Reversion to the mean: the real long COVID
The pandemic may be over, but so is tech exuberance.
In March 2020, we were warned that a demand shock would crush the economy, and tech would not be spared. Sequoia’s famous doomsday memo predicted a tech recession:
Private financings could soften significantly…what would you do if fundraising on attractive terms proves difficult in 2020 and 2021?
With softening sales, you might find that your customer lifetime values have declined, in turn suggesting the need to rein in customer acquisition spending to maintain consistent returns on marketing spending.
But the prognosis was precisely wrong.
Most internet and software businesses would have been better off following the opposite strategy during the pandemic. Fundraising became easier than ever, and demand skyrocketed for tech substitutes to real-world products. Getting predictions wrong by two years means massive missed opportunities.
The pandemic brought unprecedented volatility to the tech industry – but with two years of hindsight, what has changed?
2020-2021: tech exponentialism
Months after the “imminent” tech crash in March 2020, the demand shock never came for most categories. In fact, indicators of tech acceleration surfaced by early summer: exponential e-commerce adoption, shortening software sales cycles, SaaS vitamins becoming painkillers.
Instead of a collapse, summer 2020 was the start of a historic tech bull run.
Two factors fueled tech growth:
Falling interest rates: Rates dropped to near-zero, making a dollar of profit in 2030 nearly as appealing as a dollar in 2021. Companies with profits far in the future – think high-growth and high-risk companies – got undue credit for those future dollars.
Rising tech demand: COVID raised the demand floor for online alternatives to physical businesses. E-commerce doubled. Food delivery tripled. Video conferencing quintupled. This created a credible argument for tech domination: DoorDash would crush restaurants, Zoom would render offices obsolete, Peloton would eliminate gyms.
The consensus view by summer of 2020 was that the pandemic would dramatically accelerate existing tech trends. Wishful thinking reigned king: if two months of COVID doubled tech trends, what could two years do?
There were three potential scenarios for the future of technological growth:
Exponential: Pandemic era acceleration represented the new baseline growth rate. Instead of growing 10% annually, e-commerce and other categories would have a new sustained level of high growth.
Step function: COVID established a “new normal” for tech demand, a step function above pre-pandemic levels. Old growth rates would continue post-COVID, just from a higher base established in 2020. This was a popular view among VCs and tech pundits.
Mean-reversion: Acceleration through 2020 and 2021 was an aberration, and tech market growth will revert to the pre-COVID growth trendline.
In e-commerce, for example, these three stories you could tell about the future in summer 2020 would look something like this:
Public market prices started to reflect a combination of exponential and step function tech growth. This meant that all tech companies were undervalued by a wide margin pre-COVID. At-home fitness, security, e-commerce, video conferencing – if software was going to eat the world overnight, the respective stocks needed to reprice as such.
Tech growth engines have demonstrated a unique property in the last decade: durable growth rates. Many big tech and software companies – Facebook, Google, Amazon, Salesforce, Atlassian, PayPal – compounded at 20%+ for more than a decade.
When tech companies see growth acceleration, it is often permanent – look at Microsoft’s revenue from 2012-2017 (5.5% CAGR) versus 2017-2022 (15.6% CAGR). Once its growth engine proved it could compound faster, the growth rate had a new floor.
Did COVID establish a new baseline growth rate for tech?
2022: tech reverts to the mean
Mean-reversion seemed to be the least popular story to tell – who wants to rain on the tech parade? But the last six months clarified that much of the tech acceleration was transient.
Tech massively over-earned in 2021 – generating growth and profits at a temporarily high rate – driven by substitute demand from the pre-COVID economy. Companies with accelerated temporary growth got credit for a permanent acceleration.
But investors recently realized that underwriting growth permanence from COVID acceleration was wrong. The ensuing correction was harsh, but note that many high-growth companies are trading very close to their pre-COVID prices:
Under the surface, many pandemic era trends assumed to be permanent turned out to be temporary, and many seemingly temporary trends may be permanent. E-commerce seemed like the most inevitably permanent COVID trend. But in fact, e-commerce penetration has reverted to the trend line over the past few quarters:
Many other tech categories follow a similar path of mean-reversion. I wrote an article in April 2020, When Tailwinds Vanish, which was seemingly disproven just a few months later: tech had a banner year in 2020. But the last six months proved the opposite: tech’s over-earning during the pandemic means that tailwind decay will be even harsher now.
