community analytics 101

The Context Problem

The fundamental underlying principle of analytics, including community analytics, is the idea of comparing...which is also called context. It's essentially what separates analytics from just plain data. Without appropriate context, you don't know whether the data is telling you that you are strong or weak. Figure 1 shows the total industry workforce for Manitoba communities as a percent of population. Note the wide range of data and how the averages for each geographic type varies. It's why our reports provide 3 distinct custom selected comparison community groups. We believe that choosing the comparison community groups is the most important analytic decision because it ultimately determines what you define as a strength and as a weakness. We recommend first comparison group to be of nearby communities, the second of similar communities and a third big picture view. These are our top community analytic problems:

  1. No Context: Not everyone values the full power of data. Some organizations value primary data (collected through interviews) above secondary data, which is typically collected by Statistics Canada or other government sources. The over-valuation of primary data means that community analytics start and end with "stakeholder" interviews. Although interviews can be a great source of local anecdotal information, they fail in providing context...a wider picture comparison of other similar communities. And in the end, they can't provide targets for what a community could look like in the future. We recommend analytics reports that can serve as an additional resource to interviews.
  2. Small Context: The conventional practice of comparing a community on a per capita basis with the provincial average is a case in point. Provincial averages are some of easiest comparable data to gather. But by definition, it's an average of strong and weak communities, which creates average statistics that are meaningless in many cases.
  3. Weak Context:  The second easiest comparable data to gather in community analysis is historical data (i.e. the previous census). This essentially provides a directional trend analysis. And while this type of analysis is helpful, it's still defines weak context because it still can't tell you whether you are strong or weak. The ability to compare your community with other similar communities is the only way to know your strengths and weaknesses.
  4. Small Data: Analytics requires data aggregation...but over-aggregated data can easily hide the real stories within a community. For example, industry labour analysis often uses 2-digit data (20 industry segments) to analyze the economic engine of a community. When industry sub-segments include both strong and weak areas, aggregated data will effectively bury the real strengths and weaknesses of a community. The problem is that you don't know until you actually analyze the industry sub-segments. As a result, we use Statistics Canada special run 3 and 4 digit data (427 segments).
  5. The Problem with Averages. Community profile statistics are typically calculated on a per capita basis in order to compare communities of different sizes. Furthermore, when communities are aggregated together as a comparison base, they are often weighted by population. In fact, Statistics Canada provincial community profile data is actually a weighted average of all communities within a province. This means that the comparison of a community to the provincial average is effectively a comparison to the largest communities in a province. For many smaller communities this is often inappropriate. Exceed reports offer either simple or weighted averages.
  6. Report Graphics Don't Show Comparisons Well. Not all graphics are created equal and the more data you're trying to display, the higher the degree of difficulty. And since good analytics equals good comparison groups, good analytics effectively means more data. The unfortunate practical consequence is often a binary decision...good graphics or good analytics. For example, pie charts are pretty, but force the viewer to compare angles which are more difficult for our eyes than a bar chart. What's even worse is trying to compare several pies where the viewer is asked to compare slices located within the pie and between pies. We solve this by prioritizing good analytics...but by using colors (i.e. heat maps) and bar charts as often as possible, which are excellent at showing comparisons.