SignalRank’s anti-portfolio
Bessemer popularized the public anti-portfolio, by honoring on their website the startups where they did not invest which went on to blossom into successful companies. This list includes Airbnb, Apple, ebay, Facebook, Fedex, Google, Intel, Paypal, Snap, Tesla & Zoom. Apple was deemed “outrageously expensive” at a $60m valuation. Apple’s market cap today is almost $3.5 trillion. 🤦
A benefit of being a purely quantitative investor is that there is transparency around what would have qualified for our model. This enables rigorous backtesting to demonstrate to potential investors the validity of our approach. It also allows us to generate an anti-portfolio of companies which qualified for our product but we missed.
We are conscious that there is no such concept as an '“anti-portfolio” in public market indices, as access is not an issue. This highlights how any index of private markets needs to demonstrate high quality access to be able to confidently assert that their index is representative of the underlying asset class they are seeking to reflect.
SignalRank has now invested in 25 Series Bs since May 2023, making SignalRank the #2 most active investor globally in high quality Series Bs. Given this milestone, we thought it could be a good opportunity to delve into our investment process in a little more detail, consider why we have missed certain companies and propose how we might be able to do better.
SignalRank’s process in a nutshell
Figure 1. Overview of SignalRank’s process
SignalRank exclusively invests in first priced Series B rounds (Figure 1). We invest in companies with high company scores at Series B, where we primarily care about the quality of the Series B round, but we also take into account the quality of prior rounds. Quality at every round (“round score”) is defined by the combined investor score of each investor participating in each round. Investor scores assess the relative quality of an investor for each round, looking at various metrics from the last five years, including MOIC, number of power law companies and power law efficiency.
Companies with high company scores perform admirably. By way of example, here is a table showing the #1 ranked qualifying Series B per vintage since 2012 (Figure 2).
Figure 2. Top ranked qualifying Series B per vintage since 2012
This model is dynamic to the market based on a thresholding system to ensure that we are only investing in the top percentiles of each Series B vintage. The thresholds assess the “heat” of the Series B market and the quality of any given Series B round relative to other Series Bs in the prior 90 days. This is protection against a repeat of the 2021 phenomenon where high quality investors were over investing. There is significant benefit in having a quantitative model which has now seen a full cycle.
There are roughly 1,500 to 2,000 Series Bs completed globally each year. We like Series Bs which already raised high quality rounds at seed & Series A (where we call these post A / pre B companies “candidates”). Roughly 100 companies raise a qualifying Series B each year, representing c.5% of all Series Bs.
An annual cohort of qualifying Series Bs delivers on average at least 5.0x MOIC in five years. 30% of these companies go on to attract valuations of $1bn+ (compared to the market average for all Series Bs of c.10%).
We then ask the question of what is the minimum number of investments that we need to make each year to approximate the average of these qualifying Series Bs based on our Monte Carlo simulation. The answer we arrive at is between ~30 per year.
The algorithm is sifting out companies which do not have the characteristics we like. It is a negative algorithm that seeks to identify true negatives (where it is c.90% accurate in doing so). Put another way, we are NOT saying that an individual qualifier will be the next Rippling / Databricks. But a pool of assets with the characteristics we like has a higher probability of success (because we are eliminating zeroes). This pool will include multiple power law companies.
We seek to access all of the qualifying Series Bs with a view of investing in 30 per year. We invest on a first come first served basis. The selection of the 30 qualifiers is therefore a sample of all qualifying companies.
This model works as along as you are seeing the full distribution of qualifiers including the power law companies. If we are unable to access the next Stripe / Figma / Plaid etc, it could impact our returns for that particular cohort. Access is everything.
The next iteration of the model will include looking at ways to improve our access to the top 30 companies from each cohort of 100 qualifiers (while simultaneously managing overfitting algorithms). The start of this process requires analysis of our anti-portfolio to understand why we are not currently investing in all of the best qualifying companies. Let’s look at this now.
SignalRank’s anti-portfolio
SignalRank started to invest in May 2023. There have been 185 qualifying Series Bs since then. SignalRank has seen 49 of these qualifiers (27%), of which we invested in 20 (11%).
Why are we only seeing 27% of qualifiers?
No seed partner. In the majority of cases, the company had a micro VC / seed manager on the cap table but we simply did not have a relationship with the relevant micro VC. This suggests that SignalRank continues to fly somewhat under the radar (and/or that our partnership team needs to get out more). In truth, this is about capital availability: we are currently writing <$1m checks (which is not interesting enough for some seed managers); once we are at scale with checks of up to $5m per investment, we anticipate a surplus of demand. In the meantime, we need to balance supply of capital with demand from seed managers.
No micro VCs on the cap table. In many other cases, there were zero seed partners on the cap table with whom we could partner. If a company is exclusively backed by say Sequoia Capital at seed and Series A, then by Andreessen Horowitz at Series B, then there are zero seed partners with whom we could work with. We have a potential product here which we believe will be very disruptive: founder pro rata. We simply finance the founders’ pro rata with the same economic model we have for our seed partners (ie 20% upside), thereby helping the founder defend their position and voting rights. It is the most founder friendly capital on the market. Watch this space.
Of the 49 qualifiers we have seen, why have we only invested in 20?
Too late to the round / no seed partner. The main reason is similar to above. We heard on the grapevine that these rounds were happening but did not have a relationship with a relevant seed manager. This group also covers cases where we were slightly too late to the round and/or where our current check size is too small (as some seed partners only want to work with us when we can write checks approaching $5m).
Oversubscribed. In a small number of cases, the round was too competitive and our seed partner was unable to secure access because the lead investor squeezed out existing investors. This does happen every so often, but not regularly as the lead investors can attract a poor reputation among seed investors (for not allowing them to invest their pro rata) who in turn will recommend to their founders not to take money from these same lead investors.
Failed to qualify on old model. We have upgraded our model this year. We previously excluded insider rounds from qualification which led us to rejecting a number of companies which would now qualify. In some cases, the Series B looked like an insider round when we saw it, but then went on to be an externally led round.
Wrong share class. We do not buy common shares, so passed on a couple of rounds where we were offered secondaries at the time of the Series B.
That’s all well and good. But which high ranking qualifying Series Bs have we seen but missed that are now considered to be “hot”?
Perplexity – we were offered access into the Series B+ round (not a priced B)
Hebbia – we were offered common secondary (not preferred B stock)
Eleven Labs – insider round at the time we saw it (failed under old model)
Hadrian - insider round at the time we saw it (failed under old model)
Motherduck - we were too late to the round (lacked a seed partner at the time)
It is inevitable that some of our anti portfolio companies will be compelling. Some of the above misses sting. But we will continue to iterate and tweak the model to enable us to improve the quality of our portfolio.
Some concluding thoughts on our anti-portfolio
We give ourselves a B- for our access and approach in 2023. We already saw significant improvement in 2024, marking ourselves as a B+ last year. The table below shows how we have invested in two of the top 10 ranked qualifying Series Bs in 2024, compared to zero in 2023. Let’s strive for further improvement in 2025.
Figure 3. SignalRank’s access to top 10 ranked qualifying Series Bs in 2023 & 2024
