To Build Value in Tech, Build Different
10 min read
Forget better, faster and cheaper. If you’re a technology CEO trying to build real value, my advice is to do something different.
Yes, I realize this sounds like the famous “think different” Apple advertising campaign from more than 20 years ago. But I think the core adage remains true. Many new technologies, once you get past the marketing, are really just improved versions of old technologies. Each generation of Intel CPUs is faster than the last (sometimes because of a better architecture, sometimes because of an improved manufacturing process); Snowflake is a data warehouse that works in the cloud, rather than on-premises; and Dell really just created a cheaper way to buy a personal computer.
By contrast, companies like MongoDB—which reported stronger-than-expected financial results in early June and saw its market value rise significantly—is, in my mind, building something very novel. The company’s founders, Dwight Merriman and Eliot Horowitz, had an idea early on for a truly different type of database, one that was very natural and intuitive for developers and able to scale out on cheap hardware and work well in agile-development processes. In pursuit of that goal, they built a database to support “documents” (which allowed for heterogeneity) rather than tables (which were designed around regularity) but traded off many other features and capabilities.
The downside of different
Doing something differently is very different from doing it better. When you do something differently, your product can be worse for some people and better for others. Often opinions of “better” or “worse” are subjective, and users will be divided about which choice they prefer.
Look at the automobile industry. Tesla, another company that has forged its own, unique product path, is currently valued at multiples more than GM, Ford and Stellantis (Fiat+Chrysler plus some others) combined. My wife and I have owned multiple Teslas over the years and loved them. They have advantages and disadvantages relative to gas-powered vehicles; if we didn’t care about reducing our carbon footprint—or worried more about reducing our lithium footprint! —the choice between gas cars and EVs would be less clear.
Among electric vehicles, Ford builds a very good electric pickup truck, while Tesla has yet to deliver (and its truck is controversially styled). Is the Model Y “better” than the Mustang Mach E? Motor Trend thinks it is—but mostly because it costs less and had (at the time of the review) a broader network of charging stations, but Car and Driver gives Ford the nod.
In the end, this debate doesn’t really matter. Fundamentally, Tesla succeeded at disrupting the auto industry because it didn’t try to be the same—it tried to do something new. And “new” was somewhat controversial: The company built a car that was heavier and couldn’t deliver full power for a sustained period without overheating; it also took away most of the existing in-car “user interface” (buttons and switches), which certainly raised eyebrows. What’s more, Tesla enabled (and required!) the driver to do nearly everything through a tablet mounted on the dashboard. While other luxury and near-luxury cars are adding heads-up displays, so drivers don’t have to glance even a bit downwards at the dashboard, Tesla built cars with no instruments in front of the driver.
On the other hand, Tesla also built cars that don’t use gas, that can (in popular perception, if not quite fully in fact) drive themselves and that can accelerate from zero to 60 miles per hour in under two seconds. The company built cars with doors that open up, with a yoke instead of a steering wheel (for a while, though the company eventually put the steering wheel back), and that make (seriously) seven kinds of fart noises.
As it turns out, the maker of farting cars without steering wheels or instrument panels was the first new U.S. auto brand to create value comparable to (and in this case, an order of magnitude more than) the major manufacturers since 1941, when Jeep was founded. Not that haven’t been dozens of attempts: DeLorean, Vector, American Motors and Saturn are some interesting recent examples.
Technology gets different
What about technology? As it turns out, the tech industry, of which MongoDB is a part, obviously, has minted lots of companies that started out with profound innovations. Not all of them stayed truly different and panned out, but I think it’s instructive to look at all the U.S. technology companies currently valued over $500 billion:
- Apple has been all about different for many decades. Mac fans may think Apple’s products are “better” but really, it’s mostly an aesthetic choice. I use an iPhone and I am typing this on my Macbook air, but I can’t say they are objectively better than Windows-based laptops, and they are definitely more expensive.
- Microsoft grew rapidly by building an operating system for personal computers. They had competitors, but mostly they grew because that little niche of different–much smaller and cheaper, but also much less capable–computers grew.
- Google. Ironically, despite feeling like a “different” kind of company in its early days, with founders in control via super-voting stock and telling shareholders they didn’t intend to be profitable–Google may be the closest on the list to a company that won by being better: better at search and better at selling search advertisements.
- Amazon started selling books on the Internet. At the time, that seemed like a dumb way to buy books! It took days for them to arrive, you paid shipping fees and you couldn’t leaf through the pages to see if you liked it. But you didn’t have to drive there, you didn’t have to talk to anyone about your purchase and they had a very deep catalog of books, including some that were unlikely to be in your local store. Amazon offered a different way to buy books with pros and cons.
- NVIDIA wasn’t a better CPU, it was a specialized processor for graphics, which was an interesting niche and different from what Intel was doing. Then AI happened, and it turns out the different processors for graphics were better at that… and since 2019, NVIDIA’s market cap grew from ~$140B to ~$1T, while Intel’s fell from ~$250B to ~$130B. NVIDIA set out to be different and wound up better at something new.
