Can social media drive revenue? The answer varies from a frustrated no by many direct thinkers—those marketers sharply focused on conversion and attribution metrics—to an emphatic yes by brand thinkers facing unprecedented reach and frequency. We’ll use the Hidden Markov Model (HMM) to show what’s possible in social and appeal to both sides.
Direct vs Brand
Direct thinkers generally run direct response efforts and understand transactional eCommerce, catalogues and home shopping. Their direct attribution models are great for funding decisions around keywords or shopping cart optimization. And because they live in a world of measurement and attribution they often carry strong negative opinions about social media marketing. They’ll point to reports like IBM’s Black Friday Report 2012 where Twitter and Facebook referred 0% and 0.68% of Black Friday sales respectively as proof that social media doesn’t sell.
Brand thinkers love social. The ability to generate awareness, recall and preference to millions of fans and followers is too strong an opportunity to miss. Certain industries—retail and fashion in particular—know they work within a highly judgemental social component. People want to wear or not wear certain things, own certain brands, and be seen shopping at certain stores. This all has a social context and platforms like Facebook and Twitter are ideal for spreading it.
Hot or Not?
For simplicity we’ll assume a customer is either in the mood to buy something or not; there’s no lukewarm state in this scenario. So our Hidden Markov Model shows two hidden buying states, cold and hot.
Illustration: Hidden Markov Model (HMM)
As data scientists we don’t often talk publicly about the work we’ve been doing to distinguish between the two states but we’re about to break that embargo. We can infer a person’s state by examining indicators like clickstream pattern. Some businesses have done well by building machines aimed at sending the right message at the right time but those are quite rare.
Each state emits observations and carries a probability. Somebody in a cold buying state, for instance, is far more likely to be socializing (95%, to use a hypothetical probability) than searching online (4%) or visiting a brand website (1%). These are the grey arrows above. The searches and site visits could be for entertainment or information but since they’re in a cold state they aren’t driven by a desire to buy something. Looking at the orange arrows, those in a hot buying state are far more likely to visit a brand website (60%) or a search engine (30%) to research a purchase. There’s some socializing to read reviews or ask for advice but it’s much lower (10%).
People in a hot buying state can convert to cold through distractions or interruptions—perhaps as much as 10% (see the blue arrows). But cold-to-hot transitions are much less frequent occurrences.
From a direct marketing point of view, brand messages served to people in a cold state are wasted. In real life shoppers may pick up a pack of gum when they run out to buy milk but there’s very little impulse buying online. It requires a cold-to-hot state shift and the motivation to search or visit a site.
Brand thinkers argue otherwise. That new BMW owner may be 42 years old today but he was just a kid when he saw his first BMW ad on television and that generated latent demand. That one ad plus all the other instances of brand messaging along the way generated awareness, recall, preference, and ultimately loyalty and retention. There’s also an element of social contagion, aided by natural homophily, that makes a brand marketer’s efforts that much more powerful. Marketing Science contains well documented evidence that social effects are high and vital for new product adoption. And thanks to social media’s growing reach, these effects are increasing by the day.
This is where data scientists and the digital analytics industry can play a key role. We’re excellent at recording and predicting the performance of direct attribution channels. There’s certainly room for improvement but the optimization opportunities are generally not limited by technology or methodology. And it enables statements like IBM’s 0.68% number, which is impressive because it’s so precise and timely.
Social media can drive underlining preference, loyalty and retention. With permission, it can be observed and optimized. Brand thinkers inherently get this. The results aren’t nearly as precise and blanketed as direct marketers can produce but it’s still a highly measurable area all marketers are sure to understand. And brand leaders on both sides are tapping into a growing supply of technologists and data scientists to turn this opportunity into revenue.