Data Business Evaluation

An easy-to-use cheat sheet

I was recently asked to speak to a group of private equity investors about my career, what I have learned from it, and what opportunities I see because of those experiences.

Given my familiarity with data businesses, I chose to tie things together with a focus on them, particularly how to identify and evaluate them quickly.

Why data businesses? Because they can be phenomenal businesses with extremely high gross margins — as good or better than software-as-a-service (SaaS). Often data businesses can be the best businesses within the industries that they serve.

As a way to help crystalize my thinking on the subject and to provide a framework to help the investors separate the wheat from the chafe, I created a Data Business Evaluation Cheat Sheet.

For those yearning for a definition of data businesses, let’s purposefully keep it broad — say businesses that sell some form of data or information.

I broke out four evaluation criteria: Data Sources, Data Uses, Nice-to-Haves, and Business Models.

Each of those criteria have a variety of flavors, and, while the flavors are listed in the columns of the cheat sheet in order of the amount of value they typically create, it is important to say that many successful data businesses have been created based on each criterion, even the lower ranked criteria, or combinations thereof.

Data Sources:

How a data business produces the data that it sells is of critical importance to the value of the business, primarily in terms of building a moat versus competition and/or creating a meaningful point of differentiation.

Generally, the best source of data is proprietary data that the company intentionally produces via a costly to replicate process such as the TV broadcast capture network operated by Onclusive. Sometimes a company is lucky enough to produce this asset as a form of data exhaust from its core business operations such as salary data from Glassdoor. Data exhaust can be used to power another part of the same company that produces it, such as Amazon using purchase data to power its advertising business. Sometimes data exhaust is used to power a separate data business or businesses such as NCS which leverages register data from Catalina Marketing’s coupon business.

A close second in terms of a valuable data source are those businesses that benefit from a network effect when gathering data. Typically, there is a give-to-get structure in place where each data provider gets some valuable data in return for the data that it supplies. When the data that everyone gets benefits from the participation of each incremental participant, there is a network effect. TruthSet is an example of a company that runs a network effect empowered give-to-get model by gathering demographic data from data providers and returning to them ratings on the accuracy of their data.

Data aggregation can be a valuable way to assemble a data asset as well, but the value typically hinges on the difficulty of assembling the data…if it is too easy to do, others will do it as well and create price competition. Often the value comes in aggregating a long tail of data that is costly to do more than once either for the suppliers or a competitive aggregator. Equilar, for example, aggregates compensation data from innumerable corporate filings to create compensation benchmarking data. Interestingly, Equilar has used the data exhaust from its compensation data business which sells to H.R. departments, to create a relationship mapping business which sells to sales and marketing departments and deal oriented teams in financial service organizations.

Rounding out the set of key data sources are Data Enrichment and Data Analytics. Data enrichment creates value by enhancing another data source, sometimes by adding proprietary data, blending data with other data, structuring data, etc. Data Analytics companies make their mark by making sense of data from other sources often via insights that make mountains of data more digestible and actionable.

Data Uses:

When you can find a data set that is used between companies, it can be quite valuable because it is often difficult to displace. The best examples of this are when other companies trade value based on the data and the data fluctuates regularly. In this case, the data has become a currency. Examples of this include FICO scores from FICO or TV viewership data from Nielsen. In special occasions, data used between companies can also benefit from a network effect which tends to create value to the data provider. In fact, it can be the network effect that drives a data set to become a currency.

Data that is used to power workflows can also be very valuable, because moving away from it may require customers to change the way they work. The more frequent and the higher value the actions and decisions supported by the data, the better.

Typically, the more numerous the users of a data set within a customer and the more types of companies that can make use of the data, the more valuable it is. One way to evaluate this is to observe how easily and often a data set is combined with other data sets. Relatedly, some companies make a great business out of providing unique identifiers to enable data combinations such as Datavant which provides patient IDs to facilitate the exchange of healthcare data. Others, such as Dun & Bradstreet, use unique identifiers such as the DUNS Number, to support other parts of their business — credit reporting in this case.

Data Nice-to-Haves:

Data Nice-to-Haves are features of data businesses that typically do not sustain a data business on their own, but can significantly enhance the value of a data business. Benchmarks / Norms and Historical / Longitudinal data are typically attributes of data businesses that have been around for an extended period of time and can be difficult to replicate. The same goes for branded data though sometimes a brand can be extended to a new dataset once the brand is established — think J. D. Power applying its name to its acquisitions over the years.

Data Business Models:

The highest gross margin data businesses are generally those employing a syndicated business model in which they are selling the same set of data over and over again to different parties with no customization or with configuration that is handled automatically, preferably by the customer. The most stable data businesses tend to employ a subscription business model in which customers subscribe to a data set for an extended period of time. Subscriptions models are clearly better when the subscriptions are long term or, at least, auto-renewing.

Not surprisingly, the best data businesses are generally syndicated subscription models. On the other end, custom data businesses that produce data for clients in a one-off or project-based manner generally struggle to attain high margins and predictability, but can be solid businesses if the data manufacturing processes are optimized and they benefit from some of the other key data business attributes.

Financials:

The intent of this overview and associated cheat sheet are to analyze data business fundamentals, not to dig into typical financials. That said, typical healthy data businesses often have gross margins in the 65–75% range and EBITDA of 25–30%. Both of those number can be higher for exceptional businesses with the best of the characteristics above. Even in relatively slow growing industries, these businesses can trade at 15–18x EBITDA due to their resilience and the fact that they often require only modest R&D and marketing expense.

My parting thought to this group of investors, who were focused on a variety of verticals, was that though they may not consider themselves data business experts, they ought to be on the lookout for great data businesses in their areas of expertise. It is their detailed understanding of how their various industries work and what data their companies rely upon that will point them in the right direction and possibly uncover some data business gems for their partners who do invest in data businesses.

I’d like to thank several data business aficionados that helped inspire and refine my thinking on the subject: John Burbank, David Clark, Conor Flemming, Andrew Feigenson, Travis May, and Andrew Somosi. In particular, I’d recommend Travis’ related article, which was of particular inspiration to me, as additional reading on this subject.

Data Business Evaluation was originally published in Towards Data Science on Medium, where people are continuing the conversation by highlighting and responding to this story.