Five mistakes to do with data– How to become data-driven
We talk a lot about how we are a data-driven company and how awesome it is that we do “big data”, but in fact, we should stop saying “big data”. We need to say we have the right data! The difference between the two is the difference between flying an airplane or crashing it while trying to take off.
When it comes to big data, we have developed some bad habits and they are not allowing us to use the right data.
We use excel to connect data sources
YOU DO NOT USE EXCEL FOR ‘BIG DATA’. Why not? It’s not that excel is not capable of operating this connection (nothing like an excel crash due to the number of rows) and allows you to work with the data, it’s the fact you need to do extra work after the data was aggregated by the BI. This is the time to have a conversation with your analyst and explain him what you actually do with your data.
We love the pivot look
Somehow along the way, we got used to looking at numbers in a pivot table. We actually got so used to it, that now we find it hard to trust a simple graph. It’s time to ditch the old, ugly pivot style and move towards a VISUALIZED dashboard! No more overpopulated tables of data, only clean graphs.
We have 3000 kpis to every single session
“I just need one more KPI to make my decision!” This is our biggest crime in collecting data; we need to have them all. We got used to having so many KPIs we don’t even stop and think about the numbers we are using; we keep adding more and more KPIs in the hope they will reveal to us something the other KPIs haven’t yet. With all of this, we just forget the most important question:
Why do I need this data?
It’s about time we reduce noise. Take your funnels, analyze them and give weight to different operations, and create one final score. Ask yourself all the time: Why do I need this and how should it change what I am doing today?
Separate bidding decisions and performance research
Say what? During the day, online marketers mostly work on either adjusting bidding or creating new ads.
To adjust bids use only one number. This operation should be easy, you set one goal for your campaigns, you can adjust it based on your ROI, you can adjust it based on your installs, or based on your registrations, You cannot adjust bids based on more than one KPI, it should be an easy operation.
When it comes to performance research and ads creation, you should have the creative freedom to find correlation between different KPIs. Even here we can have shortcuts, but you can use as many KPIs as you think will help you find data.
We don’t classify our data enough
If we need to find a correlation, we need to have a true classification. We need to know what an image contained, what the leading concept and text was. We need to find the lookalike campaigns and group them, as this will shorten our data processing and will turn the big data into small data, as it should.
We complain the data is wrong
Data will never be perfect; all the time you keep adding more and more KPIs. I hear so many complaints about the data quality, but people don’t understand that fewer KPIs will allow the BI to actually focus on the important numbers and make sure they are correct, instead of spreading the effort all over the board searching for errors they will never find.
Big data is good to have, right data is more important. Use fewer KPIs, don’t rework your data in excel, and use as many graphs as you wish, just not pivot tables!
And for the last point… You should involve your analyst more in day to day business. Show them the way you use data, and they will be smart enough to find ways to shorten your working time. If you don’t have a full-time data scientist, hire someone based on need, and allow them to process your data to answer the organization’s questions.
Beomce a data-driven person and focus on the important parts