Ten Top Tips for Analysts in the Police and Fire Service

More than 20 analysts from police and fire services joined us for day one of our 2019 Client Conference last week in Worcester. This year we revealed some of the tips and tricks we use in analysing and presenting data. The agenda covered the universal truths of data analysis, some common pitfalls in analysing incidents, how to validate models, shift pattern design, and best practices for data visualisation, among other topics.

In this blog post, we’d like to share some of the hottest tips from the day.

  1. Check your work. Ask yourself whether the story your data is telling you sounds right. If you don’t believe it, or it contradicts your main hypothesis, check your data and assumptions.
  2. Take care with Excel. Home Office codes and some Athena numbers can look like dates to Excel and end up getting automatically converted to the wrong format. Commas in free text might lead to unexpected results in a comma separated values (.csv) file, too.
  3. Use equal time periods. Why did total demand drop in February and then rise again in March? Perhaps because February has fewer days. You can see significant variations in incidents if one month has five weekends and the months before and after only have four. To avoid this, it can be better to look at demand by 28-day months. If you’re doing a weekly analysis, trim the year to complete weeks.
  4. Use public data to validate. There is lots of information in the public domain that you can use to sanity-check the data you’re working with. For example, crime data and stop and search numbers are available by force, month, geographic area (LSOA) and outcome. There’s lots of traffic accident data, drink driver data (around Christmas and summer campaigns), and police resource data available too.
  5. Know your limits. Mark Twain famously said: “A lie can travel halfway around the world before the truth can get its boots on.” That’s true of bad data too. There are lots of examples of statistics that have become gospel before they’ve been properly checked. Know the limits of your analysis and be explicit about them, to avoid misunderstandings. Your data is your responsibility.
  6. Be flexible in workshops. We run workshops to get a more detailed understanding of the baseline, and it’s important to go into them fully prepared with a checklist of questions. In the session, though, you need to strike a balance between getting the information you want, and letting the participants talk freely. Time is limited, but you can gain some golden insights by letting attendees express their views.
  7. Use bar charts. There are all kinds of data visualisations available today (some of them closer to art than function). Bar charts are usually a good safe choice, though. Everyone knows how to read a bar chart, and it’s easy to compare the different values on them. Be careful with 3D bar charts, where some bars might be obscured and it can be hard to read values off the axis.
  8. Defaults are not your friend. You can choose colours, fonts, and designs that help you to tell your story. It has huge impact, for example, to change the colour of those bars you want to highlight in a bar chart, so they pop off the page. Use contrasting colours that work for those with colour blindness. Consistency of colours is another good tip – if you used green for South and Blue for North on the first chart, then use if for the other charts too.
  9. Beware of “chart junk”. Avoid magazine-style illustrations combined with your graph that are distracting and cloud the message. Do use good annotations, but ask yourself whether everything else on the slide is helping your audience to understand your message.
  10. You might not want the “best” shift pattern. When creating shift patterns, start by defining the problem in detail. Understand the resource requirements, all the demand requirements and any constraints, such as the need to work in teams or to support flexible working. XIMES Shift Pattern Design can find shifts that provide the best match between resources and demand, but that pattern might not support team working, and it could be overly complicated. You might prefer a shift pattern that doesn’t match demand so perfectly, but that meets the full requirements much better. Part of the power of XIMES is the capability to explore lots of options – use it! Remember too that shift pattern design is subjective, so involve the team to secure their buy-in and show them the options you have examined so that they understand the compromises that need to be made.

If you’d like to explore any of these ideas further with us, feel free to get in touch.