5 Challenges on the Way to the Data-Driven Peak

You may have heard people talking about data-driven "something". And maybe you envied them as you also want to be data-driven. At your company, you may have even tried to be data-driven. It just somehow didn't work out as expected. Data relationship started out so sweet. With honeymoon when you started to use Segment, bought some data lake solution, and had those meetings with excited people. But it turned into a mess. Why? Why? Why?

Don't be depressed; you are not alone.

At Dripit we work hard to turn raw data into insights. Doing that, we have found again and again that there are five crucial steps behind every insightful piece of data. Each of these steps is also a great opportunity to fail. And we have certainly had failures at all of them, however, this has allowed us to truly understand the whole data-driven process.

It is a bit like a journey to the top of a mountain. You can see the mountain in the distance, and it is beautiful, but you don't see the dangers that lie between you and the peak.

It’s always further than it looks. It’s always taller than it looks. And it’s always harder than it looks.
– The 3 rules of mountaineering

1. Choosing the mountain to climb

You have to come up with a question which, if answered, will lead to a change in the way you do something important. Yes, this is the exact definition of qualitative KPI (key performance indicator). Knowing how many unique visitors per day you had is cute! Knowing which of your acquisition channels worked to generate conversions which in the next six month turned into a high CLV and a good upsell opportunity, is valuable!

2. Mapping the road

What kind of quantitative data do you need to answer this question?
Might seem easy but it is not. Sometimes there is no data you can just grab from your data universe, lake, swamp, or whatever fancy name you have for your big data storage (because no one has small data anymore, right?). A lot of time you will have to think on the meta level. What kind of data you should get to derive the information you are ACTUALLY looking for.
In the process, you may realize that the question itself must be changed, and you would be back to square one.

3. Assembling the team and gear

Great, you have figured out the qualitative question and quantitative data you need. Now the easy part - getting the data. Yeah, about that! One piece might come from CRM, another - from tracking data or a spreadsheet. You have to combine this information, stick it together, and normalize! At this point you may introduce one more thing - other people. They have their bright side - they know what you don't know, and they can help. But they also have their dark side - their lives, their challenges, and their priorities you have to fight with.

4. Journey to the mountain

Amazing, step 4! Question - check, data - check. Database Garry from IT department and his Mysql/JS/GA skills - check. We are nearly there. Well, not exactly. Now we have the data. And most of the time this is where things will get stuck. There is data, and answering simple questions is easy. But those hard ones are, well, hard!
You have to "play" with the data, "make friends" with the data. This requires certain tools and ability to use them. Graphical UI tools may be limited in functionality, or they may be too complicated to use. Picking some programming language like R or Python (Pandas library in particular) is time-consuming.
And again, you may have to work with someone else as this part is out of Garry's scope. You need someone who can discuss business side. One more piece of the human puzzle in your everyday work.

5. Last 100m

Bam, you got some results. Time to deliver it. Different stakeholders might need those results. For each of them there might be a different "Aha!" moment (i.e., the action behind that figure). Someone in a higher position in a company needs to see the big picture. Others might want to access a bit more detailed view, or ability to slice and dice the data. Without this one, all the previous steps are useless. And again, there are times when there is a great result, but it is not presented in a usable manner.

Now you have it, your framework for data-driven insights. A roadmap and red flags to be aware of.
What are your thoughts? What is your experience with climbing that data mountain? Leave a comment below.