The Future of Analytics

Augmented, Personalized and Omnichannel

“It’s very straightforward. I want to push a button and get a ride. That’s what it’s about.” Travis Kalanick, Uber co-founder

This very statement represents the most basic need of human beings — to make things easy for us — the desire to improve our efficiency has been the single biggest driver of our progress as a race. So naturally, when Uber came out, it was an instant hit with the users. The global demand for cabs ever since has increased many folds and Uber has become an epitome of one of the most complex and sophisticated pieces of technology delivered in the most simple and intuitive manner such as the push of a button.

As a digital professional, not a day goes by when I either don’t read, hear, or preach that we need to be more data-driven.

It’s no surprise that as computing and storage have become cheaper than ever, we are storing and processing more data than we ever did. It is estimated that there are 2.5 quintillion bytes of data created each day and 90 percent of the data in the world was generated over the last two years alone.

Given this volume and velocity of data, the desire to build complex analysis and generate actionable insights has also grown substantially. And we, the humans are struggling to keep up — toggling between dashboards, drudging through reports, frustrating gaps in data, and too little time to process it all, are becoming too common in workplaces. This is despite us having the most sophisticated analytics tool such as Google 360 and Adobe Omniture for tracking and Tableau and Power Bi for reporting at our disposal.

Despite billions spent on Business Intelligence (BI), adoption hovers at 35% percent, according to Gartner, leaving millions of business professionals without access to the information they need to make smarter decisions.

What's the problem here?

I’m arguing that analytics doesn’t have a ‘training or people’s lack of desire’ problem as it is often suggested, but it’s an integration and a user experience problem.

Let’s take e-commerce analytics as an example. The 3 reasons why users are struggling to become truly data-driven using traditional platforms are:

1. More data: Our ability to collect and store data on every step of the consumer journey has grown exponentially.
2. More complexity: The consumer journey in today’s omnichannel (with ever-growing channels) is so complex and dynamic that it’s extremely hard to manually ascertain correlation and causation of various events and outcomes. E.g. multi-channel attribution is still hard to get right.
3. Need for speed: from insight to action is greater than ever before. Gone are the days, when campaigns could be devised over several days and weeks. The competitive nature of e-commerce requires real-time access to insights, which drive instant, automated, and rule-based actions.

How do we address this challenge?

Before we delve into the solution, let’s define the scope. Continuing the theme of e-commerce, there are 4 types of insights that the users need:

DESCRIPTIVE:

These are the most basic set of insights that describe “what happened”. A timely snapshot of key metrics that enable a user to understand the health of the business. Exhibit 1 is an example.

DIAGNOSTIC:

This set of insights begin to explain ‘the why’ of ‘what happened’. For example, you can see that your Customer Acquisition Cost (CAC) has been dropping, but don’t know why. The following insights (Exhibit 2) unveil that the contribution of acquisitions via email has been growing, and since it is one of the cheapest channels of customer acquisition, your average acquisition cost has been declining.

PREDICTIVE: This is where things get interesting. An intelligent system can use historical data, run statistical models, and make predictions on key future outcomes. A recent example of this is the latest addition to predictive analytics in Google 360 suite, where you can now create target audiences by purchase probabilities. See Exhibit 3 below.

Taking it one step further, you can identify channels that show the highest probability of purchase and then allocate a greater share of the marketing spend on such channels. See Exhibit 4.

PRESCRIPTIVE: This is Nirvana! The system would analyze complex data, establish causal relationships, and recommend the next best action to achieve your business goals. This system would act as a business analyst who analyzes data, extracts insights, and prescribes action. Exhibit 5 shows a good example.

To really get an integrated and seamless experience covering all of the above-listed types of insights, the analytics capabilities will need to be Augmented, Personalized, and have Omni-Channel delivery.

1. Augmented Data & Analytics: The current descriptive analytics and insights across customer, product, marketing, platform, and operations will need to be augmented in the following ways:

a. Relational — an intelligent engine that can auto-build relationships between key variables and run real-time queries in the form of “natural language”, can eliminate the modeling work required for legacy BI tools and return queries at lightning speed, just as Google or Amazon do. This would save analysts countless hours of work and resources.

b. Contextual — examples of context would be ‘who is the user’ and ‘which device are they using’. Just as when Spotify recognizes that you are accessing it on Carplay, it recommends ‘songs for the road’, the analytics platform should recognize that a sales analyst is using his Alexa device, so should highlight what she would care for most — key changes in sale for the day, week, and month.

c. Predictive — the engine needs an element of AI, that enables it to use historical patterns and future signals (both first and third party data can be used as inputs), to predict key outcomes such as sales growth or conversion rate. The system would improve over time as it learns from the outcomes vs. predictions it makes.

2. Personalized insights

These insights would be driven both by a user’s explicit and implicit inputs. For instance, if I’m a Product Manager focused on improving conversion or adoption, then the system would not only automatically track and update conversion rates, but would also recommend looking at other metrics or analyses based on my role. Using collaborative filtering, if the system knows that other product managers have used Cohort Analysis in the past, it would generate insights on Cohort Analysis and automatically recommend such insights.

3. Omni-channel Delivery

Finally, this advanced analytics system must have the ability to tailor and deliver insights across all the key channels and formats — Dashboards in BI tools, Alerts in email and Slack, Search that uses natural language and AI assistants that can deliver on-demand insights through voice, video, or any other desired format.

While there are players who have built one or more of these capabilities independently, such as “Thoughtspot” for Search-driven analytics, there is a need for integration and creating a seamless experience. Imagine if we could eliminate the need for a business analyst, who creates specialized and complex reports, and instead, empower every front line user with real-time and personalized insights with recommended actions that combine human and artificial intelligence.

When this happens, “be data-driven” will no longer be a motto, but will become a way of life, just as pushing a button and hailing a ride has.

Amit Rawal is a Sloan Fellow at Stanford’s Graduate School of Business. He has spent the last decade in building and scaling e-commerce ventures for 40%+ of the world’s population. At Stanford, he is focused on bringing together tech, design, and data to create joyful shopping experiences. He is a data geek and loves tracking all kinds of health and wellness metrics. He can be reached at [email protected].

Links: Linkedin, Twitter, Instagram, Website