What Are the 4 Main Analytical Models?

The four main analytical models organisations can deploy are:

1. descriptive
2. diagnostic
3. predictive
4. prescriptive.

As you move from descriptive to prescriptive analytics, each model offers increasing value to an organisation. But, at the same time, they increase in complexity.

Adapted from Gartner’s Analytics Value Escalator[1]

Descriptive analytics

Descriptive analytics answer the question: What happened?

This is the most common type of analytics found in business. It generally uses historical data from a single internal source to pinpoint when an event occurred.

For example:

• How many sales did we make in the last week/day/hour?
• Which customers required the most help from our customer service team?
• How many people viewed our website?
• Which product had the most defects?

Descriptive analytics are often displayed on dashboards and in reports, which are convenient ways to consume data and inform decisions. Descriptive analytics account for most of the statistics we use, including basic aggregation (e.g. count or sum of values filtered from a column or data), averages, and percentage changes.

Diagnostic analytics

Diagnostic analytics help us to answer the next question: Why did it happen?

To do this, analysts dive deeper into an organisation’s historical data, combining multiple sources in search of patterns, trends, and correlations.

Why would you use diagnostic analytics?

• Identify anomalies: Analysts use the results from descriptive analysis to identify areas that need further investigation and raise questions that can’t be answered by simply looking at the data. For example: Why have sales increased in a region that had no change in marketing?
• Drill down into data: To explain anomalies, analysts must find patterns outside existing data sets to identify correlations. They might need to use techniques such as data mining, and use data from external sources.
• Determine causal relationships: Having identified anomalies and searched for patterns that could be correlated, analysts use more advanced statistical techniques to determine whether these are related.

Traditionally, data analysts performed diagnostic analytics manually, but as data volume, variety, and velocity increase, fully manual analysis is no longer feasible. Instead, modern diagnostic analytics solutions employ machine-learning techniques to augment the analyst’s skills. Computers can process vast amounts of data and recognise patterns, detect anomalies, and expose ‘unusual’ events, and they can apply analytical techniques from a portfolio of algorithms to identify drivers of change and determine causation.

Predictive analytics

As an organisation increases its analytical maturity and embarks on predictive analytics, it shifts its focus from understanding historical events to creating insights about a current or future state. Predictive analytics is at the intersection of classical statistical analysis and modern artificial intelligence (AI) techniques. It tries to answer the question: What will happen next?

It’s impossible to predict exactly what will happen in the future, but by employing predictive analytics, organisations identify the likelihood of possible outcomes and can increase the chance of taking the best course of action. We see predictive analytics used in many sectors.

For example:

• Aerospace – Predictive analytics are used to predict the effect of specific maintenance operations on aircraft reliability, fuel use, availability, and uptime.
• Financial services – Predictive analytics are used to develop credit-risk models and forecast financial market trends.
• Manufacturing – Predictive analytics are used to predict the location and rate of machine failures, and to optimise ordering and delivery of raw materials based on projected future demands.
• Online retail – Systems monitor customer behaviour, and predictive models determine whether providing additional product information or incentives will increase the likelihood of a sale.

Simple predictive models can be created using tools such as Excel or Tableau. As these models start to accommodate more variables, with more complex relationships, these analytics become the responsibility of data scientists.

Prescriptive analytics

Prescriptive analytics is the most complex type of analytics. It combines internal data, external sources, and machine-learning techniques to provide the most effective outcomes. In prescriptive analytics, a decision-making process is applied to descriptive and predictive models to find the combinations of existing conditions and possible decisions that are likely to have the most effect in the future. This process is both complex and resource intensive but, when done well, can provide immense value to an organisation.

Applications of prescriptive analytics include:

• risk management[2]
• improving healthcare[3]
• guided marketing, selling and pricing[4].

As the most complex form of analytics, prescriptive analytics not only pose technical challenges, but are also influenced by external factors such as government regulation, market risk, and existing organisational behaviour. If you are considering deploying prescriptive analytics, be sure you have a solid business case that identifies why machine-generated recommendations are appropriate and trustworthy for each decision.

As well as identifying and programming each decision, data scientists developing prescriptive models need to prevent missteps by ensuring that all possible outcomes are considered. After deploying these systems they must test their models repeatedly, to ensure they are making meaningful recommendations and there’s no risk of costly mistakes.

Your thoughts

The more value a type of analytics adds, the more complex it is to implement. Many organisations progress ‘up’ the levels of analytics, starting with descriptive analytics.

Read Unlocking the value of data for improved performance from Tableau to understand the value of different types of analytics.

Then consider:

• What are the benefits of progressing ‘up’ the levels of analytics?

References

1. Gartner’s analytic value escalator [Image]. Gartner; 2012. Available from: https://www.flickr.com/photos/27772229@N07/8267855748/
2. Hare J. Use prescriptive analytics to reduce the risk of decisions [Internet]. Forbes; 2016. Available from: https://www.forbes.com/sites/gartnergroup/2016/03/24/use-prescriptive-analytics-to-reduce-the-risk-of-decisions/#188d02e06958
3. Kuttappa S. Optimise healthcare delivery and reduce costs with prescriptive analytics [Blog]. 2020 Apr 14. Available from: https://www.ibmbigdatahub.com/blog/prescriptive-analytics-cure-transforming-healthcare-industry
4. Dent C, Burns D, Sherrard S. Do this, not that: prescriptive analytics in sales and marketing [Internet]. 2019 Aug 27. Available from: https://www.bain.com/insights/do-this-not-that-prescriptive-analytics-in-sales-and-marketing/