Four types of data analytics you must know and what to expect

Hi guys, I would like to share about four types of data analytics that I know, and it’s common knowledge, I just want to share it based on my perspective. This topic will be very useful for any of you working in and with the analytics team.

Photo by Luke Chesser on Unsplash

Descriptive analytics

In short, descriptive analytics tells us what happened in the past. In general, when you are doing or asking for this type of analytics, the result is typically by doing a benchmark for specific metrics in a specific area, such as marketing, sales, product, operations, etc. After you set the requirements, the analytics team will try to provide the data needed by determining where is the data stored, how to extract the data, and finally to put it on a display using data visualization, exploratory data analysis, and so on.

For example:

1. The marketing team would like to know which paid ads channel generated the best return.
2. The sales team would like to know which product category and user segmentation generated the highest revenue.
3. The product team would like to know the conversion rate funnel from their feature, so they can assess which funnel needs to be optimized using the A/B test.

Diagnostic analytics

In short, diagnostic analytics tells us why X happened in the past. The result of this type usually makes us understand about which factors impacted the most and least, positively or negatively.

For example:

1. [The marketing team case] After further checking, it is known that the marketing budget for paid ad A higher than B, but paid ad B generates higher returns because the conversion is much better than A. [Insight] We can focus our budget on paid ad B rather than paid ad A.
2. [The sales team case] After further checking, it is known that from the total of 10 product categories, 30% of the sales contributed by a single product category, which is Electronic, but in terms of revenue, 70% of it contributed by the Food category because the margin rate is significantly higher. [Insight] We can focus our campaign next month on the Food category rather than the Electronic.
3. [The product team case] After further checking, the number of traffic that comes to our website average is 1M, but only 50K converted until purchase, which was caused by a significant drop when reaching the payment page. [Insight] Lot of things can be done, such as check to Engineering is there any problem with our payment page that maybe makes our customers can’t continue their purchases? Or maybe our Pay button location is not good enough to be found?

Predictive analytics

In short, predictive analytics tells us what is most likely happen in the future. Usually, we do this type of analytics after we have done the descriptive and diagnostic part because those things are essentially needed before we do anything related to predictive analytics. We can use statistical modeling and machine learning to determine some metrics which is heavily related to probability and also confidence rate. It helps the decision-makers because it gives them a picture of what will happen in the future.

For example:

1. [The marketing team case] After looking further, we find that the performance pattern of paid ad channels is not always B the better one. Using historical data that we have, we can create statistical modeling to know on average, how much return we can get in the future if we allocate the marketing cost 50:50, 60:40, 75:25, and so on.
2. [The sales team case] After checking further, we know that the Electronic category only has several purchases higher than the Food around payday, on another date, it’s always the Food as the champion. We can use predictive analytics to predict which date we need to start focusing on the Electronic before getting back to focusing on the Food category.
3. [The product team case] After doing further checking, we can use predictive analytics for each customer segmentation, time of purchases, and so on, what kind of treatment we have to give to our customers that makes they want to continue their purchases, maybe give them pop-up of additional cashback after they stay for more than > 15 seconds on the payment page.

Prescriptive analytics

In short, prescriptive analytics tells us what to do in the future. This part is where actionable insights are generated. Usually, it is produced after the diagnostic or predictive type. You should know this type is the hardest type to produce because not all solutions can be produced from our data analysis. Based on this article:

According to Prescriptive Analytics Takes Analytics Maturity Model to a New Level, a Gartner Report has indicated that only three percent of surveyed businesses are utilizing prescriptive analytics, whereas about 30 percent are actively using predictive analytics tools.

As you can read, it’s only 3%, and as far as I know, some of the top companies that can produce prescriptive analytics as a habit such as Microsoft, IBM, Oracle, and so on. So, you don’t have to be discouraged if still can’t produce prescriptive analytics as a habit, but it will be great if we can start learning from now, doesn’t it? :)

For example:

1. [The marketing team case] After final checking, statistical modeling, AB testing, and else, we can recommend something like “Let’s split our marketing paid ad budget to A and B by X:Y, because it will generate the highest result next month.”
2. [The sales team case] After final checking, statistical modeling, AB testing, and else, we can recommend something like “Let’s focus our discount coupon for this user segmentation A when purchasing B product category as much as 10%, for the other type, not more than 5%, and for the champion segmentation, we don’t have to give them any discount because we believe they will still purchase without any discount from us.”
3. [The product team case] After final checking, statistical modeling, AB testing, and else, we can recommend something like “Let’s change our Pay button from red to green from 1 am to 6 pm because the color contrast has proven to increase our conversion rate by 20% and for 6 pm to 1 am reverts it to red because based on research, data analysis, survey, and ab test, it’s proven to increase our conversion rate by 10%.”

Conclusion

To my knowledge, a data-driven culture is a great thing that needs to be understandable not only by the analytics team, but also by teams that work with the analytics team, so we could know what kind of result we can expect.