Are we becoming too obsessed with big data and The Algorithm? Are we missing human input?
I thought this recent article in the HBR ‘Data Is Great — But It’s Not a Replacement for Talking to Customers’ provided a balanced look at our sometimes evangelical obsession with big data (versus more traditional research techniques) in understanding human behaviour and the motivations behind decision making.
Like a lot of sectors, the world of research/insights can become siloed based on preferences for particular approaches. Increasingly it feels like it is the big data/analytics crowd versus those who conduct what is generally called social or market research (i.e talking to people).
One or the Other
I have seen a number of consultancies across a range of sectors recently who not only espouse the virtues and advantages of using big data to understand what people do and what they might do in the future (fair enough), but passionately denigrate techniques that involve actually talking to people to aid this understanding. (It should be noted that some of the research traditionalists can be similarly averse to harnessing the power of big data.)
We’re all liars
On one level big data evangelists have a point. Asking people questions (whether through surveys, focus groups or other means) can be fraught. There are a variety of reasons for this, but mostly it is because people don't always tell the truth (whether consciously or not).
We like to appear better than we are. "How many standard alcoholic drinks in a week do you consume?"
We feel we should give socially desirable answers. "I strongly agree we need to reduce carbon emissions to zero by 2050."
A difficult or sensitive topic can make us answer in a particular way that saves embarrassment.
We believe we can influence the outcome of the research in our favour. "If I express my opinion even more strongly perhaps things will change."
We simply can't remember. "How many times have you exercised in the last 12 months?"
Our answers can reflect the current media and broader discourse but not longer term views, e.g. attitudes towards the Monarchy based on the Harry and Meghan interview
And getting representative samples of people is getting more difficult. See how pollsters got some elections wrong over the last few years.
Hating on The Algorithm
Big data's great advantage is that it can reduce some of this human error (your FitBit knows how much you have exercised and in the case of social media they know what absolutely everyone on the platform is doing, not just a sample). But in many cases it is still inferring information about you. Yes, big data knows if you watched the Harry and Meghan interview and yes it knows your social media activity and the general sentiment of your interactions with friends while you watched, but does it really know your views on the Monarchy? The Algorithm would probably say yes, but would it be right?
Perhaps due to recent scandals around how tech companies use our data, or perhaps because when you buy a pair of jeans online you still see the same ad for them pop up on your social media feeds for weeks, it seems we increasingly loathe The Algorithm and the big data behind it. Even if we rarely give it credit for the way it and the big data behind it actually helps us - Google Maps finding the quickest route when you're driving, Spotify offering up a suggested playlist that really bangs etc. Oh, and QR codes for contact tracing for COVID-19 seems to work pretty well!
When The Algorithm is good it can be very very good, when it's not it can make you wonder what all the fuss is about.
Why don’t we have both?
The work I do for the film industry is a good case study of how traditional research techniques can marry (and sometimes clash) with the smart analysis of big data.
There are many ways to understand how an audience engages with a film. Some of them quite forensic and borderline Orwellian. You can film people as they watch the film and timecode reactions to key scenes almost to the second. You can track their eyeballs as they watch to measure emotional connection. Netflix doesn't release audience data on its content, but you can bet that along with knowing you watched all of The Crown, they probably have a bit more intel about you that you may or may want them to know.
But asking someone "what did you think of the film?" and hearing the tone in their voice as they tell you how the lead character reminded them of their mother or hearing them laugh as they recount one of the scenes or analysing the different words and phrases they use to describe the film can tell you a lot about how the film connects with people, beyond eyeballs and algorithms. When judging how someone reacts to creative content like a film, it is hard to rely on big data alone. Human input, in all its complexities (and flaws), can sometimes be an advantage. Data or discussion? It doesn't have to be either/or. Why don’t we have both?
PS. In the days to come I will no doubt pore over the analytics for this blog to measure its cut-through and make some conclusions about how ‘successful' it was. The data will probably be very helpful. But I'm also keen to know what you think, so if you fancy a chat feel free to give me a yell.