Databricks: how to follow the “enterprise software CVC playbook”

Barely over a year since it was launched, Databricks Ventures, the investment arm of the US data analytics company, has already announced nine investments — and it has closed several more deals that have yet to be announced.

To some extent the, team is making hay while the sun shines. The global data analytics market is growing fast — set to expand nearly 25% annually for the next five years, to a value of some $115bn by 2028, per Market Growth Reports. There is simply a lot to invest in when it comes to data, dubbed “the new oil” for some time now.

But unit chief Andrew Ferguson has also adopted a streamlined investment strategy to be able to cover as much ground as possible. They are opting to be followers in deals, finding good coinvestors to lead and do the operational heavy lifting. By not leading rounds, Databricks Ventures would not compete directly with VCs for deals, streamlining the investment process as the CVC itself is not setting valuations or negotiating deal documents.

“We don’t lead the round, we come alongside the lead investor – like an Andreessen or an NEA or someone like that – and we take a small part of that big round,” says Ferguson.

Databricks is following a certain kind of blueprint, one set by companies like Intel and Salesforce, that Ferguson considers the “modern enterprise software CVC playbook”. It means starting with small tickets, then growing into a full-fledged CVC and eventually leading rounds across all stages. Databricks is in the first, slightly passive stage of this, and may eventually look move into a more active role.

Right now, the key thing is to get efficiencies from riding along well with others.

“I’d say the model that that we launched with was trying to be both founder-friendly and also VC-friendly, so that way we can get into deals without conflicting either with what the founder’s looking for or what the lead VC is looking for,” he says.

“We can often get the strategic return we need without writing a gigantic check in the round. Similarly, we try to come up with strategic terms that are very lightweight and amenable to both the founder of the company and then also the other VCs, so that we can sort of play nicely with all the constituents in a financing round.”

For similar reasons, the unit doesn’t take any board seats, to avoid any additional friction that may come with having detailed information rights or commercial agreement signed at the same time as the investment.

“We’re trying to come up with, at least in this initial foray into Databricks Ventures, a streamlined process that doesn’t include those things like a board seat. That lets us get to yes or no on investments in a pretty quick and flexible way. Whereas if you had all these other strings attached to the investment dollars, it just makes the process a little bit harder.”

Power of relationships

Coinvestors are important in any deal, but when your strategy is one of following investment partners into deals, those relationships become more crucial. Investors that have led rounds that Databricks Ventures has joined include the likes of Kleiner Perkins, Andreessen Horowitz, Bessemer Venture Partners, SoftBank Vision Fund and Stripes. Relationships also become a great source of deal origination.

“We’ve been able to develop really good relationships with VCs that are investing in and around the data ecosystem, and often they’re the ones that are coming to us asking us to participate in rounds that they’re leading because they see the value that Databricks Ventures can provide to the company in a financing round. Some of it is market signalling, sort of a stamp of approval from Databricks as a leader in the data and analytics space.”

While it doesn’t lead, Databricks Ventures tries to act as a network multiplier for its counterparties – if a startup tells them there is room for more investors in a round, they help make introductions to round out the cap table, as well as due diligence for new investors that come on board even if Databricks itself doesn’t end up taking part in a round.

“The other thing VCs will often do is ask us what we think about some company. We’ve got super-smart software engineers, super smart product managers, field engineers, customers, that we can basically go to to help provide a little back-channel diligence as they’re thinking about an investment,” says Ferguson.      

“I’d say we have a unique ability to do technical product and customer diligence just by virtue of these companies often already being partners of Databricks, or us having joint customers with the prospective portfolio companies.”

Ferguson says the unit carries out quarterly check-ins with its VC network to be aware of each other’s priorities and focus areas.

Evergreen flexibility

The CVC invests out of a single-limited partner vehicle, named the Lakehouse Fund after Databricks’ Lakehouse data platform. The evergreen fund doesn’t have a set size and affords them the flexibility to invest as much or as little as they want without the time pressures that would come with your typical 10-year fund.

“If there are 10 great companies to invest in a year, fantastic. We’ll do it. And if there are only five or six, that’s also fine,” says Ferguson.

The unit typically does not go in earlier than series A, primarily because they need portfolio companies to already be mature enough to confidently put their solutions in front of Databricks’ enterprise customers.

Its portfolio currently includes collaborative training data platform Labelbox, collaborative analytics company Hex, machine learning features company Tecton and data productivity platform Matillion.

Picking winners

Databricks Ventures also avoids investing too early because can be hard it is to determine which companies will float to the top in a crowded field, especially when you might have as many as five or more startups backed by big-name investors in the same sub-segment.

Internally, Databricks may determine it needs good relationships with the top five prospects in a particular area pending stronger indicators of who may be a market leader. That complicates the investment side of things, though, as it may require waiting out the first few rounds until its customer base has a better idea of its preference.

The sector is characterised by the speed of technological change, and the artificial intelligence and machine learning tools segment is particularly difficult to predict, says Ferguson.

“Any individual company have a great technology, but it may be that these two pieces of technology really need to come together before it’ll get broad adoption.”

Taking the backseat is a savvy model for a new CVC, but it doesn’t mean it won’t evolve as time goes on.

Ferguson says: “I’d say our current model’s working pretty well, but certainly, Databricks is a company that has great ambition. So no immediate plans, but certainly if we’re successful, I think the scope of the venture program will grow over time”