How Glean Raised a $1M Pre-Seed Round
It was October 2011.
Seven months earlier, Aster Data — the big data pioneer where I was employee #1 — was acquired by Teradata. The acquisition left a halo around everyone involved and, with big data at peak hype, by Q4 many of of us were spinning up companies in adjacent areas.
Dheeraj, Mohit and Ajeet were hard at work on Nutanix. Manav, Hari and Raghu had teamed up on Instart Logic. Sharmila, John and Vaibhav had co-founded Clearstory Data. And I had partnered with my old roommate from Stanford, Jeff Zabel, to create an entirely new type of BI platform called Glean (oh…you thought I was talking about that Glean?).
Given the pedigrees of our respective teams and Aster Data’s reputation amongst investors, there was no shortage of interest in the companies being formed by Silicon Valley’s newest mafia. By late-2011, Nutanix had already raised two rounds of funding from top-tier VCs, while Instart Logic recently closed a seed round from Wing Ventures’ precursor fund. ActionIQ, Workspan, ThoughtSpot, Level Up Analytics, Cyberhaven and Cohesity were still to come. But in October, it was our turn to go to market.
Start Your Engines
The Aster Data acquisition gave our alumni a huge collective advantage when it came to fundraising: we could literally get meetings with any VC we wanted. In fact, many investors were so eager to meet with Aster Data’s alumni that they were reaching out the moment they heard rumor that one of us had left Teradata.
With such an advantage, Jeff and I figured that it would be easy for us to raise our first round of funding. Many of the Aster Data spinouts that had already raised VC funding had done so pre-product. We were coming to market not only with our experience and reputations, but with a functioning prototype. We thought it would be a cake walk.
But ours was to be a different journey. In the end, our experience had far more in common with that of relatively unknown, first-time founders than it did with those of my former colleagues.
And it taught me a lot about how early-stage investors think.
The First Try
Jeff and I started working together in May 2011. At the time, Jeff was wrapping up a stint in Münich leading development of the world’s first automotive integration with the Apple iPhone as head of BMW Connected (a precursor to the Apple CarPlay platform we all know and love today). By October, he had left BMW, returned to the Bay Area and was full time on Glean.
We were focused on creating a business intelligence platform that would enable less-technical users to make analytical decisions without having to rely on corporate data teams. The incumbent BI tools (MicroStrategy, Cognos, Business Objects and even Tableau), were so archaic in their interfaces and bloated with features that anyone without a strong data background struggled to use them. As a result, most organizations had entire teams of analysts that did nothing but create dashboards and reports for other employees. We saw an opportunity to change that.
We captured our vision — BI for Me — in this two-page overview:
We sent the Glean overview to several dozen investors in my network (mostly VCs whom I had previously met plus a handful who had reached out to me). While the term high-velocity fundraising hadn’t yet been coined, that was effectively what we were doing in scheduling so many investor meetings back-to-back. We naively assumed that our round would be instantly competitive and planned to spend only a couple of weeks fundraising.
Every single investor we reached out to took our meeting — the Aster Data halo was strong and all of them were eager to see what we were up to. Here’s the deck we shared with them (sadly, a few of the images in the deck have been lost to time):
With our early (but functional) prototype, a credible, experienced team and the halo of a well-publicized exit, we figured this would be a no-brainer for investors. But the VCs we met with had other ideas.
After three weeks of meetings, it was clear that we weren’t going to raise our round. But there was a silver lining: by meeting with so many VCs in such a short period of time, it was impossible to ignore the consistency of the feedback we were receiving:
We like the two of you as founders.
You’re credible and have the right mix of skills and experience to bring this to market.
The prototype is very promising.
But…
We aren’t convinced there’s a market for this.
Why was our experience so different from those of the other Aster Data spin-outs?
Simple: we were the only one not building a traditional enterprise software company with a well-known, well-understood go-to-market. Our company had a risk that none of those founded by my former colleagues had: market risk.
In 2011, nobody had ever tried product-led growth for a data analytics product. Our collective skills and experience, while credible, didn’t provide any supporting evidence that our market hypothesis was correct.
And to investors, market matters most.
