How startups can avoid the “Frankenstein data stack” trap

data
Learn to count the data you have or risk creating a monster.

It’s not always about ‘big’ data. In the early stages, startups need to collect data quickly to make informed decisions in a fast-paced environment.

All companies that achieve any sort of data scale require a cloud data warehouse.

However, in the quest for speed, many startup founders cobble together multiple low-quality data sources, hoping it offers practical insight. This is a problem Peter Fishman, co-founder and CEO of Mozart Data, frequently saw at fast-growing startups in the Bay Area.

Speaking with BetaKit, Fishman explained more about how startups fall into what he calls a “Frankenstein data stack” and why it’s so bad.

Learn to count before predicting the future

From a data collection perspective, Fishman takes issue with the startup adage to “skate where the puck is going,” a famous Walter Gretzky quote often mistakenly attributed to his son, Wayne. To him, this advice implies that early-stage startups need to become predictive machines quickly, using data to understand where the metaphorical puck is going so leaders can pivot quickly. Unfortunately, said Fishman, this approach is a trap.

When a startup is obsessed with prediction, they go into hyper-data collection mode. However, as an early-stage company, they often lack internal data. What Fishman sees founders doing, then, is one of two things: look entirely outward, seeking data inputs from investors, other startups in their community, or even friends and family to help understand their businesses; or they try to create advanced modelling to solve complex problems their startup might face one day.

“They often try to overcomplicate things,” said Fishman.

He added that the result of both approaches is a “Frankenstein data stack” that tries to solve complex problems before a startup can effectively count its data.

Instead of going for complexity, Fishman recommends founders “learn to count” first, offering a sports metaphor of his own to combat Gretzky: Timbits Soccer, the youth soccer league sponsored by Tim Hortons across Canada. In youth soccer, the kids know they need to get the ball in the net but aren’t yet skilled at making that happen. So instead of running complex plays, the entire team moves toward the ball and pivots when the ball gets kicked in a different direction, moving like one big toddler blob.

While this may seem like chaos, and it is, the chaos contains an important startup lesson: learn to identify the correct thing to chase, and then develop a mechanism for chasing it.

“Counting things really well is the first step,” said Fishman.

From regular counting to savvy counting

Fishman said that counting well, at the early stages, is only about three things: revenue, users (or customers), and user activity. Once you have that tracked properly and automatically, you can “count.”

The next step is to count in a more “savvy” way, which Fishman described as “something that slices and dices the world in a way that’s much more actionable.” This involves more in-depth data such as journey mapping, time- or campaign-based cohort analysis, and acquisition channel attribution.

RELATED: How startups can survive a market downturn

These more detailed analytics provide actionable insights you might be able to do something about: for example, if you realize that everyone from your spring campaign on Facebook is churning, but the Google campaign produced customers who renew, you can turn off Facebook ads, try something new, or try to mimic the Google campaign insights to get similar results.

The real challenge of ‘savvy counting’ is that startup leaders often have dozens of different data sources, resulting in multiple silos. Fishman said it takes three steps to bring these silos back together without hindering your team’s ability to perform:

1. Set up a data warehouse: a third-party software that integrates with all possible data sources via API, bringing everything into one place.
2. Clean up data: remove duplicates or manually verify discrepancies so you can trust the data in the warehouse.
3. Connect your data warehouse to a business intelligence tool: create automated reports that provide those “savvy” insights.

“All companies that achieve any sort of data scale require a cloud data warehouse,” said Fishman.

Don’t be like Dr. Frankenstein

One of Dr. Frankenstein’s primary faults was not his brilliance, but the fact that he didn’t think through the implications of his creation. For early-stage startups unwilling to consider the implications of their decisions, they may end up creating a similar monster.

“The rich insights—the ones that that sort of businesses care about, that growth teams care about, that marketers care about, that RevOps teams care about the ones—that really are meaningful and change the direction of your business often do come from bringing together a bunch of pieces of data to generate something more nuanced,” said Fishman.


Learn more about making the most of your data here.


Photo courtesy of Unsplash.

Stefan Palios

Stefan Palios

Stefan is a Nova Scotia-based entrepreneur and writer passionate about the people behind tech. He's interviewed over 100 entrepreneurs on topics like management, scaling, diversity and inclusion, and sharing their personal stories. Follow him on Twitter @stefanpalios or send an email to stefan.palios@gmail.com.

0 replies on “How startups can avoid the “Frankenstein data stack” trap”