We all agree that Dashboard does not serve as an effective tool for taking action, so, where is the future of Product Analytics solutions headed?

By 2027, it is forecasted that the global market for product analytics would be worth $30 billion.

As a result of the demand for businesses of all sizes to understand consumer behavior and generate more revenue from their data, the market is experiencing such phenomenal growth.

The current state of data products is insufficient to provide companies with actionable next steps in data.

Which direction will product analytics take in order to provide us with more value from data in the future?

My thesis.

It is essential to integrate more advanced algorithms, explainable insights and actionable recommendations into data products to allow us to maximize the use of the data.

I would like to share three scenarios for how the industry can evolve from where it is today to where it can be in the future.

1. Metric, event or query-based analytics.
2. Algorithm-based analytics.
3. Counterfactual explanations.

1.Query-based analytics.

The first scenario is where the industry is already operating. With the current product analytics solutions, it’s easy to visualize basic metrics, dimensions, and query data to answer simple questions, here are a few examples:

-For the past 30 days, how many users have logged in?
-What is the number of daily active users for the last 3 months?
-What is the retention curve for users of mobile apps versus web apps? How does using a X vs Y feature affect retention?
-What is the drop-off for Z feature?

These insights don’t give us a deeper understanding of how or why it is affecting a particular behavior in retention or revenue, for example. What level of actionability do these questions or visualizations provide? As a result of these insights, what can we do to improve our product next?

2. Algorithm-based analytics.

To answer more complex questions in our companies, data scientists analyze information and deploy machine learning models to provide answers that enable us to make decisions.

For better action, what types of analytical products should we design?

It is important for business people and product managers to understand insights quickly in order to accelerate the decision-making process.

The shift from metrics-based or events-based analysis to algorithms-based analysis is where the next scenario of actionable insight takes place.

Can we build analytics products to answer more specific questions?
Could algorithms be applied to the data to answer more advanced questions, this is an example:

-In the first 10 days after creating an account, what is the difference (event-based activity) between those who upgrade and those who don’t?

Would data products be able to automate and provide a response to this question?

Amplitude is an example of a product in the industry that incorporates advanced analytics.

Among other features, Amplitude’s Personas chart groups your users into clusters based on the similarities of their event behavior. Users who behave the same way will be placed into the same cluster. There’s no explicit, pre-specified rule that defines a cluster.

The Personas chart you can quickly do exploratory data analyses of the ways in which your user base navigates your product. It can help you surface similarities between user cohorts you may not have thought to look for. And it can guide you through the process of creating a comprehensive set of user personas for your product, which you can then use to drive engagement and retention.

The results most companies will get will be far beyond what they were once receiving through query-based analytics, where they had to figure out what data to query, how to correlate it, and how to make sense of it.

3. Counterfactual explanations.

A counterfactual explanation is a concept in artificial intelligence that reveals how little changes in input data will generate a different result, allowing us to understand how an algorithm’s results might change if certain initial conditions were different. For example, It would be extremely valuable for a person who has been denied a loan to receive information about the variables that have the greatest influence on the loan approval.

Can we design data products that implement predictive analytics to play with specific levers, parameters, what-if scenarios, and see what the projections may be based on the input data you provide?

With data modeling, you can predict how each of these levers, dynamic parameters, will affect the outcome so you can take the most impactful next steps.

The following are some questions we might explore in this scenario:

-When new users take an Y key action at least X times on the day they sign up, are they more likely to continue using it after 30 days?

-Is it possible to identify churn behavior one month after an account is created? Are there any key events that can be activated to increase the retention of these customers?

-Are there any customers with a high probability of churning in the next two months?

-Which actions or events influenced users’ decision to finish onboarding?

 Are you building a data product? How are you utilizing machine learning algorithms to answer advanced questions?

In what direction do you see product analytics heading in the future?