One of the most common discussion point in our internal meetings is – What insights are beneficial for customers ?
Here is the way I train myself and the team –
1. User is not asking for tons of data. User is asking for one actionable insight.
An aha moment from the data that she did not know before.
2. User wants insights. It can come from a summary view too. Though, at the same time, a column/pie chart having more than 5 columns is not useful.
3. Think like a box-plot. so many insights in the whiskers of it. A great way to summarize an entire population and used by traders frequently.
4. An insight can be generated from one data type or more. Tougher to do it on one data type. For example, technology acquisition by a user is a data point but technology acquisition over time is a good insight.
Easier to generate insights when one variable is observed over time or another variable.
5. Insights change with time. Intent data is an insight today…but when everyone starts using it, then it wont give results and won’t be of use. It will become a commodity. It will lose the aha part.
6. Overdose of insights does not work. If I compare many categorical variables and measures, you can generate graphs but no use of them.
7. Every insight is within the limitation of data behind it. For example, there are more than 11K+ companies hiring in AWS related technologies this month but more than 40 % of them are IT services and staffing agencies. If you remove this 40 %, suddenly the insight changes.