Back to the Drawing Board
While the outcome of our first attempt at fundraising was discouraging, we firmly believed that we were on to something.
We hadn’t been building Glean in a bubble. Not only had I seen the pain point firsthand at many of Aster Data’s customers, we had socialized the concept with dozens of potential users, as well as data team leads and corporate executives. The feedback we received was consistently and overwhelmingly positive (even after taking into account The Mom Test). But we needed to do better.
We needed to get more specific.
“Life might have its failures, but this was not it. The only true failure can come if you quit.” - Great-great-aunt Rose
For the next three months, we were laser focused on three things:
Nailing the initial use case(s) for Glean
Gathering more evidence that there was, in fact, a market for this
Figuring our what specific functionality was necessary for an MVP (and building as much of it as we could)
In the final weeks of 2011, we spoke with hundreds upon hundreds of people. Business and data users at companies like Neiman Marcus, MySpace, LinkedIn, Nike and Ubisoft, executives (aka buyers) and data team leads in organizations large and small, and other experts in the data analytics space.
As we dug in, we began to see a very specific, very acute pain point that many of the organizations we spoke with were facing: business units were increasingly relying on a new wave of “cloud services” like Salesforce, Marketo and SurveyMonkey, but they had no way to analyze the data contained within those services (at least, not without considerable help from their internal data teams).
In those days, there were virtually no integrations between cloud services and on-premise databases (Salesforce being the most notable exception). That left companies wanting to analyze their “cloud data” with only two options:
Download the data in a giant CSV file and try to manually make sense of it
Build a custom connector to each service in order to pull the data into the corporate data warehouse, then build custom reports and dashboards for business users on top of the data warehouse
Both of these options required considerable effort on the part of internal data teams, meaning literally no one was doing it.
That was our wedge.
A More Focused Prototype
The insight hit us like a ton of bricks.
Before Aster Data was acquired, we had started experimenting with deploying our data warehouse in the cloud (in fact, we deployed the world's first production cloud data warehouse on AWS for ShareThis in 2008…it was barely functional). Snowflake hadn’t yet been founded, but the opportunity seemed crystal clear: data was increasingly being generated by cloud-native services. Of course we would need cloud-native data analytics.
With the help of Gail Yui, an experienced graphic designer who would eventually become the third leg of our product development stool, we kicked off 2012 in earnest. We rapidly iterated on our prototype with a goal of highlighting its potential for cloud-native BI.
We already had a fairly advanced machine learning engine that did a reasonable job of automatically identifying and classifying CSV data (that’s right kids, machine learning wasn’t invented by OpenAI 🤣). We identified every single cloud service we thought someone might want to use for data analysis — everything from business services like Salesforce and SurveyMonkey to personal fitness offerings from Fitbit and Nike+ —then downloaded sample CSV files from those services and used them to train our algorithm.
We didn’t build any actual integrations — open APIs weren’t a thing yet and, even if we had access, we didn’t have the engineering resources to build them. Instead, we focused on making our data classification layer as accurate as we could for the CSV files one could download from those services. We reasoned that if Glean could handle the standardized CSV files users could currently download, then we could credibly argue that we could eventually build a direct integration.
Proving a Market
The second thing we needed to do was to prove beyond a shadow of a doubt there was a market for what we were building.
We had a two-pronged approach to this:
Take our new-and-improved prototype back to the potential users, data team leads and execs we had been talking to and get as many of them as we could to agree to take reference calls with potential investors
Reach back out to the experts in the data space we had been talking to and try to get verbal commitments from some of them to angel invest (the SAFE hadn’t yet been invented, so there wasn’t an easy way to raise angel funding in advance of a VC round)
After circling back with the potential users, data team leads and execs, a number of them agreed to take reference calls and/or gave permission for their names to be included in our fundraising materials as potential future customers. These ranged from the President of Neiman Marcus Online to the head of Stanford University’s alumni fundraising organization to countless individual users.
When we showed our updated prototype and explained our vision to experts in the data space, we quickly secured commitments from a number of them to angel invest, including:
Mayank Bawa and Tasso Argyros, co-founders of Aster Data
Dave Kellogg, CEO of Host Analytics (also ex-CEO of Mark Logic and ex-GM of Salesforce)
Jonathan Goldman, Anu Tewary and Mike Greenfield, all previously of LinkedIn’s data team (Jon was the inventor of the “people you may know” algorithm that is used in literally every social media platform today)
(We also subsequently raised from angel investors who weren't in the data space, like Jerry Neumann and David Cohen.)
