Difference between descriptive and decision-driving dashboard

Popular phrases these days are “data-driven” decisions, using “power of data”, “actionable insights” and others. In its core the value added of a BI team is to provide these deliverables. Company’s investment in development of “data analytics” is just like any other, it is expected to deliver value. This value is often the ability to build dashboards that enables the stakeholders to make decisions. So how exactly to make these dashboards? This article will give you 3 examples how to turn a regular fancy dashboard into dashboards that empowers data-driven decisions.

Development of BI professionals is a mission of ours. This post is a must if:

• You are working a BI job and are looking to enhance your dashboard building skills;

• You are a senior manager in Finance looking to improve your decision-making;

• You are a BI Manager tasked with developing data-driven culture;

• You are a student and love data analytics and want to start with the most valuable basics;

As always we will be using Tableau as a tool of choice. The data set that will be used in this example is a simulated account receivables operations. Here are the details:

The Case

Overall we are tasked to analyze the data for payments and open invoices and the work of the AR department. We have data on what invoices are paid, when are paid and also what invoice are still not paid. Also there is data for customers and products. This is pretty much your standard AR data set that most transactional systems such as Oracle or SAP can give you.

So having the data set we created two dashboards in Tableau. One that is more descriptive, easy make, with eye-catching design, and another dashboard dedicated to empowering data-driven decisions and giving actionable insights. It is the same data just a matter of extracting value from it via data analytics.

This is the first dashboard:

Overall here we see descriptive vizulizations:

• Showing Trends of collections;

• Simple Variance Analysis;

• Summary of AR debt by different dimensions;

• Top clients.

These types of analyses are pretty standard, but those are descriptive in nature. Knowing the trend or the variance of collections by product is great and informative but it does not lead anyone to make any decisions. Personally I classify those types of analytics to be great for “power-point” type of tasks and poor for driving decisions.

On the other hand we have this dashboard:

On first glance it is very rudimentary and poorly spaced, but it does give a lot more to users than the first one. There are several key analytics that can be derived from it.

• Since we have information on the invoices then AR people can take action directly on particular invoices;

• Implementing analysis for bad debt provision gives senior Finance stakeholders insight into the impact of AR operations on P&L and can spark the right conversations;

• Clients are classified by performance which helps turn attention to dealing with the problematic ones.

We still keep descriptive vizualizations like the trend and collections breakdown by teams and the heatmap, but it is not the focus, instead the main space is dedicated to those charts that lead to decisions.

 How to turn descriptive analysis into decision engine

There are 4 key concepts in turning descriptive dashboard into one that provides actionable insights:

Give details

High-level data is great for overview and trending but rarely gives the opportunity to make decisions. Usually stakeholders want to see what is behind those high-level numbers and they can act on it. In this case we are showing the full list of invoices with all kinds of information on them. I like to do that in a table because I can show a lot more attributes in this table format. Additionally, Tableau has great drill down capabilities so it is always a good idea to apply it. Usually I apply the drill-down action from all other charts to the details table. This means the user can see the details (in this case list of invoices) behind every different analysis that is shown. Seeing these details makes it possible to bottom level stakeholders (in this case credit controllers) to use the dashboard, as they will work on invoice level and need these details, this makes it more operational rather than just high-level overview for senior stakeholders.

Embed modelling (e.g. customer classification)

Most dashboards are descriptive in nature, but if you want to make your project special and give the stakeholders valuable insights then you need to apply some sort of modelling to it. Showing the results is not enough, the job of the BI analyst include also inferring conclusions by applying modelling. It is a big task where you need to have domain knowledge, data science skills and overall business understanding. Also you must be able to explain to your stakeholders your model. One of the best examples is applying classification models to group some of your data (customers, invoices, accounts, employees etc.) into groups. In our case we have applied a simple scoring method for invoice (based on when it is paid) and from there classified the customers into good or bad payers. This allows to check out each customer and know if we have issues with them. Also showing that inferred classification in the detail table multiplies the value to stakeholders as they now know from where each invoice has come.

Show risks (breakdown of due debt by customer class)

Business analytics is all about profit, cost and risk. When building dashboard on a particular data set you always need to think about the business risk and how your data can add value in understanding the risks to the business. Granted this requires quite a lot domain knowledge and business understanding, but here is where the BI value goes into hyperdrive. Bringing the underlying table to life means extracting the risk and showing it in understandable way to the stakeholders. The value for them is that there is no way just look at the data table or the trend chart and know what kind of a risks to the business are hidden in this table.

In our example we use the classification of customers and then breakdown the due debt into those categories. Basically we are saying XX amount is sitting with “Performing Customers”, YY amount is sitting with “Bad customers”. This immediately opens the eyes of senior stakeholders and identifies the risk – they know what amount may not be paid or is at risk of not being collected.

When building dashboards always think and model the risk to the business, this will greatly help your stakeholders and they will be very very happy. Also BI people are in great position as we have both the business knowledge and data analytics skills to identify those risks out of the data, because for a lets say Financial Director with no background in analytics maybe difficult to uncover all the possibilities that the underlying data gives him.

Give opportunity to the user to manipulate the modelling

Another great option to improve your dashboard is to allow stakeholders to manipulate themselves the analysis that is performed. Basically giving them opportunity to perform “what-if” analysis on your already existing model. In Tableau this is done by the use of Parameters that users can then change in the live dashboard. Of course this complicates any modelling that is applied in the dashboard because it needs to be dynamic and think about different scenarios that users can exploit and how this will affect the data on the dashboard. However this is a very powerful tool for business analysts as it gives them the chance to apply their own analytical skills on your dashboard and this greatly enhances their experience.

In our case we calculate the provision based on the customer score. Basically what % of the outstanding debt of particular type of client must be provisioned. Then the percentages for each customer group themselves are via Parameters that the user can manipulate and see what the impact on the total provision will be.