Jonathan Lakin : “Companies that design for difference win every time”

When it comes to customer analytics, most companies think of what that data will enable them to do to their customers: sell more, increase plans, reduce costs, and so on. Jonathan Lakin (GEMBA ’03) had a different idea: using analytics to listen to customers and understand what they’re trying to do – their intent. Thus, his company, Intent HQ, was born in 2010. Advances in artificial intelligence (AI) make it possible to sift through billions of signals from millions of customers to determine what matters to each person, and operationalize that insight.

It hasn’t been easy: “The pace of AI is outstripping our ability to leverage it,” says Lakin.

The challenges are technical, requiring an entirely new way of processing data, as well as organizational, as companies need to change their old ways of working to take advantage of the AI opportunity.

There are also challenges around privacy, which is where Intent HQ has focused its efforts to develop customer AI that “redefines what’s possible both from a customer insight and a privacy perspective.”

Intent HQ sponsors IESE’s Chair on Changing Consumer Behavior, currently held by marketing professor José Luis Nueno. In this interview with Prof. Nueno, Lakin elaborates on managing these challenges when “playing to change the game.”

How do telcos, one of your key clients, illustrate the shift that many companies need to go through?

Telcos are trying to reinvent themselves and they look to firms like ours to be change catalysts. But reinvention is hard, especially for a sector that has been around since the 19th century. And while telcos’ reputation for “invented here” might have been true in the past, today it is Apple, Samsung and the OTT players that get the glory and the love.

Increased competition and 4G maturity have made commoditization more prevalent. 5G requires a considerable capital investment whose benefits won’t be immediately transformative. Wall Street hasn’t helped, either. By still heralding telcos as blue-chip dividend stocks, it puts unrelenting pressure on management to deliver consistent cash flow and margins.

All this requires a regime change. Because many of us in Intent HQ come from telco backgrounds, we get it. We take a practitioner approach to implement our AI-based solutions, a few use-cases at a time. The capabilities that we provide are helping to break down age-old company silos, forcing telcos not just to rethink their business models but to turn them on their heads. We help them establish an ecosystem to co-generate value for end customers, because collaboration and partnership now hold the key.

What else might companies be missing?

Many consumer organizations have forgotten about the data gold sitting right under their noses — it’s just hidden in a lot of noise.

Our customer AI looks at the customer activity data through a number of different lenses. Our IntentGraph has a semantic and a behavioral component. The semantic graph looks at language, so when someone says “Queen,” for example, we know whether they meant the band or the monarch. The behavioral graph looks at people who exhibit similar social behavior, such as going to the same type of restaurant or being persistent complainers.

We start with the data we have — every tiny signal, every interaction — and then build out from there. It’s combining human behavior with quantitative measures that’s key to business transformation, particularly in detecting which customer behavior drives your KPIs.

What common mistake do you see analytics teams make?

Using averages, such as the average of all the interactions that happen in a day. While storing an average might seem a reasonable trade-off to spending money on storing every single piece of data, for many kinds of analyses, particularly those involving humans, we are unwittingly trading the cost for accuracy. Think of all the ways your own behavior is a bit quirky or different from the average person. By using averages, the products or experiences we design will be naturally less compelling.

Companies that design for difference win every time. The challenge is twofold: sifting all the data to find the key customer signal, and then operationalizing that insight at every touchpoint. Achieving this requires a completely different approach, not only to how you compute your data but to how you orchestrate it across the plethora of platforms and systems that make up today’s complex IT landscape.

How are you dealing with those challenges?

Think about it like the MP3 player. The core innovation was the algorithm that compressed the raw music file while retaining close to the original sound quality. That reshaped an entire industry and changed the way we relate to music forever. The same will be true for encoding raw customer activity data. To make data easy enough for data scientists to use every day, we need to compress and transform it without losing the meaning.

This is one of the things we do at Intent HQ: taking raw customer activity data, finding the signal, deciphering the meaning, removing sensitive data, turning it into vectors (for greater privacy) and making it available for data scientists.

This new behavioral dataset can then be used to improve all the customer propensity models across an organization. This means organizations can start to talk to their customers more like a friend rather than as average revenue per user.

If organizations prove to be as thirsty for data as people are for music, the change for industries will be as great as with the MP3, with competitive advantage for first movers.

How about privacy concerns?

Customers want to be treated personally but they also demand privacy; it’s not an either/or. Most organizations, however, think about privacy one-dimensionally: as compliance with legislation. But privacy is ultimately about trust: demonstrating it, rewarding it and building it directly into the customer experience.

Building it into the experience fabric means thinking about privacy at every step of data flow. For example, in the data layer, automatically removing sensitive data classes, like politics or health, before they can be used. In data processing, preventing the processing of data for restricted categories. And at the customer touchpoint, offering a graduated consent approach that builds a greater sense of value exchange.

We think it’s crucial to go beyond the legislation and offer customers “the right to reasonable inferences,” essentially sharing the inferences we make about customers, both so they can correct them and so that organizations can be transparent, which furthers trust.

This sounds like a complex, expensive undertaking. How do you do it?

Here’s a sobering statistic: 87% of our models never make it to production. Most of the complexity originates in the data layer for machine learning.

Given our focus on customer AI, we first needed to build a platform and tools to make it easy to experiment and move seamlessly from discovery to production.

Having learned from that process, we next built a new kind of platform with an entirely different architecture to enable near real-time data with analytics, effectively marrying what have historically been two separate domains: transaction systems and analytics.

Now, we can move to production in days and weeks, rather than months. It’s a long and painful but ultimately rewarding process.

Did COVID-19 affect your models if they were trained on prior behavior?

Churn is an area that saw obvious impact. In lockdown, churn was significantly reduced across most industries. For customers, changing their service wasn’t at the forefront of their minds, and the friction of making changes with stores closed was higher. This meant figuring out which customers were suffering hardship and providing solutions to mitigate churn later.

This goes back to what I said earlier: focusing on understanding what a customer is trying to do, rather than just analyzing their current activity.

What management profile is necessary for what you do?

Talent alone isn’t sufficient. You need to marry technical expertise with data curiosity and commercial savvy.

When I did the Global Executive MBA at IESE, I recall a finance professor talking about developing a “smell for the numbers.” I loved that concept. We look for people who can “smell the data,” which is really about having an instinct or natural curiosity for finding the insight and stories in the data, and weaving together the many strands into a picture of the future you want to create. This kind of data curiosity is in short supply, and there’s always a strong demand for it.

Which is more important for a manager: good instincts or lots of good data?

Both! You’ve got to have some intuition to select the types of data that will give you actionable insights. Knowing which data is probably good, which is unreliable, when to reach for more or when to make the decision: these choices require instincts.

On the other hand, without good data, you’re lost. Nobody wants to be the proverbial drunk, looking for his keys only under the streetlamp because it’s too dark to see them anywhere else. You need to shine the light where it is needed — by pulling data together from diverse sources into a single system to create new insights. Or you might use existing data in entirely new functions, such as using data from operations to drive marketing.

It’s wonderful how often the arrival of more data yields better insights, which will shape your instincts and amplify what those instincts can achieve. Good instincts will react to new information. Instincts can guide choices about data governance as well as selecting the actions the data will drive. Then, those actions will generate more new data and, if done right, more new insights. We learn and improve.

This goes back to that rare skill of data curiosity. That’s what good management is all about.

A version of this interview is published in IESE Business School Insight #156.