The ultimate dashboard guide - Do's and Don'ts
As an introduction and disclaimer: I love dashboards. Probably I'm biased because of my McKinsey background, but I do think they are an amazing tool in case you use them properly. Unfortunately, as in many other areas, people do it wrong: companies abuse the purpose of dashboards, especially in operations heavy, complicated businesses, where there are so many metrics and events to monitor. In this article I try to share some of the Do's and Don'ts that I learned about dashboards based on my experience so far.
When I took over the Managing Director role at foodpanda (previously NetPincér) in Hungary, one of my first things was to set up a management dashboard, right after creating a KPI tree for the business. The order is not a coincidence: the dashboard should come after you had understood your business and set up your KPI tree. (See my previous article how to set up a KPI tree.) This dashboard included only the most important metrics of the business, but helped me prioritise in my mind as well as provided some level of confidence that I knew what was happening on the ground while serving several thousands of orders per hour.
But dashboards do not work in all situations. There are management or strategic dashboards, similarly to the one that I created, and there are operational dashboards. While I'm a big supporter of the first one - and I'll cover this type of dashboards in this article -, I don't believe in operational dashboards at all. If we look at a random description of the dashboard, we realise that by definition it can refer only to management dashboards:
A dashboard is a visual display of the most important information needed to achieve one or more objectives; consolidated and arranged on a single screen so the information can be monitored at a glance.
Operations is usually the most complicated function of a company where dozens of metrics need to be fine tuned and hundreds of often unforeseen events need to be managed day by day. You wouldn't only need to manage a great magnitude of KPIs at the same time, you would also require near real-time data for many of them in order to effectively steer this function. While many companies try to tackle this by creating a pile of dashboards, this method usually doesn't work. In this jungle of dashboards operations managers are getting completely lost that results in not using those dashboards anymore. Or even if they do open them, there are so many that they have no chance checking all of them in time to identify issues or opportunities early enough. To manage operations, you need the concept of observability. I'll dedicate a separate article to operations observability, so for now let's just accept that we shouldn't create dozens of dashboards for our operations teams, because that doesn't only go against the purpose of dashboards, but is also a waste of time in most of the situations.
Shall then operations executives drop dashboards completely? I don't argue that, but they should rather use management dashboards that are simple and provide them a quick overview of their business. This sounds easy, but in reality it's very hard because people want peace of mind and will ask themselves: what if I won't recognise a spike or a drop in this KPI if it's not on my dashboard? Unfortunately you need to accept that dashboards are not built to identify every potential issue or opportunity that you may have in your business (for that, you can use observability), and thus they shouldn't give you a false sense of comfort. Dashboards shall help you understand if your business is going into the right direction in general, which areas you need to investigate or what topics you must prioritise. To indeed support these objectives, you need to choose very wisely which KPIs you add to your dashboard. As mentioned before, you shouldn't create an overcomplicated dashboard with too many metrics, as you should be able to understand the status of your business at a glance. I recommend adding (1) a few high-level business KPIs that can be monitored on a daily basis (i.e., orders, sales can be monitored with that frequency, EBITDA cannot), as well as (2) a few lower level KPIs that are key results based on your actual OKRs and thus by definition important for your strategy in that quarter. As the high-level business KPIs will be most probably lagging indicators, you should try to include some leading indicators that are related to your OKRs. It's difficult to define a rule of thumb how many metrics shall be on a dashboard, you should rather evaluate its simplicity by asking yourself: do I understand everything within 10 seconds looking at it?
If you know what to include in your dashboard, the next essential step is to ensure that you have a clear and aligned definition of all KPIs on it. It's best practice to include a separate tab next to your dashboard that describes (1) the sources of the data and (2) the definitions of the metrics, or how they were calculated. I have witnessed in so many cases that one team had a completely different view of their progress than another one, in spite of being aligned on the KPIs that measured their performance. This resulted from looking at dashboards that were constructed based on either different data sources or different definitions. Because of that potential pitfall I also recommend sharing the same dashboard transparently with your entire team - that ensures that you all look at the same metrics and evaluate performance according to the same standards.
Timing also matters. You should make sure that your dashboard is regularly updated.How frequently shall that be? Once again, it's difficult to share an exact number, but you should consider the right frequency based on the factors that drive the selected metrics. As a rule of thumb, the update shouldn't happen more often than a few hours, otherwise we're not talking about dashboards anymore, but we arrive back at the concept of observability. On the other hand, your dashboard shouldn't be updated too late. Most people start their day looking at dashboards and if the data is not there yet by the morning, you will sooner or later disregard that dashboard. It's also worth adding transparently to the screen how frequently it gets updated and when was the latest refresh so that everyone knows how up-to-date is the data they're looking at.
If you selected only a handful number of KPIs for your dashboard, showing trends won't make it more crowded but will actually add tremendous amount of context. As such I would propose not only to show a snapshot of the metrics, but also the week-over-week, month-over-month or year-over-year comparisons to understand if things are going in the right direction. Moreover, it may also make sense to show month-to-date or year-to-date numbers, especially if you want to monitor if you're on the right track to reach a monthly or yearly objective.
As mentioned before, you should be able to understand the content of your dashboard within seconds, and visualisation is key for that. It can be as simple as adding colour coding to the metrics (i.e., green = on track, yellow = requires attention, red = significantly below target), or you can include charts wherever they add value. There are several great (and highly valued) companies in the market that offer dashboard creation & visualisation. In case you have a larger company with a dedicated BI team that will develop the dashboard for you, most probably they will use Tableau or Looker. In case you want to set up a dashboard yourself, I recommend checking out actiondesk that is an exciting startup allowing dashboard creation directly from Google Spreadsheets.
Lastly, don't forget to iterate on your dashboards from time to time. Modify your dashboard based on your own learnings (e.g., you want to look at MoM trends instead of YoY) as well as because strategic priorities, and more frequently OKRs change. If you don't update your dashboard continuously, it will become irrelevant and useless quickly.
1) Naturally, you also need clean data for your dashboards, otherwise the foundation of it is missing. This is something you, as a business stakeholder, can't do much about, however it's important to keep it in your mind and you should urge your data or engineering team to work on that. Unless the data about your business is complete, transformed and ready for analytics, you may receive false information on your dashboard and you may make the wrong decisions. But that's a completely different topic that will be discussed in a different article.
2) How to ensure that you don't miss out on an important operational event or a sudden change in business performance? What to do with lower level operational KPIs? How to surface operational events that require immediate action? As I mentioned before, in order to manage such cases, you need operations observability, meaning near real-time operations monitoring & alerting. This will be the topic of my next article - stay tuned.