#DataStories is an ongoing series where we interview the people doing incredibly interesting things with Twitter data. This week we’re speaking with Shawn O’Neal of Unilever about the work his company has done to build a mature insights capability with Twitter data.
This is the overarching program for how we look at all our “people data”: traditional CRM, social, contact with customer service, any information from marketing initiatives or consumer research. We wanted to create one IT structure and one focal point for the organization so everyone knows where the data sits and how to make use of it. The PDCs are actual physical spaces. The social listening aspect is heavily fueled by Twitter data.
Twitter data is important to us because it gives us a window to what’s happening right now in the world. When one of our brands has questions they want answered, they don’t have to wait six weeks for traditional market research. With Twitter data, brands know whether an ad is working right away or whether to investigate feedback for a new product innovation. Twitter data gives us the first hint at what’s popular and what’s not, and a view into what people are saying and how they’re saying it. (Editorial note: Unilever makes some of the world’s most well known consumer brands ranging from Dove and Axe to Lipton and Hellmann’s.)
When Stan Sthanunathan (Unilever SVP of Consumer & Market Insights) came on board, he brought a vision for what social listening would be. Stan wanted to do for social listening what the first iPod did to MP3 players: it had the same functionality as other MP3 players, but it made music sexy and interactive. Social listening by itself can be pretty thin, but if you add other elements that bring it to life – such as weather data and sales data – you get a lot more context and make it more interesting for the brands. Stan also brought the idea of having a physical space where brand teams could engage with what’s going on socially.
There are several factors for this. First, you have to dig deeply into social listening to get value. If you only give existing tools to brands, people may not know how to use them or the tech might not be very deep. This sets up an environment where people are less likely to understand what’s happening. If you put brand managers into an environment like the PDCs with people who know how to dig deep into data with them, it unlocks the context they need.
If I do a focus group with 30 people, it’s arguably statistically significant. But if I can give you 1,000 Tweets and show you 10 of them that say something specific about an ad, the brand team is far more likely to change the advertisement. The Tweets add humanity to numbers, to charts and graphs, and statistical correlations.
Social analytics allows us to make qualitative research quantitative. We’re able to step into people’s lives. Rather than rely on just focus groups, we are able to do research with hundreds of people.
Our favorite story is from one of our ice cream businesses, Ben and Jerry’s. The team came to us and said look, it’s premium-priced. We’re always arguing about whether we’re too expensive or too cheap. Can you tell us what the consumer thinks?
We did some sentiment analysis on the social media conversation, which was generally pretty positive. We overlaid the price of the product in the stores, and what we saw was that the while the price was changing, there was no correlation between price and sentiment. So price didn’t seem to be a big concern to consumers.
We also did some qualitative analysis as part of this process and saw that while a few people were talking about price and a few about taste, there were huge amounts of conversation about buying the product, e.g. “Should I go buy some right now?”
This caused us to look more carefully at sales data, and we saw that loads of people were buying on Saturdays. But when we looked at social conversation, we saw lots of conversation on Thursdays and Fridays. We then saw that some conversation spikes were bigger than others, and when we overlaid our weather data, we saw that there were big spikes when it was raining. We then looked back at those conversations, and we saw conversations like, “it’s raining, i’m going to go buy some ice cream and watch a movie.”
Through Twitter data, we discovered ice cream was a premeditated purchase.
This discovery changed how we approach our marketing. We had the belief that ice cream is an impulse category, so we spent money on in-store activation rather than brand building. We also used to think we should advertise on a Saturday, but we decided we should really advertise on a Thursday because that’s when people are thinking about purchasing.
Stakeholders from our various brands come to the PDC with questions like “We’re worried about the price of this product. Can you help us understand if everyone else is worried as well?” Based on a request such as that, we first investigate and may - for example - find that Twitter sentiment around the price is actually flat. This, however, then leads us to explore even more pressing questions such as: “If they’re not talking about price, what are they talking about? And when are they talking about this? And finally, why?”
We don’t want a world where the business asks a narrow social media question of the PDC. We want people constantly going back and forth between the business and the PDC thinking about related and even larger questions. The goal above all else is to actually learn something.
It’s a journey we’ve been on for a while. The spectrum for analytics ranges from highly technical (write code, do data systems for a living) to people who have never written code but consult on the business demands and desires.
Previously, we’ve had those two sides talking but not understanding each other. Right now, some of the best business people understand coding, statistics and modeling well enough to understand the limits of their questions and their asks. When building an analytics team, we want as many people across the spectrum as we can get.
For us, social listening tends to be a skillset that you can train an entry-level person to do and they can grow into it. Yet people who have it done it for a few years get better at storytelling with the data. At first, you can pair people with the baseline analytics skills and those who can do the storytelling. You’ll also want to add highly technical people into the mix that can build tools, apps, models and more.
You also want people who REALLY know the business. In Unilever’s case, the existing Consumer & Market Insights people we engage and bring into the social listening world are acting as the business liaisons who help us understand the brands.
Ultimately, we’ve had to hire a lot of specialists in social listening and highly technical people who can work with data in a “new world” data structure.
Stan brought with him a vision to inspire, provoke and transform – people who do those three things are the ones who change the world. People who tell great stories as a skillset are seeking to inspire others. Sometimes and quite often, you use storytelling to provoke a response…and the goal of provoking is to transform. If you’re always doing those three things, you’re going to be successful.
Our team never hires someone for functional expertise alone. I believe in hiring smart and ambitious people who have long futures in an organization. I hire on potential knowing that they can pick up skills. Most of what I do is effectively changing the rules, so if I hire someone just on skills, I spend a lot of time trying to get them to change. We may hire 1 in 5 people who have the specific experience, and those I hire — in part — to keep the others from running off the tracks.
Thanks to Shawn O’Neal of Unilever for taking the time to speak with us.