Competing With Data & Analytics
What to Read Next
There are 52 million Latinos in the United States, with $1.5 trillion of purchasing power.
Entravision Communications Corporation a Spanish-language media company, reaches about 96% of that U.S. Latino audience through its numerous television and radio stations and digital platforms. It uses that extraordinary reach to provide media solutions to marketers interested in tapping into the Latino consumer market.
Entravision’s sweet spot is its ability to offer hyper-local and regionalized channel marketing — what Latinos in Los Angeles are interested in versus what Latinos in Yuma or Tampa are interested in. But to provide more efficient and deeper insights to marketers, Entravision executive Franklin Rios decided to utilize empirical and transactional data obtained through licensing agreements. The increased data coming into Entravision’s traditional data environments caused bottlenecks — in processing power, in latency, in response times. As a result, the company embarked on a “big data” infrastructure implementation that included new analytics tools, the cloud and Hadoop.
The addition of external data, coupled with the new platform and new algorithms, resulted in fine-grained behavioral insights. But the story doesn’t end there. Entravision clients became interested in the new analytics results, outside of their implications for media buys. And with increased client demand for those insights, an analytics business division was created. In effect, analytics transformed Entravision’s business model.
In March 2012, Rios became the president of Luminar Insights, an entirely new business unit within Entravision that specializes in big data Latino insights, not just for media buyers, but also for customers in other industries, such as CPG (consumer packaged goods) and retail. Rios spoke with MIT Sloan Management Review contributing editor Renee Boucher Ferguson about the process of analytics innovation that led to the development of Luminar.
How did Luminar come out of a big data infrastructure implementation?
We launched the big data environment and we started showing results to clients, in connection to increasing their media budgets with the Entravision properties. And it was actually a client-driven request, where they were saying, “hey, can you do a predictable model for me on these other markets, for these other categories?” And the requests had nothing to do with media; it was purely to do with driving better efficiency, or purely for them to know what the market penetration was in a particular DMA [designated market area], and what a 2% or 3% lift could mean to them, and how the competition was stacking against them for the Latino consumer.
So, those questions became separate from our television or radio conversations, and became more of the analytics conversation. As a result of that — and we felt that we needed to have, for lack of a better word, separation of church and state — we wanted to establish a new brand in order for it to carry its own weight.
That’s where Luminar came into play, because we needed to create a new brand that was independent, so to speak, from the broadcast properties, and establish a new brand that could then be the source of analytical insights for the U.S. Latino consumer.
What was the process like of going before your board for approval of a new business unit based on big data? Was it a no-brainer, or something you had to really weigh?
It was an interesting process, because I first approached my chairman, Walter Ulloa, and I said to him, “Walter, what we’re doing for the media properties is creating a lot of traction and it is working. Everyone can see that it’s working, because it was producing results internally.” I said, “I think there is a market for this, based on the conversations I’m having with the clients, where this could be a stand-alone service with its own brand and its own entity.”
So we started those conversations. I went off and we did some market analysis and segmentation, in order to find out how much of an opportunity was out there for this. We realized this could be a whole new revenue stream for the company. A revenue stream that is very different from our traditional media revenue stream, because this is very much a technology and professional services engagement type of model, versus media where you’re selling spots and dots. And so it is a departure from our core business in that particular area. But as it relates to the entire strategy of the company, it is very much complementary to the fact that we’re serving the same end customer, which is the Latino consumer.
So, we presented it to the board. The board again deliberated on the pros and cons: the revenue upside, the fact of having another business unit in the company doing a completely different type of approach to how revenue is created, and to how Latino consumer information is being gathered and reported. And between the first time that we approached Walter to the time that we had the green light from him and the board, it was about five months, where they said, “Fine, run with it, establish the brand — establish Luminar,” which coincidently, means to illuminate or enlighten in Spanish, which we felt was very appropriate to the services we provide our clients.
Well, we didn’t have a name. The code name for the project at the beginning was called “LatinMath.” We engaged with an agency to help us come up with a good name, and that’s where Luminar came about. In the U.S. now we have 20 people. We have nine data scientists and mathematicians in Mexico, and seven in Argentina.
Now that you have transitioned from focusing on internal analytics to delivering big data as a service, what does that shift mean for your organization?
At the beginning, people were scratching their heads saying, “Why are you doing this? Is this really going to create any value to me and to my customers?” And the only way to prove that is to truly demonstrate the value when you’re in front of a customer and they’re starting to open their eyes and they’re like, “This is great. We want more of it.”
So at the beginning, it took a couple of internal champions and a couple of people we had to engage with in a more individual, one-to-one manner, in order to have those small wins that then would propagate throughout the organization. Because success gets around really quickly. From that perspective, it was very good.
From the perspective of the company, our internal operations — finance and ops — they had to think differently, because when you’re not selling just media and now you’re selling analytical models and professional services and data environments — big data environments — you have to change the way that you capitalize the investment.
So it has been a shift in thinking for multiple areas of the organization, including our board. They have to get their heads around the fact that there is an opportunity here, based on what had been created. And now that that has been established and it’s ongoing and it’s a matter of growing the business, it’s a matter of the board saying, “Okay, how fast can you do it?”
Now there is buy-in. Because we have proven, through the engagements of several customers, that the model is working. And customers are coming back for more, asking for more insights.
You mentioned that there was some one-on-one that you had to do to change the thinking around Luminar. Who did you have to approach to start that shift in thinking?
