Data analytics is the process of analyzing large quantities of data. There are many different lenses that an analyst might take when approaching an analytics project. We can divide these into four overarching categories:
- Descriptive Analytics
- Diagnostic Analytics
- Predictive Analytics
- Prescriptive Analytics
All four of these involve analyzing data, the difference lies in the outcome of the analysis. We’ll dive into each one below.
Descriptive analytics aims to describe a current or past situation. It does not seek to pinpoint cause and effect relationships or to solve a problem. Questions that may be explored using descriptive analytics are:
- How many people live in each county?
- Are there more men or women enrolled in college?
- What is the most common language used in a region?
- How much did revenue grow year-over-year for my business?
- How many people used my product last year?
An important element of descriptive analytics is comparing metrics for context. Simply finding that the past year’s revenue was $670,000 does not say much with nothing to compare it to. To add context, you may compare the metric to the prior year’s revenue or to the amount of revenue you had been expecting. Other metrics may be compared as parts of a whole. For example, you may contextualize the fact that 10 million people in California speak Spanish by comparing it to the state’s total population of 40 million – about one quarter of Californians speak Spanish.
Descriptive analytics is often the first step in a larger project. When a business or organization is trying to develop a strategy or solve a problem, they must first use descriptive analytics to understand the current situation.
Diagnostic analytics builds on descriptive analytics to understand why something is happening or has happened. For example, if you use descriptive analytics to find that the number of people using your product has fallen over the past month, you might use diagnostic analytics to pinpoint the reason behind the dropoff.
Methods used in diagnostic analytics include drilling down into the data and seeking correlations between variables. Drilling down means looking at data on a more granular level. To drill down into the loss of users, you might split out premium users and non-premium users and see whether both subsets have decreased or if only one has. From there you might go even more granular by splitting out daily users and weekly users. Isolating the decrease in users can provide a clue as to what issue is causing users to drop off.
Finding correlation between variables can also provide clues about the reasons behind trends. Linear regression is a more advanced analytics technique that can help isolate the correlation between particular variables without confusing the effects of a third variable.
It’s difficult to definitively find a single reason for a trend or phenomenon. However, with diagnostic analytics techniques, data analysts can develop reasonably plausible hypotheses.
Questions that may be explored using diagnostic analytics include:
- Why has the number of users decreased since last month?
- Why has the population of this county increased so much?
- Why are more women than men enrolled in college?
- Why do more people speak Spanish than Japanese?
- What caused the increase in company revenue this year?
Predictive analytics uses large amounts of data alongside algorithms and, sometimes, machine learning to project a likely outcome based on historical and present trends. Predictive analytics is quite difficult because there is no absolutely fool-proof way to predict the future. Furthermore, most analysts who perform predictive analytics try to predict specific metrics rather than general trends. For example, predicting that revenue will increase over the year is somewhat helpful, but predicting how much revenue will increase is much more useful for business planning purposes.
Predictive analytics uses advanced statistical techniques like modeling, simulation, machine learning, and regression.
Questions that you could explore using predictive analytics include:
- How much revenue will my company generate this year?
- How many students will enroll in college next fall?
- Which product will be most in-demand next week?
- What will the population of the county be in the year 2030?
Prescriptive analytics combines all of the previous types of analytics to try and develop a prescribed set of actions to be taken in order to achieve or avoid a particular outcome.
In order to master prescriptive analytics, analysts must use descriptive analytics to understand the situation as it stands, diagnostic analytics to get an idea of which variables impact one another, and predictive analytics to understand a range of possible future outcomes. Prescriptive analytics is then used to choose a course of action that is likely to produce a desired outcome.
Like predictive analytics, prescriptive analytics also uses advanced techniques like regression analysis and machine learning.
Questions that can be explored using prescriptive analytics include:
- What products should we stock in order to double revenue?
- Which classes should be included in the curriculum in order to increase college enrollment by 50%?
- Should the county increase the number of post offices as population increases?
- How will creating a Spanish-language version of my website affect user metrics?
To summarize, the four overarching categories of data analytics are descriptive analytics, diagnostic analytics, predictive analytics, and prescriptive analytics. A well-rounded data analyst should aim to master each one of these to be able to complete advanced projects with robust analysis. For an entry-level analyst, descriptive analytics is a great place to start!