Subscribe to DSC Newsletter

The Evolution of a Data-Driven Startup

People often talk about a startup’s evolution in terms of funding, sales, or product engagement milestones. But companies also change over time in the ways that they interact with data.

Increasingly, the companies that achieve massive growth are anticipating future data needs and progressing quickly through four distinct phases of data.

Whether you’re an early-stage founder and or an analyst hired after this evolution has already begun, take time to identify where you are in theses phases. Setting up the right team, tools, and processes during each phase can help your entire team more effectively use data to achieve other outward-facing milestones.

Phase 0: DATA?! But we haven’t even built anything yet.

Before you launch your product, the majority of your data is qualitative. As you talk to potential users, you’ll (hopefully) start to see the feedback center around specific topics.

These qualitative data points—the problems for which you are creating a painkiller—are likely the inputs that will inform your initial data strategy.

Once people actually start using your product, their feedback can lead product development in unexpected directions. You’ll want data to validate this feedback, so make sure your product is instrumented in a way that will allow you to retroactively measure new things. Build robust data tracking into your culture. Encourage engineers to make detailed event tracking an integral part of the development process from day one. The longer you wait, the harder it becomes.

Before we launched Mode, we made a decision to classify “Run” events—query executions against a database—into a few categories: runs against private databases, runs against the public database, and runs from the report view. Analysis of this granular event data, corroborated by feedback, helped us identify two strong user types: analysts who write queries and business folks who view them. As a result, we’ve been able to better understand in-product behavior and build features to engage each group.

Takeaways

  1. Use early, qualitative customer insights to inform your quantitative data strategy.
  2. Track events as if your key metrics will change. That means tracking everything with as much granularity as possible.

Phase 1: What’s happening?

After your product launches, your first challenge is to figure out what’s happening. Questions in this phase often sound like:

  • How many people signed up today? Where’d they come from?
  • Are people coming back? How often?

With out-of-the-box reporting products like Google Analytics or Mixpanel, you’ll be able to get to a lot of answers quickly. Qualitative information from ongoing customer conversations will still guide many of your decisions—and simple measures of engagement like DAUs are enough to ensure you’re headed in the right direction. Be careful not to over-engineer metrics just yet—at this point, agility matters, and you can’t afford analysis paralysis.

As you progress through this phase, you’ll likely start to measure actions that are unique to your product. In building these dashboards, try centering them aroundforward-looking metrics. Picking the right metrics can be tough. It can be tempting to focus on revenue, for example, but it’s a trailing indicator and won’t be helpful when making decisions about what to build next (some better examples here).

Once built, get these dashboards in front of the whole company on a regular basis. When health-of-the-business questions crop up, answer them with context anddata. By talking about metrics early and often, you’ll begin to cultivate a culture of data literacy and establish credibility for data (and the people working with it). As the team becomes accustomed to seeing data, they’ll recognize patterns, more quickly notice anomalies, and navigate decisions substantiated with data.

Takeaways

  1. As you take stock of what’s happening, leverage ad-hoc analysis of raw data to develop core metrics that fit your business.
  2. Establish a culture of data literacy by talking transparently about metrics with the whole team early and often.

Phase 2: Wait, but why?

So, once you’ve nailed the “whats,”—you will start hearing more “whys:”

  • Why do customers who invite friends stick around longer?
  • Why do customers interact with feature X but not feature Y?
  • Why did growth suddenly evaporate last Wednesday?

The tools that helped you measure standard metrics aren’t flexible enough to answer this new class of questions—they’re built to be broadly applicable rather than tailored to your business. If you’ve built custom dashboards, those are often a good starting place for answering the “whys.” You can use the methodologies that drive the dashboards as a jumping off place to answer these tougher questions.

The questions to be answered are often unique and involve bespoke research—we see companies at this point really dive into their databases and use SQL to answer these questions. Depending on the state the data is in, we see many companies start building out their analytics teams with data engineers or analysts.

Lots of ad-hoc work can get tedious and block high value projects, so it’s important to begin automating solutions to common problems and democratizing access to the information. Folks will be able to self-serve some answers on their own—the “whats” are still important but there’s not much reason for analysts to be spending their time reactively pulling data.

Takeaways

  1. When your data questions increasingly start with “why” it’s likely time tohire an analytics team.
  2. Make your analytics team as effective as possible by automating answers to “what” and “how” questions—and give everyone in your company access to the information.

Phase 3: Where can we go next?

Often, this phase sneaks up quietly. The nuanced difference between the “why” stage and this one is proactive discovery.

To move into stage three, you’ll need to make sure a few things are in place:

  1. Routine reporting that does not require human intervention.
  2. When reporting problems do arise, the technical and organizational infrastructure is in place to respond quickly.
  3. Because your infrastructure is solid, reporting problems are rare in the first place.

With these achievements unlocked, reverance for data pervades company culture. Decision makers are empowered to make data-driven decisions independently and analysts can spend more time on forward-looking, high-leverage projects. This is the sort of thing happening at Uber, Facebook, and other companies with strong data culture.

In phase three, you also start to see your time horizon increase. You can start to dedicate parts of your analytics team to focusing on product decisions that affect work half a year out.

As an example, Google has run experiments on page load times to understand how site performance impacts behavior. They’ve used this information to make better decisions about allocating resources towards site performance improvements. Intentionally making your site worse may sound crazy, but it’s incredibly efficient in identifying the opportunity for improvement. For companies who have solved most recurring problems and are advanced enough to pull this off, the dividends can be tremendous.

Takeaways

  1. Enabling everyone throughout your company to use data effectively canfree up analysts to work proactively on high-value projects.

If you can’t start from scratch…

Not everyone has the benefits of building data infrastructure and an analytics team within a company from day one. If you’re jumping into a company mid-evolution start by assessing where your company lies between these phases. With a complete view into how your company currently works with data, you’ll be able kick off meaningful projects—be it building the right team, developing stronger tracking and warehousing practices, or implementing new data tools. You’ll empower everyone to use more data more effectively, allowing your whole team to achieve faster growth and dominate your market.

This post originally appeared on the Mode Blog.

Views: 1693

Comment

You need to be a member of Data Science Central to add comments!

Join Data Science Central

Follow Us

Videos

  • Add Videos
  • View All

Resources

© 2017   Data Science Central   Powered by

Badges  |  Report an Issue  |  Privacy Policy  |  Terms of Service