Market research has changed dramatically and company CEOs today are extremely fortunate because that change has made market research much more accurate, predictive and affordable than it was in the past. This brand of market research is not done using focus groups or telephone surveys. Those methods are not only expensive, but it is next to impossible to get the diverse sample that is required to be able to use them to predict future trends. Today’s market researchers have ditched the phone and the focus group rooms for social media sites and replaced the facilitator with sophisticated algorithms that can ensure sampling is done accurately. Using these techniques, you can find out far in advance when your products and brand are about to soar or take a beating.
You are probably wondering, as many do “but are social media sites really representative of what people are thinking? Not everyone is on social media, isn’t it all young people?” The answer is that there are now enough people on social media that we can get a representative sample of the population. You know how telephone pollsters always ask about your household income? That’s because they want to make sure that not everyone in their survey makes between $60,000 to 90,000 a year. They need to survey low, middle and high income people according to how they are represented in the population. Ideally they are also getting a good mix of ethnicities, religions, education levels and other important demographics. The problem, of course, is that when asked how much money we make, most of us lie. In fact, a study by MIT Professor Alex Pentland shows that people lie on surveys about 50 percent of the time. How do you feel about making important strategic decisions with data that is 50 percent false?
With social media research, we are not asking any questions. We use sampling and topic modeling to accurately “survey” what people are saying in their natural environment. Topic modeling is not the same as keyword searches. Keyword searches are notoriously unreliable. When is Bell a telephone company brand and when is it something you ring? The word “sick” has a negative connotation for a medical doctor and a positive one for a teenager. Topic Modeling corrects for this by putting words in the context of the other text and also the demographic of the person speaking. And using proper sampling algorithms, we know the income level, education, race, religion and family make-up of a person without having to ask them.
I want to be clear, this is not social media monitoring. Social media monitoring doesn’t use sampling, is not scientific and cannot be used to tell you how the population as a whole (i.e. those not on social media) is feeling. You have to know how to read the data properly. Failure to do so can lead to costly mistakes.
Technologies that simply give you daily "sentiment readings" are, at best, unhelpful and at worst, very misleading. Known as “noise” in the data science world, I know many entrepreneurs who have their teams reading various sentiment scores and including them in their KPIs without realizing that they mean absolutely nothing, and should definitely not be actioned, even if you could divine what action to take. These readings simply do not give you actionable insights, or clues as to what you can do right now to benefit your company. And more importantly, they don’t tell you when the change has become statistically significant -- the point at which action needs to be taken. To do that, you need what is called changepoint analysis. In statistical terms, it’s the point at which your distribution has changed. In business terms, it is the point at which you and your team need to implement change or face the consequences. Fortunately, we see this measure before it actually affects your bottom line. So you have time to change. Change before that and you might be messing with a good thing.
We see these statistically significant changes happening, on average, about 12 to 18 months before sales are significantly affected. Any CEO not monitoring these trends on a monthly basis is at a severe disadvantage to his competition. And speaking of the competition, most brands now monitor not only their own companies but those of their competitors as well, to see where their vulnerabilities lie.
Market research studies that used to cost on the order of $25,000 to 50,000 can now be done for $2,500 and are much more accurate. Marketing managers are always being asked to show ROI on their campaigns. Now you can. Run the campaign and then spend the $2,500 to see how far you moved the needle. Running it monthly (which is now affordable) shows how your brand perception changes over time and with each campaign. You need to be able to pivot quickly the moment the market begins to shift, not wait until you see it at the cash register.
We had one client who came to us after a “sudden” dip in sales caused concern. The first thing we did was analyze whether this change was detectable through social media analysis and at what point. Sure enough, a full 24 months before they saw changes in sales, a changepoint analysis showed their main market was growing discontented. What propped them up was a secondary market that was riding on the coattails of the first cohort. But as the disenchantment grew, the secondary market fled too, and sales dipped precipitously and suddenly. They had no inkling of this two years prior, when our data showed it had started.
This is true across all industry sectors, big and small. Your customers are talking about you, and the chatter grows exponentially. You need to be vigilante and know when you have a problem. That is your job as a marketing manager. There is no longer any excuse for saying “I didn’t know.” It is the marketer’s job to see the trends before they show up on the CFO’s spreadsheet. And now you have the tools you need to keep your numbers soaring.