Bessemer’s Emerging Cloud Index (EMCLOUD), which is probably the best proxy index for public high-growth tech, is trading within 10% of March 2020 levels. So at a crude aggregate pricing level, COVID was net neutral for tech. But even compared to the major indices, high-growth technology companies fared neutrally relative to the old guard:
I have no doubt tech will continue to compound, but the laws of physics still apply. Tech and non-tech products are largely substitutes: fitness bikes versus gyms, gaming versus sports, takeout versus restaurant dining.
The nation’s COVID recovery was zero-sum with tech exuberance.
The mid-2020s: stagflation is a lingering symptom
The tech exponentialism we saw during the pandemic was not only driven by a demand spike – it was exacerbated by low interest rates and money printing.
But it turns out you can’t print $10T without anyone noticing.
Macro factors like inflation didn't matter in tech for the 1990s through the 2010s. They were offset by total internet and software growth rates in the high 20s and 30s. Inflation was always below 4% – as a percentage of growth, it was an irrelevant headwind.
But many tech category tailwinds are flipping to headwinds as we revert to the mean – when you’re growing sub-30% and inflation is 10% a year, inflation becomes a far more consequential growth headwind.
Inflation sneaks in indirectly – it’s not simply the cost of gas and groceries. There are several hidden inflationary factors for startups today:
Labor cost: Rising salary expectations are the most obvious and direct impact to startups. When inflation became common knowledge in the 1980s, unions pre-negotiated annual wage increases – this could happen in tech. When salaries rise, you have to burn more capital to build the same products. While labor cost inflation will slowly restabilize, expect regular wage hikes for the next several years.
Productivity: In a more competitive tech talent market, employees expect not just higher compensation, but less work – a double whammy. The four-day work week is already gaining popularity. And many employers are acquiescing for fear of attrition.
Equity burn: This is a painful inflationary backdrop to every startup, and under-appreciated even at the most sophisticated companies. The option pool is each startup’s internal Federal Reserve, setting an “inflation rate” for your share price. As competition for talent increases, so does your option pool inflation rate. Every year, your share price is offset by option pool expansion, which can range from 2-5% for growth-stage companies. Public companies are offenders too, with stock-based compensation constituting 10%+ of revenue for many tech companies.
DCF impact: If your cash flows are far in the future (high-growth and sub-scale businesses), USD inflation and rising rates hurt you the most. As interest rates rise to match inflation, your valuation gets less credit for profit generated in ten years from now. If your average dollar of profit comes in thirty years from now, your stock would be down ⅔ from mid-2020 even if nothing changed about the odds of that dollar being earned.
Labor squeezes and compensation inflation will take years to stabilize. Tailwind reversion plus inflation will usher in an age of tech stagflation. Startups need to demonstrate exceptional growth and efficiency to overcome these inflationary headwinds.
How should we navigate tech stagflation?
Reversion to the mean takes time. But the easy wins are over, for both VCs and founders. In a way, the Sequoia memo from March 2020 is useful now, but because of the harsh return to normalcy, not because of a sudden demand shock.
Look at anything that grew massively in the past 2 years with a cautious eye: crypto gold rushes, venture capital expansion, diligence-free fundraises. I’m not saying these trends are not durable – a subset of pandemic-era shifts certainly will be. But mean-reversion will be a strong gravitational force across categories over the coming years.
Investors, founders, and employees must relentlessly seek signal within the noise generated over the past two years to assess what changes are durable vs. transitory.
The challenge for investors is to find non-obvious but permanent shifts. Unfortunately, ubiquitous 100x ARR multiples is unlikely to be one of them.
Investors and the media cheered on VC firm AUM expansion over the past 24 months. Deal sizes and deployment speed skyrocketed. Law firms were backlogged for months due to unprecedented deal pace. Hordes of venture tourists arrived.
2021 was peak VC imperialism.
The most bullish pandemic-era tech investors are now in a world of hurt. SoftBank registered a $27b loss on their Vision Fund in Q1. Tiger Global is down 44% on its public hedge fund through April, not counting the even steeper May correction. Their private book valuations are likely down even more (at least they have preferred shares).