Fundraising: Take Two
With our narrowed focus, improved prototype, and clear evidence in support of our market hypothesis, it was time for us to try fundraising again.
But there was one more adjustment we planned to make: this time, we cast a wider net.
Our first attempt at fundraising was focused exclusively on VCs who had deep knowledge of the enterprise data space. While those investors immediately understood what we were trying to build, we were struct by how many of them reacted to our go-to-market hypothesis with some version of “that’s never going to work” (a far stronger stance than simply “I’m not convinced”). So, rather than just go back to the VCs who had previously rejected us, we took time to fill our fundraising funnel.
We built a target list that included more than 50 additional VCs that were active early-stage investors in B2B software companies but didn’t necessarily have a prior investment in a data analytics company. Then, we figured out how to get introductions to each and every one of them (reasoning that the Aster Data exit might not have been on their radar). The list of connectors included Jud Valeski, the cofounder of a young startup named Gnip who had attended the same high school as Jeff.
Over the coming weeks, we met with dozens and dozens of VCs. This time around, the majority of them were far more interested in what we were building and the early proof points we came armed with. We progressed into multiple meetings with a large number of firms, but ultimately, the majority of the investors we spoke with landed on the same conclusion as six month’s prior:
We aren’t convinced there’s a market for this.
While our first experience with fundraising was very much an exercise in our own naivety, our second attempt showed me for the first time how few early-stage VCs are actually willing to make non-consensus investments.
More than half of the investors we met with outright rejected the evidence we presented about a fundamental market shift. Unmistakeable evidence that data generation was moving to the cloud, large enterprises were struggling with this shift and cloud-native data analytics platforms would need to follow (I have no doubt that many of these same investors passed on Snowflake).
Trae Stephens of Founders Fund recently wrote about this phenomenon
But this isn’t about sour grapes. This is about fundraising being a numbers game.
Had we stuck to the “easy intros” — the VCs who were familiar with Aster Data and were looking to invest in Aster Data alumni founding another Aster Data — we might never have raised funding. Ultimately, we found success with investors whose view of the future of data analytics wasn’t colored by the industry’s recent past.
Which brings us back to Jeff’s high school friend, Jud. Jud introduced us to Ryan McIntyre of Foundry Group, a relatively new firm based in Boulder, Colorado. Ryan was a deeply technical investor who had firsthand experience with paradigm shifts, having cofounded dot-com pioneer, Excite. Unlike many of the other VCs we had pitched, Ryan had limited experience as an investor in data analytics platforms. As a result, his diligence was focused less on the recent history of the data analytics industry and more on the potential of the shift to the cloud.
Over the coming weeks, we met with Ryan multiple times in person and had calls with his partners in Boulder. We were excited about how they thought about our opportunity and Foundry quickly rose to the top of our list. All of which ultimately led to the moment of truth.
And when the dust settled, we successfully closed a $1M pre-seed round led by Foundry Group.
Epilogue: What About the Name?
You might be wondering why we gave up the name Glean and introduced ourselves to the world as DataHero.
This was another lesson for me: sometimes, your lawyers are wrong.
Jeff and I absolutely loved the name Glean. We thought it was far and away the best name for a data analytics company — especially one trying to bring data analytics to business users. We were so committed to the name that we had already started discussions about securing the domain glean.com while we were fundraising. But then we met with our lawyers.
Like many startups, we worked with one of the big 3 Silicon Valley firms. And our counsel was adamant that we should not under any circumstances brand the company as Glean. “You’ll never be able to get the trademark,” they said. “You won’t be able to protect the name.”
While they may have been right from a legal standpoint, it’s pretty apparent these days that trademarks aren’t everything. But we capitulated and abandoned our original name. And while the name DataHero served us well, I always lamented the fact that we gave up on Glean.
So it made me absolutely giddy when, 10 years later, a new startup called Glean came out of stealth to democratize data insights.