- Tesla, which we discussed earlier.
- Facebook enabled people to communicate online in new and different ways. You might blame them for the decline of civil discourse and the erosion of the value of objective truth, or credit them with freedom in the Middle East. More modestly, you could say they allowed people to share interests and build human connections at a scale never before known or share boring kid and pet pics or unrealistically curated and highly processed photos of their life, but they clearly enabled something that wasn’t happening before.
I would score it as six of the seven tech companies with the highest current public market valuations doing something different, and one doing something better.
Back to data
Two of my defining professional experiences after spending 9 years at Oracle were with tech companies that tried, with varying degrees of success, to build and sell a different kind of database. MarkLogic had an Extensible Markup Language (XML) database that you queried in XQuery. When I first met the team in 2003, “NoSQL” wasn’t a thing, and most prospects’ reaction to the idea of using a database that wasn’t relational and wasn’t SQL was only slightly more favorable than if I had walked into the room and suggested cannibalism. But there were a few, mostly publishers who had lots of Standard Generalized Markup Language (SGML) and some XML that thought it would help them with some very specific problems. And while MarkLogic’s technology didn’t really pan out, the company’s motivation to do something different was well-advised.
Seven years later, we saw the proof: Google had Bigtable, a repository that allowed users to store terabytes or even petabytes of data; Facebook had built its Cassandra distributed storage system; and Yahoo had built its powerful open-source Hadoop storage system. Suddenly, the world was more open to a different kind of database. When I joined MongoDB in 2011, the company had a database that didn’t do multi-document transactions or “joins” (combining data from two or more database tables; this, along with transactions, are two of the most powerful features of a relational database that have been fundamental to its success) and wasn’t nearly as reliable or secure as the more mature offerings on the market.
But the product was very natural and intuitive for developers working with JSON and Javascript, and, at least in principle, scaled out on cheap hardware and worked well for agile development and evolving schemas. I spent four years there as the product matured significantly and we began to build the go-to-market strategy. MongoDB’s growth since then is well-documented; to me, it shows that “different” is a pretty good way to build enduring value.
The two most valuable new entrants in the database market right now are Snowflake and Databricks*.
Snowflake pioneered the concept of a cloud data warehouse; essentially, this allowed customers to perform the types of analytical queries they had performed with technologies like Teradata in their own data centers, but in a cloud-based system. With the strong desire of many IT organizations to move to the cloud for many reasons (scalability, cost and simplicity all high on the list), Snowflake grew rapidly even at scale and became the highest-valued software IPO in history.
Since IPO, the company grew to a peak valuation of over $100B, and while it is down from that peak, it is, at this writing, still worth over $50 billion. To enable the transition of the data warehouse to the cloud, Snowflake had to develop complex technology to separate the compute responsibilities of the database from the data management responsibilities and make those technologies work at a scale that they could support the complex requirements of hundreds of large enterprises.
Meanwhile, Databricks took a very different path to creating a similarly ($38 billion in the most-recent funding round) valuable data company in the cloud. They began by commercializing Spark, a distributed-processing framework that allowed users to more easily write programs to take advantage of large clusters to rapidly and reliably process large sets of data. It wasn’t exactly a database or an ETL tool—it was something new and different—but it seemed to deliver on some of the expectations that Hadoop had set, and it was growing quickly.
It turns out this architecture enables customers to implement “Lakehouses” that combine some of the best capabilities of data warehouses and data lakes. Meanwhile, Databricks ignored the traditional approach of selling an enterprise version of their product on premises–despite strong demand from attractive customers–and focused 100% on building a great cloud offering.
Snowflake has done a great job of making the established technology of a data warehouse, for which there was already a large market, available in the cloud. Databricks has instead developed a new technology to meet the needs of emerging data science and AI workloads. Snowflake has a little more revenue and a higher valuation right now; we will see over the coming decade which company proves to be more important to the industry and builds greater enduring value for shareholders.
As excited as I am about the potential of new and different technologies like MongoDB and Databricks to build the data infrastructure of the future and create broader market value, I would caution entrepreneurs that the path of building and selling something different isn’t easy. In the early days of a company, your product is immature. Early adopters understand its strengths and weaknesses and how it helps them, but mainstream users are often confused about if, how and where to apply the new technology.
In retrospect, the markets and use cases to focus on look clear, but prospectively, it’s hard to sort out where to focus, whom to sell to and how to describe your product. You will inevitably confuse some users and disappoint others.
It takes hard work, persistence and a good bit of luck to eventually reach the level of success of a MongoDB or a Databricks, never mind a Tesla or an Apple. If you are in it to make a quick or easy buck, I recommend doing something else—just about anything else. You might, like I did at MarkLogic, spend 7+ years of passionate hard work in a market that never quite achieves liftoff.
Or, you could have the right, innovative idea at exactly the right time—and everything else will fall into place.