I wouldn’t be surprised if some of these crossover firms get margin calls – their public books often run on leverage – and some could implode entirely. Going forward, LPs will grade new managers and markups on illiquid positions much more harshly. The private market correction is somewhat of a slow motion train wreck – the markdowns and failed fundraises don’t happen immediately.
The silver lining for investors is that many companies with stalled growth due to mean-reversion are now valued on a much more punitive profit multiple. Shopify, Netflix and Zoom, for example, saw multiple compression of 80% as their earnings reports showed slowing growth. Valuations for true compounders should renormalize as their growth engines prove durability.
Where should investors focus in an ecosystem reverting to the mean?
Wish list: In 2021, every software company with a pulse warranted a 100x valuation multiple. This meant mediocre companies were massively overvalued. But the inverse is also true: in a market reset, multiple correction means that the best companies will go on sale, too. Come up with a target list of your top 10 picks so you’re ready if and when companies reach renormalized prices.
Counter-cyclicality: Investors always tell their LPs that they invest in secular shifts – they’re currently figuring out how secular some of those shifts really are. But an even better strategy may be to find counter-cyclical categories: energy, healthcare, infrastructure, and defense are contenders for truly macro-proof industries.
Consolidation: We’re likely to see strategic M&A and rollups, particularly in mid-stage software startups that don't achieve escape velocity. Companies that cannot grow fast enough to justify high burn will need to find acquirers as private funding markets soften.
Remote work: Hybrid and fully remote work is far more permanent than people expected. Remember that in March 2020, people expected a return to the office within a couple of weeks. The return date slipped for weeks, then months, and in many cases, indefinitely. Today, hybrid work is commonplace for white collar workers at large companies, and many startups are building products to support the shift.
Crypto caution: Crypto and web3 exploded during COVID. Non-BTC crypto is up over 10x since 2020. Billions flowed into projects promising fast profits. Enthusiasts argue crypto disruption of traditional financial institutions is imminent, but the 2021 run-up seemed more driven by a collective belief in the ROI of lottery tickets – a symptom of easy money. Is web3 overvalued?
Institutional mistrust: Distrust of centralized institutions – government, big tech, schools – skyrocketed following pandemic-era authoritarianism. These second-order tailwinds could linger for years, opening up opportunities for startups building the future of privacy, transparency, decentralization, and education.
In a return to normalcy, founders should revert to 2019 benchmarks for fundraising and operations, and be careful in absorbing any pandemic tailwinds they benefited from.
Agile spend: Some spend, like online ads, is easy to control: if you want to cut your ad spend in half, you can do so immediately. But some costs are hard to unwind: Peloton massively ramping up hardware spend assumed a sustained demand level that was in fact temporary. Hiring to meet demand spikes is tricky, too: see Hopin’s recent layoffs as a cautionary tale. Avoid fixed costs for variable demand, and be ready to unwind them quickly if demand softens.
Fundraising expectations: VCs have been underwriting your business based on high-growth public comps, so look to public companies’ valuations and strategies as guidance for how to benchmark your business and build out your investor narrative. Getting fundraising tips from founders who raised in 2021 is irrelevant – it is far more useful to ask founders who raised pre-pandemic.
Isolate pandemic and secular demand: An exercise in finance and data science, the best teams will isolate temporary and permanent shifts in business fundamentals – CAC, retention, LTVs. Assessing these variables requires ruthless intellectual honesty. Gauging retention means filtering through customers with a recent renewal option: are customers purchasing and engaging with your product at the same rates post-COVID?
Geography: The pandemic shed light on San Francisco’s many problems. The positive network effects of Silicon Valley are increasingly offset by negative network effects – homelessness, taxes, political groupthink. But physical offices drive productivity and culture, while free-rein remote work lowers employee accountability and learning by osmosis. Is returning to the office undervalued?
Profitability: The best time to learn how to build a finance model was when you raised outside capital. The second best time is now. Having visibility over your business levers is table stakes. A concerning number of companies raised on a simple deck or memo in 2021 – if you have never been through a challenging fundraise, you could be in a world of pain if you don’t control your financial destiny. Profit is the great equalizer.
Unsavvy employees can get crushed by the monstrous valuations of 2021 that take years to grow into.
