This podcast serves as a Part II to an earlier episode. In Part I, titled, “Making Better Decisions with Data (Episode 9)” Chris outlined basic thinking around how to develop and execute a data project by keeping cause and effect in mind.
In this second part, Chris continues along this same vein of logic to describe a dashboard reporting philosophy defined best as a Management by Exception approach to reporting. In situations where your outcomes metrics are a score, rating, or otherwise continuous, there will be normal fluctuation. Knowing when something unique is happening becomes critical to effective reporting in these circumstances.
In this episode, Chris addresses how to develop management by exception approach to your data dashboards by addressing the following key points:
- Variable Type: Not all metrics are created equal. Make sure you understand cause versus effect, inputs versus outcomes; and record the data appropriately.
- More is More: When using a “management by exception” approach to reporting, it’s okay to have a lot of metrics because you’re not going to be trying to react to everything. You’ll only be reporting on metrics that are important and unique.
- Charting Data Over Time: In many cases you’ll be charting metrics over time as either a count or average reading. Connecting these measures over time with a simple line chart will reveal a natural ebb and flow that represents the distribution of the metric.
- Unique Variation: As such, we can use a static called Standard Deviation to identify the likelihood that a given measure of this variation is “normal” or “unique.” The more severe the implications, the less sure we need to be that variation is unique before it warrants our attention
- Management by Exception: A Management by Exception Dashboard is only highlighting those situations where the threshold of tolerance for a metric is reached, and a measure is a “unique occurrences.”
Links of Interest
- More reading on Management by Exception as a business principle: https://en.wikipedia.org/wiki/Management_by_exception