As a data consultant who has worked with several startups, I have observed that the early priorities of these companies do not compel them to hire data expertise until it’s too late. 

On one hand, not making data a priority is understandable. They are busy validating product and market fit, finalizing features, and acquiring customers. They tend to hire product managers, marketers, and developers rather than data scientists. 

Then, some data needs naturally emerge, especially when the company gets to the point of product optimization or even acquisition. The CEO or CFO will start asking for data and analytics and the existing staff, most likely marketers or engineers, will do their best to produce it for them, even if that’s not their main priority.  

Even if these requests are addressed by the world’s best marketers and engineers, they can’t transform themselves into business intelligence people. They’re likely unaware of best practices to put together a data MVP (minimum viable product). They don’t know what the risks are if they don’t follow them. 

Typically, the resulting solutions are what I call scrappy, built with the best intentions and skills but not a BI mindset. For example, I’ve seen many companies use noSQL databases to store their data, hence building data warehouses on MongoDB or Elasticsearch. That’s convenient at a small scale, but as your company expands from 100 to 1,000 customers these tools cannot support enterprise data. Data aggregations that used to take five seconds might now take five minutes, 15 minutes, 50 minutes, then simply collapse. At that point there is only one option left: scrap the data infrastructure and start over. 

MongoDB or Elasticsearch are great tools but they’re built to retrieve and serve a huge number of concurrent requests in a very fast and scalable way. They are not, however, built for data warehousing at scale.

Three immediate data priorities  
With these realities in mind, here are a few practices any fast-growing small company should consider when it comes to their data.

  1. Build an MVP of your data solution.
    By definition, this MVP is going to be something small, quick, and simple. Yet it has to be built with the appropriate best practices, architecture, and expertise. You should deploy the right technology and the right data modeling. And it all has to make sense from a business perspective, of course. 
     
  2. Leverage the right talent.
    One option is to go to a consultancy, have them build an MVP of a data warehouse, and advise on your choice of BI solution. If you follow this path, expect an engagement of about three months, as well as follow-ups post-launch. If you are reluctant to use an external company based on fear of losing autonomy, you could include a structured end-of-project handover and training in whatever the consultant builds for you. An experienced consultant also can help you to locate and interview freelance or full-time data staff for your team.

    A second talent option is to hire an external consultant to advise your team as they build your data MVP. More often than not, a good in-house engineer is perfectly capable of building a data warehouse. It’s not really rocket science, but a consultant can save everyone from making costly mistakes by making sound recommendations.

    Talent option number three is to hire a data leader of sufficient seniority and skills to build the foundations of your solution. Ideally, this person is at the level of a senior BI or analytics engineer. One good thing about hiring sooner rather than later is that this person will join your startup with a long-term perspective, can make good future hires, and can scale the function along with the company. 
     
  3. Create excitement about data. 
    Whenever I join a new company I consider it my Number One responsibility to drive a data culture. That means creating excitement around data and showing people the value of what data can do. In practical terms, it’s also about finding ways to show them that it’s not really that difficult. Being able to understand the most basic KPIs to track company performance delivers a huge amount of value and does not require much of an investment.

    Once you have people’s interest, your next step is to educate these people in all things data and ultimately to drive adoption. Their level of eagerness will depend on the business specifics. In some companies where I’ve worked,  people were absolutely eager to consume data at any point in time. In others, I would show a chart and it felt like everybody wanted to run screaming out the room.  

The End of the Unicorn
One data talent trend becoming far less common is the expectation that a single data scientist will allow you to run your entire data function and answer every data-related question. Ten years ago these data unicorns might have existed, but today it’s unrealistic to expect one person to be proficient across the board. Think of data as a team effort. I would say a minimum viable data function includes three people: 

  • A data engineer to build the data warehouse and prep the data using a BI solution
  • A data scientist to interpret and identify value and visualize complex data you produce
  • A data analyst to communicate the financial and strategy implications of the data to the business, and help it make data-informed decisions

In a fully operational company, you should expect that ten percent of your employees have data roles. 

Removing competitive disadvantage
I believe we are past the point where leveraging data in your company delivers stronger competitive advantage. Data has become so central to organizational success that it’s more about removing a serious disadvantage. 

As a company leader, if you want to keep up with your competitors, if you want to win, you must come to a point in which your company decisions are data driven. If that’s not enough to convince you to get your data priorities in order, funding probably will. As small companies go out for their first round of fundraising, any potential investor today asks for data and more data. That’s when people become aware that data is important, and should be a priority from the beginning.