
Convoy's story generalises well beyond freight, because the mistake at the centre of it is one almost any company can make when a market is running in its favour. Demand is the signal every business is trained to look for, and a strong demand signal is treated, almost reflexively, as confirmation that the strategy is right. The Convoy case is a reminder that demand is not one thing. It comes in kinds, and the kinds behave very differently when conditions change.
The distinction worth carrying forward is between demand that is structural and demand that is borrowed. Structural demand reflects a durable need that persists across cycles: it may soften in a downturn, but it does not disappear, because the underlying reason for it remains. Borrowed demand is real while it lasts but rests on conditions that will revert, whether that is a stimulus programme, a temporary shift in consumer behaviour, a pricing spike or a period of cheap money. On a growth chart the two are indistinguishable. In a downturn they could not be more different.
Separating them before scaling is uncomfortable work, because it means interrogating your best numbers at the moment they look most encouraging. A few questions do most of that work:
How much of this demand would survive if conditions returned to their long-run average? Not the optimistic case or the pessimistic case, but the ordinary one.
Is the current pricing normal, or is it elevated by a temporary imbalance? A margin that depends on peak prices is a margin you do not yet reliably have.
What is actually driving the growth, the product or the environment? If the tailwinds were removed, would the curve still point up.
Do the unit economics work at trough conditions, not just at peak? A model that is only profitable at the top of the cycle is not yet profitable.
Is the capital funding the business or funding the wait for the business to work? The two feel identical until the capital gets expensive.
None of these questions can be answered precisely, and none of them will stop a favourable market from being genuinely useful. The point is not to distrust growth, but to know what kind of growth you are looking at, so that the decisions you build on top of it, how fast to scale, how much to subsidise, how much runway to hold, are calibrated to demand that will last rather than demand that is on loan. Scaling aggressively on borrowed demand is not ambition. It is building the business on the assumption that the weather will not change.
What Was the Real Lesson from Convoy's Collapse?
Convoy is not a story about a bad company or a foolish bet. It is a story about how convincingly a favourable environment can imitate a working business, and how hard it is to tell the two apart while the environment is still cooperating. The company built real technology, attracted serious capital, moved enormous volumes of freight and was, by the external measures most people use, a success. What it did not establish, because the boom made it unnecessary to establish, was that the business worked independently of the conditions holding it up.
The demand was real, and that is precisely what made it dangerous. Fake demand is easy to dismiss. Real demand, borrowed from a cycle and arriving in record volumes, is almost impossible to argue with in the moment, because every number confirms the story you want to believe. The mirage was not that the freight existed. It was the inference that freight at boom-time volumes and boom-time prices, intermediated at a thin margin and funded by cheap capital, amounted to proof of a durable enterprise. Three temporary tailwinds, mistaken for one permanent fact.
The Gap Between Riding a Market and Building a Business
The deeper lesson sits in the distance between riding a market and building a business that can survive the market reverting. Convoy rode its market skilfully. When the market turned, as a cyclical freight market always eventually does, there was less business underneath the growth than the growth implied. The freight recession and the capital-markets contraction that the CEO cited were the triggers, but a trigger only fells something already exposed. The exposure was built during the good years, one aggressively scaled, subsidised, loss-making load at a time.
For any company reading its own demand in a strong market, whether that means a pandemic tailwind, a category in fashion, a pricing spike or an era of cheap money, the question Convoy invites is not whether demand is strong. In a boom it always is. The question is which part of that demand is structural and would survive a return to normal conditions, and whether the business has been built to run on that durable core rather than on the borrowed surge sitting on top of it. Being right that a market is growing is worth a great deal. Being right about how much of today's growth will still be there when the weather changes is what determines whether you are still standing to benefit from it.
Frequently Asked Questions
Why did Convoy shut down?
Convoy shut down in October 2023 after a prolonged freight recession and a contraction in the capital markets, the two factors its CEO named directly. The deeper reason is that its business model had little cushion to absorb either. Convoy operated a two-sided digital freight marketplace on the thin margins typical of brokerage, and it had scaled during a pandemic-era boom that lifted both freight volumes and shipping rates. When volumes and rates fell back to normal and cheap funding dried up at the same time, the model could not cover its costs, and the company was unable to find a buyer before running out of money.
Was Convoy profitable?
No. Like many venture-backed companies of its era, Convoy prioritised growth and scale over profitability, funding the gap with successive rounds of investment. It subsidised both sides of its marketplace to build volume, on the assumption that scale would eventually make the economics work. That assumption was never tested against a normal market, because the company was reliant on continued cheap capital and a favourable freight cycle to sustain itself.
What is a demand mirage in business?
A demand mirage is demand that is real but borrowed from temporary conditions, so it disappears when those conditions revert. The demand is genuine while it lasts, which is what makes it convincing, but it is driven by factors external to the product, such as a stimulus programme, a temporary shift in consumer behaviour, a pricing spike or a period of cheap money, rather than by a durable, structural need. On a growth chart it looks identical to lasting demand. The difference only becomes visible when conditions return to normal.
How much money did Convoy raise before collapsing?
Convoy raised roughly $900 million over its lifetime, reaching a $1 billion valuation in 2018 and a reported peak of around $3.8 billion in 2022, with backers including Jeff Bezos, Bill Gates, Marc Benioff and Reid Hoffman. When it wound down in 2023, Flexport acquired its remaining technology and assets for about $16 million, a small fraction of that peak valuation.
What is digital freight brokerage and why is it hard to make profitable?
Digital freight brokerage uses software to match companies that need goods transported with the carriers who move them, automating work that traditional brokers do by phone and email. It is hard to make profitable because brokerage is a low-margin business, with brokers typically keeping around 13 to 15 percent of what the shipper pays. Building a two-sided marketplace usually means subsidising both shippers and carriers to reach scale, which compresses those margins further, and freight is a cyclical, commoditised market, so revenues rise and fall with rates the broker does not control.
What can companies learn from Convoy's failure about scaling on demand?
The main lesson is to distinguish structural demand from borrowed demand before scaling on it. Structural demand persists when conditions return to normal; borrowed demand rests on temporary tailwinds and clears when they fade. Before committing to aggressive growth, it is worth stress-testing whether the demand and the unit economics would still hold at average or trough conditions, rather than treating a boom as the baseline. Being right about a market's direction is not the same as building a business that survives the market reverting to normal.
How Metheus Can Help
When companies scale into a growing market, we help them test whether the demand they are building on is durable or borrowed from conditions that will not last. Being busy in a boom is not the same as having a business that survives the cycle turning, and the difference is rarely visible while the numbers are still climbing. We work across market expansion planning and growth and scale strategy, looking closely at demand quality, unit economics and how a model behaves under normal rather than peak conditions, so that growth decisions are calibrated to demand that lasts. Our role is to help you separate the structural core from the cyclical surge before you commit capital and headcount to it.