Jade Software Wed, Jun 17, '20 8 min read

Understanding the four types of data analytics

The volume of data in the world is growing exponentially. You only need to look at the significant amount of data generated from not just storing but analysing video footage. For instance, one hour of CCTV footage conservatively equates to approximately 2GB of data. That’s 24GB a day and nearly 20,000GB a year – just for one camera. Enter data analytics.

Broadly speaking, data analytics can be broken down into four areas: descriptive (what is happening), diagnostic (why it is happening), predictive (what might happen next), prescriptive (what course of action should be taken).

Each of these four types of analytics require different approaches or techniques to implement. However, just because a business may only have the expertise or inputs to undertake one type of analytics, doesn’t mean it can't progress into the more advanced types. Businesses can grow from descriptive right through to prescriptive.

So, let's take a look at each type of data analytics.

Descriptive analytics.

The first type of analytics is descriptive, which is commonly used for helping people visualise, absorb, and interpret data – in other words, describing what happened. These visualisations often take the form of frequent reports that provide insights into past performance, helping answer any questions you may have about particular events or activities.

Example: A passenger transport business uses descriptive analytics to understand its on-time performance. It finds that its vehicles on a particular route are consistently 149 seconds late at each stop.

Diagnostic analytics.

The second type of analytics is diagnostic, which seeks to drill down into a situation or activity to understand why it happened. If descriptive analytics was macro in scale, diagnostic would be the micro equivalent. This form of analytics is aided by additional data sets, such as traffic, urban planning, infrastructure or construction, weather, etc. However, the more data sets you add to an analytics programme, the higher the potential noise - skewed results.   

Example: The passenger transport business uses diagnostic analytics to understand why its vehicles are 149 seconds late. The transport company found that the opening of a new housing development had increased not only traffic on the route, but also the number of passengers travelling on their vehicles, resulting in a delay of their services.

Predictive analytics.

The third type of analytics is predictive, which springboards off the back of descriptive and diagnostic analytics to identify potential outcomes. The key to remember with this form of analytics is that, to a point, predictions are typically more accurate with each additional data source, as mentioned above. For instance, when you take retail sales data and combine it with weather data, you can predict likely sales figures according to either weather forecasts or the current weather conditions, which enables businesses to respond accordingly.

However, as many data scientists will be quick to point out, correlation does not imply causation. For instance, one could compare the sale of soup and umbrellas. While you could draw the conclusion that as more umbrellas are sold so too are more bowls of soup, the cause for such an increase in sales is more likely due to the onset of cold, rainy conditions.

One key thing to understand with predictions is they are purely that, predictions. Depending on the variables of the analysed activity or event, the level confidence of each prediction will drop as forecast period extends. Normally, you will have more confidence in a one-year prediction than a ten-year prediction. Either way, you will have information in front of you to make a more informed decision.    

Example: The passenger transport business wanted to be more proactive in its timetabling. So they used predictive analytics to forecast when they will need to amend their timetables to provide more accurate services to its customers. It found that their on-time performance of their current timetables would start reaching mildly unacceptable levels in six months and completely unacceptable levels in twelve months.

Prescriptive analytics.

Like the finale of a TV series, the final type of analytics is where the magic really happens. Prescriptive analytics is the process of scrutinising a range of potential outcomes that flow on from certain actions, then providing guidance on which action to take. This form of analytics is by far the most complex, requiring advanced modelling techniques, machine learning, algorithms, visualisation, and automation. This final form of analytics takes considerable expertise but is where businesses are set to gain the most value.

Example: The passenger transport business wanted to take their analytics to the next level, so they implemented prescriptive analytics. The business also was getting more agile, automated, and mature in its practices and added additional data sets to its analytics. They now receive alerts when their services are likely to be delayed and recommendations of how to proceed.

  • In one case, the operations manager received an alert after noticing the following week's torrential rain forecast would increase traffic congestion. The analytics recommended the manager schedule three extra vehicles during peak times.
  • In another case, the operations manager found that a bus lane was planned on a major route. The analytics recommended the manager add an extra two vehicles during peak times because of construction delays, but return to the normal number of vehicles once it opens. It also recommended extending the time between each stop by one minute as the vehicles would be able to travel the route much faster.

While most businesses are yet to achieve the full level of data analytics maturity, due in part to the global talent shortage in this area, outsourcing is a viable option, particularly through automated analytics.

Learn more in Part 3: How to realise the true potential of business intelligence and data analytics.

Talk to us about making data meaningful