Leave risky business to Tom Cruise
A bad investment usually feels a bad hangover after a failed casino trip. It usually leaves you questioning your ability to make choices. But decision-making for new business investments or new business models does not have to be as unsuccessful as a casino trip to Vegas. Sure, where there is risk, there is failure. But it is possible to lower the odds of misfires with informed corporate investment decisions. Who better to analyze, filter and process data than decidedly unemotional machines?
Whether to aid the investment decision in startups (external) or to decide on a corporate innovation investment (internal), VCs and companies look to AI tools to determine the smarter way to back intra- or entrepreneurship. It turns out that a surprising amount of the decision-making can be replaced by new technologies because it eliminates emotion and bias. One way to learn how comes from an investment algorithm applied in Hollywood. Spoiler alert: We’re not talking about the Netflix recommendation engine…
Venture capitalist bias, buh-bye
Traditionally, corporate venture capitalists rely on Excel charts, market analysis, and due diligence in conjunction with human oversight. Human beings “with their infinite wisdom” generate decisions filtered through vast experience. They mine their networks, get feedback and go over the numbers with their own take. It is more art than science, or at least the method is an interpretation which implies human-centered capacities like feelings, intuition, instincts.
Google Ventures has already dismissed this approach in favor of quantifying innovation. According to The New York Times, Google’s “rising V.C. arm focuses not on the art of the deal, but on the science of the deal. First, data is collected, collated, analyzed. Only then does the money start to flow.”
No one likes to lose time or money. Not the corporate investors of a new internal innovation project, not a VC investor—even when it’s a relatively small investment. One of the best reasons to use artificial intelligence is it removes natural biases and blind spots to better assess new business investments.
Conventional wisdom has led to blind spots in the film industry, too. ScriptBook – a tool that approximates box office revenue through screenplay analysis– can determine through machine learning if a creative choice as innocuous as a hero dressed in a blue shirt will perform better than a hero in a red shirt. For the scriptwriter who envisions the red shirt, it is harder to let go of imagining a hero sporting something other than a crimson top—until you add data to the argument.
ScriptBook was built to limit faulty decision-making. Founder Nadira Azermai (who came up with the ScriptBook concept following a stint as an intern on the box-office bomb Gigli; at minus $70 million, it was one of the greatest money-losers of all time. See more here.). Azermai blames the current human decision-making process in determining what gets greenlit. Currently, 6.3% of the total Hollywood films made are responsible for 80% of the film industry’s overall success. With figures like that it is no wonder Hollywood would consider quant innovation and investing.
The returns in tech show it is a good idea. Data scientist-slash-venture capitalist, Thomas Thurston–a Partner at the VC firm WR Hambrecht Ventures, has been consistently in the top 1% of returns within the VC industry–thinks that the answer to good investing lies in math 70-80% of the time. Growth Science is the R&D partnering organization for WR Hambrecht Ventures’ proprietary MESE computing system, a tool which delivers “unique insights into businesses that disclose little or no public data about their financial or commercial status.” All of their new business investments are evidence-based, after investing nearly a decade of time and resources in MESE.
Since 2008, MESE has served executives in guiding billions of dollars in venture capital, innovation, acquisitions and related growth initiatives. According to Thurston, people are often wrong in their venture capital investments and wait for the “hits,” or the top-performing startup investments to compensate for the turkeys.
Just like the tendency of CEOs to skip reading business plans prior to investing, executives in Hollywood personally read only about 100 out of the 25K screenplays that make it to production. Before scripts even make it to an agent’s desk, they get “coverage,” where professional readers rate the script and write a summary. It is a broken system; the few deemed worthy enough to green light are ultimately the ones being rated by “experts,” or human beings with personal tastes and preferences. ScriptBook’s machine-based analytics outperform humans 3X better. But these investment tools and algorithms are not just for the elusive Hollywood producers.
Read on to discover the latest venture capital tools and algorithms to supercharge new business investments.
Go from gut-based to evidence-based decision making
From Thurston’s perspective, AI just offers better vetting. “Innovators have not had the analytic tools that you see in a lot of other areas of business. We’re always looking for something that we can measure and quantify. In the past couple of years, innovation was all about serendipity and try and fail,” says Thurston. He notes it’s harder for corporates to permit allowing for “a learning experience” when it involves a five million dollar innovation investment.
The process is much like a giant computer simulation performing actuarial calculations giving high degrees of accuracy. In a Fortune interview, Thurston stated, “We’ve found only around 20% of the predictive value to come from details specific to the startup itself (e.g., the team), whereas 80% comes from things outside of the startup,” which he listed as the market, customers, competitors, technology trends, and timing. Growth Science, with its MESE consulting tool, can predict out of a portfolio which ventures are likely to fail or succeed, just like ScriptBook.
Typically, selected innovation investments are chosen due to emotional biases towards certain people and/or proximity. When we think of needs we tend to turn to our immediate network or neighborhood. Data-backed decision modeling can cast a wider net; startups or intrapreneurial ideas can be evaluated across genres and geographies.
Robert Bonanzinga of InReach Ventures had a pace of seeing 1,500 companies a month (of which only 100 would advance to the next round) all to net a single deal. Getting on Bonanzinga’s radar was tough; the bulk of startups were located in a tech-centric locations, like London or Silicon Valley. To compensate for proximity bias and the tendency to continually fish in the same pond, InReach got a £5M investment to help build a machine-learning tool. Now with data, InReach can ferret out 95,000 startups across all of Europe, generating 2000 potential deals. Expanded terrain means a broader scope: different people, ideas and business models. That means a greater chance of success, not to mention the wins for efficiency and productivity.
