What startups should learn from music about using data to grow

Based on my talk to growth stage startups at Scaletech (Toronto, Oct. ‘19)

Businesses are saturated in data these days. Every company is now swimming in data on every activity they do. And every company has analysts pouring over that data for one reason or another.  

But data is still often not driving growth. Why?

I don't think data or data talent are the constraining factors. I think it’s vision and focus around how you use data that is the constraining factor. I think data isn’t being used for the highest impact decisions.

So I looked back across all of the things I've done in so many different companies to use data, from fancy machine learning to simpler analytical methods and from marketing to pricing and I asked myself: What had the biggest impact? What was truly transformational? 

PART ONE: SNOWBALLING USING AUDIENCE STRATEGIES 

What made the biggest difference was when we used data to drive strategic decisions about growing audiences instead of tactical decisions. It was when we used data to be very clear on our existing audience segments so that we could better serve them. And when we used data to identify bigger adjacent audience segments that we could grow into. So I'm going to try to persuade you today that that is how we should all be using data if we want to have a huge impact with it. We should use data to drive what I call Audience Strategies.

Let’s start with the story of a brilliant businessman I had the pleasure of working with on his Audience Strategy to help drive amazing growth. David Guetta is a global superstar now, but he wasn't always a global superstar. He was a much smaller artist before 2008. 

Maxing out your market: In 2008 he had done an amazing job of building his expertise and fame with one particular audience segment that we called Nighthawks. They loved electronic music and, for them, music was all about nights out and socialising. He was a star in their world. But pretty much only in their world - most other consumer segments hadn’t heard of him or weren't that engaged with his music if they had. This was particularly true outside of France and was more extreme in the huge music markets of the US and Japan. There wasn’t much more growth to be found in his existing audience. Where would new growth come from then?

We showed there were much bigger adjacent audience segments that he needed to engage if he was going to grow. We used analytics to help him better understand his existing audience segments (and their limits), help him understand these adjacent, bigger audience segments and help him expand into them. 

How did we do that? We did that by taking audiences and clustering them into Audience Segments - distinct groups bound together by common needs. One big growth opportunity that stood out was an Audience Segment that we called Pop Idols for whom music was mainly about escapism. And right next to them were a huge but harder to reach segment we called Casuals who were average mainstream people kind of interested in music, but not that much. 

Both of these were much bigger audiences than Nighthawks. Led by David Guetta himself but with the support of his management, A&R and marketing team and underpinned, guided and focused by our Audience Segments he was able to grow from a Nighthawks-only ‘French DJ’ to a true global Pop Star. 

Snowballing, not Pivoting: But it was really important to David and really critical as a business decision that he do it in a way that wasn't a ‘pivot’. He didn't want to Pivot from being a Nighthawk-focused electronic music artist to being a pop star, he wanted and needed to be both. So he wanted to grow his market into these adjacent, bigger Audience Segments, but he wanted to retain his existing Nighthawk segment. I call that ‘Snowballing’.  

Here is one simple but brilliant illustration of how he did this, from the many examples spread across his music, products, press, marketing and events: His 2012 album, really cemented his position as a global superstar and the number one DJ in the world. He did something really clever with it. He led with music that excited the Pop Idol and the Casual segments - big vocal pop / dance where he partnered with a ton of amazing talent from the pop music world. But he didn't want to lose his old Nighthawk audience and so he also did a non-vocal version of the album focused on the Nighthawk-led electronic music scene that had been his core market to that point. So, again, instead of pivoting to go to the new audience and losing his older audience, he really cleverly like kept his old audience. 

Another great example of Snowballing comes from MaserClass. They make online classes with amazing talent like Martin Scorsese, Gordon Ramsay, Samuel L Jackson and Steph Curry. When I started working there they had built a great business selling classes to people characterised by one particular Audience Segment: people with very focused learning needs that we called the Explicit Learners. This segment had a specific need to study a particular topic in great detail. Perhaps they were learning a new career to mid-career and wanting professional development. Buying individual classes was perfect for them. That was our business model at the time.

