We had a team exercise to design a simplest-possible-but-still-useful BI dashboard / report / KPI.
Our minimalist requirements were very simple – the report should look great, be informative and useful on:
n iPhone at 2AM – even for a sleepy person;
n Paper WSJ in black and white.
These requirements were against all the industry “best” practices (such as: impress your boss with flashy reports, increase the Wow! factor etc).
We (humbly: me – myself) discovered a minimalist design using the old German principle “Less is More”.
A stock ticker.
A whatever ticker.
Let’s say we are building a DAILY DEMAND dashboard for an Animal Farm Forecasting Engine.
Here is a ticker for a daily forecast – which is informative and almost self-explanatory:
n MAPE – Mean average percentage error of the forecast. The key metric for this dashboard;
n Chicken/Egg – we are measuring prediction for chicken eggs, not alligator ones;
n 1,000 – volume, units (eggs in this case);
n @1.02% — value of the MAPE, with 1K volume the error is about 10 egg;
n ▲ 0.02pp — forecast change since yesterday, in percentage points.
MAPE/GOAT/MILK2KL@1.02% ▲ 5W ▼ 2Y
The forecast for Goat Milk has a 1.02% error with volume of 2000 liters.
The error went up by 5 percentage points since last week and down by 2 percentage points since last year.
A ticker represents a single point on a curve with:
n Context (which helps to answer the usual WTF? questions);
n Volume (which helps to assess the metric weight and possible repercussions – a 50% error with a volume of 2 units is not that bad);
n Velocity of change;
n Difference from the previous measurement.
What else do you need for a quick status report?
The ticker tape was invented more than 150 years ago. Since then Wall Street gave us many well designed reports based on stock tickers. For example – the famous Smart Money heat map
Why not use them?
All you need is to express your metrics as tickers.
Everything has been already invented!
So if you ask your BI engineers to design a minimalist report, and they come back with tables and diagrams – they are probably overpaid.