An approach on building a reporting
I told myself it was probably time to sit down about my experience on this topic.
When I discuss with business stakeholders, the main challenge is to make sure we are the same page when we discuss data.
Let’s take the example of a request to build a “sales ranking” for pens from an EC site.
1. Meaning of the data
Which item should appear first in the list:
- a $1 pen sold 1000 times?
- a 1000$ pen sold once?
The definition of “sales ranking” will differ according to the stakeholders’ intentions
One approach is to allow both to be displayed/selected:
- ranking per number of units sold
- ranking per revenue
That’s somehow what the App store did for sometimes (top paid, top-grossing).
The problem? the more you dig, the more you might be tempted to give options to view the data and transfer some complexity to the business users.
Therefore it comes to a point: what is the purpose of this ranking, and for who?
My personal experience is the more “expert” the audience is, the higher the number of options might be necessary. An EC-site manager will spend his days digging data to find insights and stories (and hopefully pinpoints profitability).
As a consumer, I want something simpler, such as an idea or even some inspiration. I want maximum output with minimal input. This result requires implementing a bias in how you calculate and present data. The App store captures this perfectly well: when you log to the app store, you see stories, categories but the notion of “sales ranking” was diluted over time. The App store presents biased data on purpose.
Let’s go back to our pen example, what kind of bias should be? Do we rename the “sales ranking” by “popularity” and compute the ranking based on units sold?
Moral questions will/should come up when you figure out what data to present and for who. For example, is the packing and shipping cost for 1000 pens at $1 profitable if you have ecology in mind? (nowadays, it’s not impossible to order for 1$ online). Would it be better to present data differently to influence a purchase?
Let’s keep this in mind and move on to a 3rd component: time.
3. Time range
The last component, what should be the time range of the sales ranking? Daily? Weekly? Monthly?
Assuming (optimistically) you can have access to the raw data, how do you define the time range:
- from 00:00 am to 23:59?
- from Monday to Sunday?
- Does monthly really mean from the first day of the month until the last?
If you can access consolidated data from the platform, you should know when (that is, precisely at what time) those data are ready to synch properly!
Conclusion: Whenever I need to create a dashboard what I do is:
1. Sit down (even virtually) with the stakeholders;
2. Grab sample data and do the ranking by “hand”
3. Simulate several outputs variation
I haven’t found a more efficient and ambiguity free approach that reaches complete consensus.
It feels like it hasn’t changed for the past 30 years.