The missing piece in analytics - socializing the insights
We recently started a discussion about the value of writing at work. This is increasingly regarded as one of the most important skills in knowledge work, a trend accelerated by the shift to working from home.
One of the roles where I’ve seen the biggest gap between the actual importance and the attention paid to writing is in analytics. We often think of our work as analysts and test analysts on a variety of technical skills, but rarely on presentation of information and insights. However the data is useless without the action it drives, and that action is often in the hands of other team members. Effective analysts are aware of that and invest heavily in ‘socializing’ their findings.
The task of communicating analytics though, is quite tedious. You often find yourself creating screenshots of charts, figuring out how to laboriously annotate them, and then integrate them in decks or long e-mail threads. More often than not, the chart is misunderstood or data is questioned and alternative hypotheses come up in an effort to get to the root finding. Why is the number of users decreasing — is it equal across geographies or is it a particularly large cohort that started churning as their free trial expired?
An e-mail thread can quickly get out of hand, and there’s no way to track all the comments given in a deck. Radical tries to solve this by moving the discussion to where the data is. Radical lets you chat, annotate charts, embed new data and richly express your thoughts, ideas and questions right inside the dashboard, regardless of whether it’s Google Analytics, Data Studio, Chartio, Amplitude, Mixpanel, Tableau or something else altogether. Better yet, it integrates neatly with Slack so it can bring the entire org, even users not used to your BI tool, to bear on the question. You’ll be driving much more actionable insights and impact when you’re using Radical to get to the bottom of any data discussion quickly.
To wrap up though, some quick tips on communicating data:
- Avoid jargon and complexity. Your goal is to get your message across, not to show off how smart or sophisticated your analysis is: “15% less users, mainly in UK” is better than “DAU net churn increasing, distributed unequally across geographies”.
- Show the point, not everything you tested. You might’ve spent a week trying out other drill-downs to discover the problem is with mobile devices, but your audience shouldn’t have to spend an hour learning the problem wasn’t with any specific language, region, cohort, etc. etc.
- Get the message through in the chart title. The chart title should read like a great newspaper headline — “Usage down 15%, 30% in UK” is much better than “Daily use over time by country”. Make it easier for your reader to see the takeaway.
- Use the simplest visualization possible to convey the finding, not the richest. There’s no use seeing 60 lines for different countries instead of a line for the average and the few outliers we care about. Sometimes, we don’t even need the timeseries: Bar charts with categories convey a lot of information if there’s no important trend over time to show.
I hope this is useful, and that you give getradical.co/data-science/ a try and see how it can change how your analytics team becomes a more effective partner to the product and business.