SignalRank AI - Track Record Matters
Over 220 Unicorns in ten years, with 540% average gain 5 years after the Series B investment
SignalRank’s purpose is to bring data analytics and AI to the venture ecosystem to improve it for all stakeholders. SignalRank scores Investors, Companies and Funding Rounds to select the top 5% of Series B rounds in real time. SignalRank is now deploying capital to support our partners maintenance of their ownership at Series B in the AI selected companies.
A reasonable person might ask themselves if SignalRank AI’s predictive data platform is too good to be true. After all, an average five year gain of 540% with between 20 and 35% unicorns in any given year (see below) would be a world beating investment performance, and it is. Why is it so good at predicting outcomes and why should investors believe it? Also, can it be relied on for similar results in future?
SignalRank AI’s Track Record
When limited partners invest in venture funds they often ask about the track record of the general partners in the fund, and services like Pitchbook and Prequin measure that performance and report on it.
For a data-driven investor, where a data platform is making all of the allocation decisions, and humans are making none of them, a track record is also important. How good has SignalRank AI been in the past?
In data driven investing that is the role of a backtest. It has an advantage over human investors, who cannot go back in time and forget everything they know, and test their capabilities. With a data driven approach it is possible to test it in real world conditions and see how it would operate. We can show what the data platform would have done in any given year.
A backtest has to be “non-cheating”. That is to say, the data platform should only have access to data prior to the date of its predictions, and have no access to knowledge of future outcomes. So a prediction for investments made in 2012 should only have data up until Dec 31st 2011. And so on.
Based on this SignalRank AI has ten years of performance data, dating back to 2012. Each annual cohort was chosen based on the prevailing data from the prior year only.
2012 is a key year because it marks the five year point after the first seed funds began investing - around 2007 - and those investments have a five year outcome based on valuation upticks and exits that forms the basis for learning from the data.
By 2012 it is possible to begin to profile seed rounds and A rounds and to learn the signals that represent success and distinguish them from the rest that do not.
How the SignalRank AI Backtest Works
SignalRank begins by profiling all seed investments and Series A investments in a five year backward looking window. How many investments were made? What happened to them in the years since the investments were made?
Using 2012 as an example, it is able to look at and score seed, Series A and Series B rounds in the prior five years - 2007,2008,2009,2010 and 2011.
By 2012 some of those investments are five years old, others are only one year old. Each year is treated separately and our model is formed by learning what significant features correlate to the best outcomes.
SignalRank runs many models and each model is unique. For this update we will use what is called the three signals model. It was the first model we developed and still is core to SignalRank AI’s predictions.
We can use 2012-2018 as our backtest period, because there is a five year future for each of these years so that we can measure actual success for the selected companies.
The table for those six years is here:
Let’s use the 2012 column so we understand how to analyze the backtest across all six years.
There were 797 Series B Rounds in our data set in 2012 as row 1 shows. SignalRank AI used the three signals method to select five companies or less than 1% of the total. This is a low number of choices, and reflects the sparsity of data up to 2011 in scoring Series B rounds. The list of companies selected in 2012 is as follows:
We can now read the columns for each of the six years.
As you can see the five year MOIC (multiple of invested capital) average is in a range from 4.18 to 16.74, and the average across all six years is 8.4x. The % Unicorns is in a range from 20% to 40%. 109 unicorns in six years from 390 selected Series B rounds, give a performance of over 27.98% unicorns.
It is worth repeating that all of these selections were made only with data from prior years and before knowing the outcomes.
The list of unicorns selected at the Series B round, the 109, is as follows:
SignalRank does not disclose all of the features and weights that are part of its predictive data platform, or of the three signals method. But it does publish which companies the platform selects. Investors in SignalRank have access to real-time scoring of seed, Series A and Series B rounds.
And here are the unicorns selected for all years 2012-2023 shown annually.
Is the past any guarantee of the future?
A common question from potential investors into SignalRanks pool of investment capital is whether the predictive data models will perform in future as well as they did in the past.
Of course it’s a good question and the answer is certainly neither ‘yes’ nor ‘no’. It’s more subtle than that.
The backtests vary year by year, and the outcomes, whether measured in unicorns produced as a percentage of selections, or MOIC at the five year point. But the range of outcomes is surprisingly stable within a range. The reason for that is that the models are tuned to avoid failures and seek out the top 5% of outcomes irrespective of macro level market changes. As such the model tunes to the market as the market changes and produces quite consistent relative outcomes.
For example, the number of Series B rounds in 2023 is significantly down from recent years. There are, as of today, 675 rounds. The selections amount to 26 companies:
Each of these has been announced on the @SignalRank Twitter Feed - like this:
As you can see in the graphic, the three signals happen at the seed, Series A and Series B round. Only companies that have all three signals are selected by SignalRank AI using this method.
26/675 means that year to date the predictive data platform has selected 3.85% of Series B rounds as qualifying. This is lower than in prior years and reflects the platform’s tuning due to there being less Series B rounds, creating a situation where quality rounds are set a higher bar.
For these companies to perform as the prior year’s selections have requires that they go on to do Series C and Series D rounds, some get acquired and others IPO. The seed, Series A and Series B rounds contain signals suggesting that the performance will be in the top 5% of performance of all Series Bs from 2023.
We will only know in 2028. But we are confident it will be the case as the core features in the signaling are resilient to changing markets and that in relative terms 2023 will not look significantly different to 2012-18.
Great Selection and access to invest?
Assuming it invested in all selected Series B rounds since 2012, SignalRank’s three signals predictive data model is already the world’s best Series B investor. With over 220 Unicorns selected in that time.
But how can SignalRank place capital into the selected rounds? these are some of the most highly prized Series B rounds each year. The answer lays in our business model.
SignalRank’s goal is to support seed and Series A investors to maintain ownership in their best companies. We have over 80 relationships to early stage fund managers who solicit our capital when a Series B comes along.
We began investing in April 2023 and have been notified of over 170 planned Series B rounds by well over 100 managers seeking to maintain ownership. We have approved five investments because they correlated to SignalRank AI’s selections. These Series Bs have Andreessen Horowitz (2 of them) Khosla Ventures, Union Square Ventures, Lightspeed Ventures, Northzone, CRV and Madrona as investors. They are highly rated companies with top decile investors.
As we said in the first newsletter:
SignalRank will raise capital to invest in Series Bs by selling up to 40 million preferred shares over time, initially priced at $25. The share price will rise as the assets appreciate.Share purchases will only be available to overseas and US-accredited investors and organizations.
Share sales will open each quarter, priced at the current fair market value, and spread over 2-3 years.
Our next raise is on June 30th with subsequent raises in September and December this year. Each quarter we will continue to deploy capital into selected Series B rounds in support of the follow on rights of our partners.
Keith Teare/CEO
To learn more go to https://signalrank.co
Appendix
For Reference, here are all selections made by SignalRank AI (three signals method) 2012-2023.