One of the first folks that we had to engage with internally was the existing head of research inside the organization. Because everything that this person had been using, forever, has been the traditional highly sampled data. So this person was thinking, well, how much different is this going to be, and what value can it bring to the national sales reps or the local sales reps, to be able to use this thing to sell more? So that was one person.
The other was specifically the general managers and the integrated salespeople at the local level. So we had to choose a couple of DMAs to concentrate on, and to prove the concept with, in order to have some success stories. Because doing a 50-DMA approach would not have been as efficient as just concentrating in two DMAs, in a couple of categories with those DMAs, where we could prove that the model worked. So it was truly a step-by-step process.
It sounds like it was a relatively quick step-by-step process, though.
It was quick because we had executive support. If Walter wouldn’t have bought into this and supported it wholeheartedly, internally, it would have not worked. So the executive team, the board supported this, and allowed us to explore and allowed us to reach out to the right people in the organization in order for this to be tested.
What sort of biases did you encounter as you evangelized the prospects of Luminar?
The biases that I saw were in the media industry, where there are standards of how things are done. And deviating from those standards is usually a no-fly zone. What played into this was the cultural, the media environment — and the traditions that the media companies have.
Just going back to, for example, the head of research. He had done all the analysis that he could with the highly sampled data that he had, for the longest time. And that’s how he reported to every local DMA. And for me to come in and say, “Look, there’s another way we can do this that will give us better results, better data and better everything,” he said, “Look, I’m already using these other tools that is the currency in the market. So you’re telling me you’re going to create a new currency? How are my clients going to believe this new currency, if the currency that has been used for the last 30 years has been working?”
In our research we’ve looked at those companies that are “digital natives” that were founded on the concept of data as a core asset. And it sounds like that’s where you are starting from with Luminar. What sort of advantage does that provide you?
Absolutely, we consider data a core asset and a core concept of Luminar, due to the fact that it is the genesis of how we can get to the end product. But truly what creates the value, the incremental value to the organization, is when we start applying our analytical models into the data. The reason why we have our data scientists and mathematicians in Mexico and Argentina is because, when you look at data through a cultural lens, it becomes more insightful than just looking at the numbers.
Because there are a lot of things that aren’t taken into consideration beyond just the numbers, particularly the cultural differences and cultural nuances that must be taken into consideration when you’re analyzing data, specifically for U.S. Latino consumers.
How is it that you apply that cultural lens? Is it the algorithms that you develop?
Yes, exactly. It’s through algorithms, and it’s also through seasonality. It’s through a lot of components that we get to have this data even richer and more insightful, because of the people that are working with it. So at the end of the day, just being an analytics company that focuses on the Latino consumer is not enough, if you don’t have the cultural insights in order to inject into the model.
I can guarantee you that we have engaged with large Fortune 500 companies that have large analytical teams, in-house, and that work with multiple different types of partners and vendors. And when we take their analytical models and we add our cultural models into it, the results are richer and different. So we sometimes end up partnering with the internal analytical teams of these companies, CPGs and retailers, in order to enrich the way that they are analyzing their data.
Have you seen any other organizations that add this sort of cultural lens to analytics?
Well, the closest that companies have gotten to any type of cultural lens has been a process of what they call ethnic encoding, which is to just identify a Latino consumer. And that’s really like step one in 50 steps, right? Because just knowing who is a Latino consumer in your internal environment does not give you a good overall look into how those Latinos are behaving outside of your environment, and how they live and how they’re purchasing things.
And so the cultural lens component on top of that adds a lot more layers to it, because it’s easy to identify a Latino if you are using a first and last name combination, but it’s also very dangerous, for example, because you can get into a lot of false positives. But if you start identifying the Latino consumer by their purchasing behavior, by the contents of their basket, by how they eat, by what they watch, by what they’re purchasing in subscriptions — whether it’s magazine or television — and all of those components put together, then you can have a better accuracy in truly identifying a Spanish-speaking-dominant Latino versus an English-dominant Latino. Because culturally, we still eat according to our cultural preferences.
Particularly since you’ve used analytics to create something completely new for your organization, at what point do you have to go back to the drawing board to reinvent how it is that you’ve used analytics to innovate?
Well, the interesting part about analytical models is that they mature, right? And they need to continually be fine-tuned in order for them to become more efficient and for them to uncover better insights. So every time that we engage with a client, we are fine-tuning the model each time, because we learn something new. There’s a new question that we haven’t been asked, that we need to go find the answer to by adding to a model or building a brand-new model, in order to get to the answer.
So from that perspective, yes, we have to continuously — I would say not innovate, but I would say we have to continuously evolve. Because innovation to me sounds more of what we just did, which is create a brand-new business unit for a traditional company that is looking into a whole new revenue stream for the Latino industry, versus evolve, meaning take what we have created and continue to make it better and better, in order for it to become more relevant and in order for it to have a faster return on investment for the clients.
So, that evolution needs to continue. We cannot sit on our laurels and pretend that no one’s going to catch up. Someone will catch up, and there’s going to be something out there that is going to make this faster to market, because big data is evolving too. Big data is moving, and there are still a lot of players out there. And there is still a lot of consolidation that needs to take place in the industry.
I think the number one thing is for us to be flexible. If we can be flexible and evolve at the same time, I think we’ll be able to continue to be spearing ahead.