Scrutinize businesses: When evaluating prospective employers, candidates often substitute valuation for job security. If you’re assessing a company that last raised in 2021, don’t take solace in the billion dollar valuation they fetched. Understand their actual revenue scale and financial health (margins, runway, growth rate) to get a sense for the magnitude of the business. Great businesses will survive, but it could take years for the fundamentals to catch up to 2021 entry prices. If you’re uncertain on how to value various companies yourself, speaking with tech investors is a reasonable substitute.
Assess product durability: When looking at job opportunities, be discerning on the durability of the product value proposition. If they experienced exponential growth during the pandemic, will it continue after trends revert to the mean? This should be fairly obvious at a category level: mission-critical infrastructure software is likely durable; a levered bet on virtual events may not be.
Find recent data points: Put increased weight on current data points from startups. Recent fundraises and signs of product-market fit carry additional merit – bonus points for indicators from the past one or two months. Of course, joining seed-stage companies is a completely different risk profile, which is relatively macro-agnostic.
Two years with COVID have been hectic, but what actually changed? A return to normalcy is the boring answer that nobody wants to hear.
But reversion to the mean seems to be an increasingly powerful explanatory force, and not just in tech:
Political unrest: From the BLM protests to the January 6 Capitol riots, protests spiked across the political spectrum during the pandemic. The reduction of face-to-face human interaction made humans far more volatile. With more human contact and less idle time, politics may stabilize going forward.
Retail trading: Retail trading volumes reached a new baseline in the pandemic – volatility, boredom, and stimulus checks are a powerful cocktail. GameStop and Dogecoin represented peak retail trading. Volumes have already started to renormalize.
Authoritarianism: From country-wide lockdowns to permanent Twitter bans, authoritarianism popularized during COVID. Thankfully, freedom antibodies have started to form: SF residents recalled the school board to focus on returning kids to school, the Twitter algorithm could become ban-resistant and open source, and the mask mandates have been struck down.
Consumer habits: Travel halted, movie theaters shuttered, retail spending and GDP plummeted in 2020. The speed of return has been uneven across categories, but travel is nearing pre-pandemic levels, GDP has reverted to the trendline, and movie tickets are on the rebound. I wouldn’t invest in a movie theater chain in 2022, but the death of the old economy is further out than the 2020 forecast suggested.
Not everything will revert to the mean. Movie theaters may never reach 2019 levels, Zoom is here to stay, and even I have been surprised by the durability of remote work. And mean-reversion does not spread at the same rate for every phenomenon: inflation will persist far longer than home fitness sales.
But many pandemic-era trends will slowly revert to their pre-2020 trendline – inflated valuations, easy money, exponential tech demand, meme investing.
Mean-reversion and inflation will be lingering symptoms of the virus for years to come.
Thank you to Harry Elliott, Brandon Camhi, Axel Ericsson, Michael Solana, Amin Mirzadegan, Philip Clark, Lisa Wehden, Pranav Singhvi, Melisa Tokmak, Russell Kaplan, and Everett Randle for their thoughts and feedback on this article.
As always, excellent piece.
A thought on step change versus reversion.
The way I think about it is segregating the step change into the following categories:
- Temporary adopters due to Covid who would never have adopted (1).
- Permanent adopters due to Covid who would never have adopted (2).
- Would have adopted eventually but did so due to Covid (3).
The proportion of category (1) vs (2) determines the permanent effect of the step change (level effect) while (3) is merely time shift and so determines the degree of mean reversion.
What really matters is the degree to which Covid actually expanded the size of the market rather than just accelerating adoption. From a financial perspective, it is, obviously, better to have cash today than tomorrow so acceleration does increase business value but the increase is actually pretty small since the majority of business value (for equity) is in its terminal value. ie. the market size effect is much larger than the timing of cash flows.
As you wrote at the end, not everything will entirely revert to the mean. For example, I do think that food delivery has a lot of category (2) where a larger number of meal occasions are now permanently fulfilled via delivery than would have been previously. Anyhow, what is clear to me, viewing it through this lens, is your point around tailwinds. Since the largest proportion of demand change is pull forward, it makes the future prospects of internet enabled businesses much bleaker post Covid.
Great work as usual