As it turns out, nearly every startup we talk to feels like they could be doing better with turning data into insight, and therefore no one has been more popular among the portfolio companies than William, the data scientist that joined Creandum to help both the firm and our portfolio companies with data. William was part of the team that built up the core analytics at Spotify, and joined Creandum beginning of the year.

With so many new team members the Creandum office is starting to get cramped, and I was one of the lucky girls that ended up sharing a desk with the data guru. After listening in on his calls and asking stupid questions during the last quarter has given me a few epiphanies that I wanted to share.

1. Data that doesn’t say anything still tells you something
Many A/B tests come back negative or neutral and it’s easy to feel like you redoing the A/B testing in vain after coming back so many times without results. But by doing A/B testing you are eliminating the unknowns. You now know that there is nothing to gain by moving the ‘sign in’ button, or that you’re already doing the right thing by letting people see the price of shipping before checkout.

Instead of thinking the negative A/B testing result gave you no increase in conversion, you should see it as a gain equal to the delta between what you are doing now and how much worse you could do (And you now know you shouldn’t do it because you A/B tested).

2. Don’t get too impressed when someone is using the phrase “machine learning”:
With my own programming career ending somewhere at building a sudoku games with javascript almost a decade ago, I am sometimes easy to impress with long techy descriptions of how the product works. Machine learning is a common word in the meeting rooms here in the Creandum office, and while the investment team is nodding their collective heads, William is rolling his eyes (he is actually too nice to roll his eyes but if he would ever roll them, he would at that point).

It turns out machine learning is often very simple ideas, scaled up by being ran on the data set 100k times, making the algorithm therefore being tweaked and improved. So now when I hear “We apply machine learning to our vast dataset to to find the most relevant videos” I know it probably means “We have sorted the videos by number of views to get to the most popular one.”

3. Being right staffed when it comes to data can look very different
When trying to get better at data, a data scientist is the buzzword to hire. So how do you find the right one for you? Not every small startup can hire a senior data scientist like William! (Which is why we got him onboard for our portfolio companies).

Luckily it’s more about mindset – being on top of data doesn’t always require a senior data scientist. If your data is relatively clean, a junior hire fresh out of school can probably provide the answers you’re looking for.

And you may not even need a data scientist. An concept Tictail is uses is that instead of having a dedicated data team, is to have very data driven and data knowledgeable people in the product team. With every product decision coming from a data mindset you’re starting with the right pieces in place.

The point is, there are many ways to get better at data. All companies we speak to past an A round said that data is a big priority, and nearly all of them admits that they are struggling. Without slowing you down, getting it structured now rather than later is probably a good favor to your future self.