How to make data-driven decisions in your startup

Collecting and analyzing data to inform decisions is key to any startup’s growth. But in a world of unlimited data, failing to choose the right metrics to measure actually inhibits high-quality decision-making.

This is a challenge that Dillon Mullaney, VP of Revenue at Mozart Data, sees regularly with clients. For him, the solution is to first envision the dashboard your business needs, which in turn informs the specific data points you need to collect. Speaking with BetaKit, Mullaney explained how he builds high-quality dashboards that drive specific, relevant action. 

Begin with the end in mind

At a foundational level, Mullaney said you only need to understand three things about your business to make data-driven decisions: how your product functions, how customers use your product, and how the company makes money. But a near-unlimited number of metrics could be used to measure these outcomes; unfortunately, most won’t be useful and even useful metrics can deliver the wrong insight if the data quality is low.

So, in a world where you could measure almost anything, how do you know where to start?

Mullaney’s advice is to work backwards. Rather than starting with a problem and measuring everything you see, Mullaney said to start with the perfect end state, then ask what data would help you measure if you’re accomplishing it.

“What would the end dashboard be that we would want to create that would help answer that question for us moving forward?” said Mullaney. “And then from there figuring out, okay, ‘What different pieces of data are we going to need in order to build out this chart?’”

A dashboard for your thoughts

Building a dashboard from end-state brainstorming applies to most startup growth challenges. Mullaney gave the example of a marketing campaign and the common question: which paid ads are driving the most signups and the most revenue? The end state is a dashboard that shows not just campaign sources (e.g. Facebook or Google) but also downstream conversions throughout your product sign-up process; it should also tie those results to a customer with a revenue number.

“I know I’m going to need data from the various ad platforms that we’re using,” said Mullaney. “I’m going to need data from our CRM or a tool like Stripe to know ‘how much are we charging our customers?’ Then we’re going to need product data that’s most likely sitting in our database.”

This is the power of knowing what you need upfront: once you have outlined the necessary data sources to understand your end-state, you can then ask what specific pieces of data will help you confirm if you’ve achieved that goal.

A crucial part of data collection is connecting disparate data sources to avoid gaps in knowledge when trying to drive a specific action. Consider, for example, a SaaS platform allowing users to sign up for a free trial, with a sales rep then tasked with reaching out to convert free users to paid plans. That sales rep should have easy access to the company’s CRM data, containing info such as when a user signed up for the platform. However, in organizations that don’t connect data sources to a dashboard, the sales rep might not have access to the product database that tracks user behaviour on the platform, meaning they are entering sales calls with only half the information.

“I’m putting myself at a disadvantage by going into these calls into a black box, because all I know is that they signed up and they’re using the product,” said Mullaney. “But had I known that they’re using the product in X, Y and Z ways, it’s going to allow me to be much more efficient on my calls.”

Capture everything, verify often

Even with a perfect dashboard, your startup will still have to take action based on what the data tells you. In the ad campaign example above, one platform might reveal itself as driving a ton of sign-ups but very little revenue. This could be a user adoption problem, an ad campaign source problem, a messaging problem, or something else entirely. You won’t know until you dig further and run experiments to isolate the cause.

Because you might need to dig deeper at any moment, Mullaney advises startups to capture all available data beyond the core pieces you’ve identified through end-state brainstorming. Even if you don’t use it today, it might become a key input for the future. 

This is what happened internally at Mozart Data. Mullaney said the company has multiple payment packages that a customer can get started with. Initially, the plan information for each customer was not being brought into Salesforce and was siloed within each individual contract. When it came time to build out a usage and billing dashboard for the account management team it was virtually impossible because of the siloed data, forcing Mozart Data to manually update each customer’s account. Now that the team has better insights into each customer it allows them to be more proactive with customer outreach.

​​“Another thing is setting up whatever your CRM might be correctly from the beginning and over-capturing data because you never know six months down the road what’s going to be a valuable piece of data for your team,” added Mullaney.

Mullaney also cautioned startup leaders to remember two things: verify data accuracy and check the data frequently. Even after data collection is automated, startup leaders should continue to work closely with the data analysts or technical operators who built the dashboards to ensure data integrity remains high. He added that the data needs to be as recent as possible.

“If we know that a campaign is going well, but we don’t find that out until a month later, that’s a huge missed opportunity if we had that data in real-time,” said Mullaney.

In making these recommendations, Mullaney is aware that young startups have severely limited resources and time. The challenge of knowing what data to measure and how to do it well is why startups need to get very good at what Mozart Data CEO Peter Fishman calls “savvy counting.” 

“We’re not saying that you need to hire a data team from the beginning,” said Mullaney. “There are a lot of great tools out there like Mozart Data that allow you to increase your data maturity early on without having to hire additional resources.”