Eliminating the mission impossibles
Just as aspiring filmmakers can pick apart a film or movie and figure out what camera angle ruined a shot and why, corporate venture capitalists can also make use of preto-typing, or what customer development specialist Sean K. Murphy termed a “Picnic in the Graveyard.” Scanning what went wrong in failed startups gives input into the obstacles specific to your innovative idea.
Autopsy (or Startup Graveyard and Startup Cemetery) is a data-based tool that can help you tweak your business model before you start, based on how similar startups have failed in the past. Get Autopsy can analyze a portfolio of past corporate investments, and turn those failure insights into actionable investment advice.
Selecting the right vehicle to invest in is half the battle. ScriptBook’s AI method of choice, machine learning, finds patterns without being given set parameters; ML algorithms are advantageous because they operate without assumptions making discoveries wholly organic. From the large number of data sets, indicators emerge. The algorithm continues to learn and find patterns hidden in the data (including from non-linear patterns) and make predictions otherwise inconceivable with statistical models.
At present, most venture capital firms already use elements of AI in their evaluations, together with the standard Excels, and incorporate narrow, weak AI and structured data. More visionary firms, however, use a mix and include unstructured data, too.
SignalFire, a San Francisco-based VC that invests in seed stage and breakout companies, tracks more than 2,000,000 data sources and a half a trillion data points to give its portfolio companies unique insights into market intelligence answering questions like: “How much does a certain product cost in different geographies,” “How much of a discount would impact revenue growth and profit margins,” “How change to an offering would compare to competitors” to help with everything from pricing to cohort analysis. Business model science is the next frontier in creating healthy operations and ensuring profitability.
Opting for a proprietary method, SignalFire built software that currently tracks 8,000,000 startups across the world, using everything from sales figures to academic publications to financial reporting. Companies that stand out appear on an easy-to-read dashboard. Eliminating the complicated reporting renders more efficient investment metrics by default. And isn’t the main question nowadays about how to measure innovation?
A crystal ball for innovation investment decisions
Top Gun performance in innovation analytics for corporate venture capital funds come from being able to use a robust mix of big data. Since the rise of the Internet and the accompanying googolplex of information, algorithms have been applied wherever they can simplify challenges. Risk in the venture capital sector remains the scariest villain, but–conversely–as the possibilities with artificial intelligence expand, it is the narrow use of structured data that mitigates specific business anxieties.
Now, custom venture capital algorithms can make predictions based on the performance of a venture’s historical data. Benchmarking, using information like business models in different industries, looking at root-cause analysis have helped unlock critical information like the amount of investment required, mapping out the top performance indicators, the predicted success factors, typical challenges, etc.—all welcome information for the risk-averse.
In the case of ScriptBook, however, with its ability to process unstructured data, (screenplays are built on emotive language and depict a visual story, after all) it becomes evident how reading business plans or interpreting a pitch can be done—and accurately to boot. ScriptBook’s analysis of the script to The Passenger in 2015 was predicted to earn $118 million at the box office. Actual revenue? Around $118 million. Which innovation manager would not like a better crystal ball to help make innovation investment decisions?
The solution cocktail
Though WR Hambrecht Ventures (with MESE), InReach Ventures and Signal Fire all invested multiple millions in a proprietary tool, ScriptBook first started with a modest investment of €25,000 (not counting the countless development hours). It is possible to employ analytics, measure innovation or use an array of metrics to improve your new business investment.
All of these tools are covering a part of the innovation investment process:
Signum sees product and service trends, detects growing markets, and reports on the most promising companies.
Klydo can turbocharge desk research plus find hidden links and relationships, which corporate/venture capitalist managers might appreciate.
Just like how predictive algorithms develop and correct, the quantitative investment industry is continuously improving, too. Jed Dougherty of Dataiku, said companies are trying to make user interfaces for machine learning and predictive modeling more visual, “so someone who knows the business very well can leverage the algorithms” meaning AI can be harnessed by HR, sales and marketing, too.
Besides, we are no longer in the era of digital transformation—the next leap indicates a quantum transformation. But giving up the sense of autonomy is one of the challenges. There is something about sitting in front of an Excel chart and physically manipulating data. It has the essence of self-actualization, whereas handing off an innovation investment decision to algorithms and robots makes one feel superfluous—much like the script readers who wrote coverage for screenplays.
When consultants stop offering advice and start offering investment predictions, WR Hambrecht Ventures’ Thurston notes that clients will not love you. “You show up and say that great, disruptive idea that they vetted as their next innovation investment has about a 12% probability of success. People won’t like you.”
It does not have to be that way. Corporate venture capitalists can use any type of data, and just like a human, autonomously learn. Once the risk is tamed, AI still can perform quantum miracles in productivity and efficiency. The business model canvas, assumption maps, business deal flow—all areas that AI can assist with right now. It makes the once notorious gut-feel approach based on personal preferences, scientific and data-driven.
Data-driven decision-making means more innovation investment targets can be fully evaluated, so there’s no bomb, whether at the box office or the bottom line. Corporates, Venture Capitalists and yes, innovation managers, can all channel their inner Tom Cruise’s and ask an algorithm his most famous line in Jerry Maguire: “Show me the money.”
-  https://www.nytimes.com/2013/06/24/technology/venture-capital-blends-more-data-crunching-into-choice-of-targets.html?_r=0
-  http://fortune.com/2015/08/05/venture-capital-hits-average/?#
-  http://www.ft.com/content/dd7fa798-bfcd-11e7-823b-ed31693349d3 Maija Palmer Dec 12 2017
-  https://medium.com/fintech-weekly-magazine/3-ways-machine-learning-supercharges-investing-8d94a7516613
-  https://medium.com/fintech-weekly-magazine/3-ways-machine-learning-supercharges-investing-8d94a7516613