As with David Guetta, pretty quickly we identified that there were other adjacent Audience Segments that were much bigger. They are best characterised as being ‘Lifelong Learners’ who are more casually interested in learning than the core Audience Segment. Unlike the core segments, they weren’t interested in buying classes one-by-one - they wanted to dip in and out of multiple topics. So we launched an ‘all you can eat’ subscription service which enabled these bigger Lifelong Learner consumer segments to access content whenever they liked. This subscription product also better met the needs of our core Explicit Learner segment (in part by unlocking the Lifelong Learner that was inside of many of them as a secondary need). We had unlocked another strong cycle of growth. We had successfully Snowballed, rather than Pivoting.

PART 2: DON’T CONFUSE BEING DATA-DRIVEN WITH BEING STRATEGIC

Hillary Clinton ran the most data-driven campaign there ever has been. She was more data-driven than any previous candidate. She had the best data and the best team of data scientists that has ever been gathered on an election campaign anywhere in the world. But they didn’t see the loss coming. Why?

The danger of personalisation or population averages vs. clustering

A great example of the dangers of two alternate approaches to clustering comes from the last presidential election in America. I say this based on my time working to build a data-driven culture in Democratic politics in 2005-07. This was when we knew data would be our secret weapon in winning elections but before I knew of the critical importance of looking at Audience Segments. We spent a lot of time trying to do two things that Seemed at the time to be great alternatives to clustering. They won us back the House and the Senate in 2006 along with a majority of Governerships. They helped Obama win the White House twice. But something was missing in 2008.

Personalisation: First, Democrats worked their socks off to optimise individual voter contacts, so that each voter got exactly the best treatment for them. Voters got the right messages through he right medium (door knocks versus phone calls etc) and with the right frequency. We spent an unbelievable amount of time, money and brainpower working out how to personalise campaigns and measuring the precise impact of doing so. Tactical optimisation to drive efficiency and effectiveness. Super important.

Population averages: We also spent a lot of time on trying to forecast the average vote across the whole population in a given state. Great effort went into creating the best polling data and using the variables in the right way to get population-level predictions of voting as good as they could be. 

What was missing? But Democrats didn’t see the loss coming. I believe Hillary needed Audience Segments to enable her to see the subtle but critical changes in voter intention that cost them the 80,000 votes that lost her the election. Simpson’s Paradox explains one reason why.

Beware Simpson’s Paradox (solution: #ClusterEverything)

As you can see from the David Guetta and MasterClass examples, clustering customers is a really exciting way to use data to drive an Audience Strategy. But there is another critical reason why you should #ClusterEverything: Simpson’s Paradox.

A clear pattern in individual Audience Segments might disappear when looked at in aggregate as population averages. Another reason to #ClusterEverything.

I suspect that looking at changes in behaviour amongst properly constructed Audience Segments would have revealed critical problems that were lost when looking population averages. I suspect that doing so would have rung enough alarm bells to shift enough resources to get the 80,000 votes that would have won Hillary the election.

Don’t confuse efficiency and effectiveness with growth

A big lesson I’ve learned in many businesses and that you can see in the Hillary example is to not confuse having loads of data and using it to do great work on efficiency or effectiveness with driving growth. 

There are lots of exciting ways to use data to drive efficiency. They are relatively easy to do. They feel necessary. They show quick results. It is addictive for companies to spend lots of time here. But you’ll end up with single or low double digit percentage decreases in cost / increases in revenue at best. There are more impactful ways to use scarce analytical talent ...

Along comes effectiveness, the natural next step from efficiency. These use cases are trickier and require more thought. They’re harder. They show results, but less quickly. The results are bigger though. Double digit percentage decreases in costs / increases in revenue for sure. Good times. But neither efficiency, nor effectiveness are going to 10x your business. That’s where Strategy comes in ...

To use data to drive strategy is even harder than using data for effectiveness. The results take longer to see. You can see why it often gets left to last and often finds itself de-prioritised. In part because it is often the same analysts, budgets, managers and leaders in charge of Strategy analytics as Effectiveness analytics and Efficiency analytics. They compete for time, attention and budgets. This is a huge mistake. 

SUMMARY

As the David Guetta and MasterClass examples show, I believe that using data to drive an Audience Strategy is the big win. I’ve always found Audience Segments to be key to doing so. They help chart a path to growth by Snowballing. And Simpson’s Paradox is another reason why you should use them. #ClusterEverything

As Hillary’s campaign shows, being data driven isn’t enough. You risk great peril by using your analytical resources on efficiency and effectiveness at the expense of Audience Strategies.