The Future of Customer Insights

Multiple trends are flooding the customer insights industry. AI-powered solutions, more and more DIY tools and “Restec”, as well as the ask to tap on new data sources. I will take here a birds-eye view to distill the essence out of the noise. This can serve as your guiding rails for your 5-year insights roadmap.

The Future of Customer Insights

Trend #1 — More data sources

Insights were all about creating a questionnaire and running a survey in the past.

This time is gone.

The classical survey will always be an important source of information for the same reason that “talking with customers” will never be old-fashioned.

But technology now equips us to get information without asking and digitalization enables us to connect all those data sources:

  1. Behavioral in-survey: Text analytics today enables us to quantify open ends. Within a survey, we can also ask for audio feedback and read not only text, but emotional value from this. We can record the video, get facial expressions, and learn something about customers’ living context. We can learn from implicit, time-based feedback about underlying attitudes.
  2. Transactional: Data about a purchase is undoubtedly more valuable than expressed willingness to purchase in a survey. As digitalization evolved, nearly all customer actions are digital now, along with growing information around the customer. Linking this data with other sources opens a whole new goldmine.
  3. Social listening: Social media debates, customer discussion groups, ecomm ratings with comments — everything that consumers talking about in public spaces can be a valuable source of information, especially at our fingertips, for virtually little cost.
  4. Predictive enrichment: Today, AI systems are trained by eye-tracking data. Without any eye-tracking, they can predict the typical eye attention. In the same way, you can predict psychological characteristics based on some basic variables. You may be able to predict customer preferences based on the type of words they use. All this is based on a simple idea: Build a predictive model using a pilot dataset, then apply it at scale by relying on a few key predictor variables.

#2 Merge Qual with Quant

The market research and customer insights world has divided itself into Venusian and Marsian: Qual vs. Quant.

True is that businesses need both.

But both fields could advance by learning from each other. Quant can improve by becoming more explorative rather than just theory-testing. Qual can improve by introducing more rigor and validity control thru the aid of quantification.

AI makes it possible to quantify text, audio, and video even more reliable than humans at low costs. Tools evolve that mimic qualitative conversations with respondents and perform short qualitative interviews.

This quantified qualitative information can now be used in quantitative modeling. It can be used to discover but also to predict impact.

Both worlds will and must be merged.

The process of hypotheses discovery and validation will not be a binary qual vs. quant. We will instead see a continuum where “pure qual in-depth interviews” on the one extreme and theory-testing models at the other extreme will be the exception.

The vast amount of research will play between those extremes.

#3 Link Actions with Outcomes

What’s the worth of understanding that, say 50% of this speaker brands customers “love the sound”. Mmmh, ok isn’t it “a lot”?

What’s the worse of knowing that 23% of speaker brands are “music enthusiasts”. Mmmmh.

The point is that this is not insights, just aggregated data.

An insight is for instance to know that controlling a most reliable music stream to the speaker is not yet always achieved but when it is, the user is by .5 scale points more likely to recommend, which translates into 23 Mio. aggregated Customer Lifetime value.

What is customer insights? The quest to learn what drives customer behavior, more specifically, which actions influence intended commercial outcome.

With this, customer insights is most and foremost, the process to understand the hidden link between actions and outcomes.

The more we have that growing wealth of data at our fingertips, the better we are capable of fulfilling this understanding of “customer insights”.

Not only that. The more data we have, the more we don’t see the forest, but just trees. It’s the job of customer insights to not just aggregate, slice and dice this data. This is not just “boring” or confusing. It is not helpful.

Helpful is when we build causal machine learning models that help us to explore how actions and outcomes are linked.

#4 Create a Learning Loop

Last year we at Success Drivers tried to optimize subject lines for emails that go out to our prospective customers. We ran a questionnaire to assess 50 different ideas.

Then we tested the two best and two of the worst-performing subject lines in real life. Guess what happened!

One of the worst-performing was outperforming regarding open rates, while the top-rated subject line did not perform at all.

We further analyzed the underlying properties of all tested subject lines with an AI called neuroflash. It gives you the world of associations behind the words used.

A good subject line was “Conquer with Billy Beans AI”. We learned that the dominant language would trigger open rates. At scale, this learning would come from Causal AI.

Next, we ask another AI (GPT3) to write us related alternatives. We tried them.

This process produced the winning subject line “Straight to the point, FIRSTNAME” which showed open rates of 50% as opposed to 10%.

Long story short: Experiments are the gold standard of insights. In the future, winning businesses will weave in continuous experimentation into the daily workstream. With AI, as the example shows above, we can speed up the learning process. Learning that would have taken two years with AB testing in the past will take two months or even two weeks.

Subject lines are just the beginning. You can do this with pictures. You can do it with each personal service interaction, with each sales call your company is doing. Track, experiment, model, iterate.

This loop will open a whole new universe of insights.

#5 Insights needs its own marketing

Insights is like watching sports.

Everyone has an opinion. Everyone feels he can see what’s going on just by luring at data. Everyone feels he can be the coach.

This is even more true as more and more people within an organization are touching data. IT does it. Data Science does it. Everyone does it somehow.

Providing access to dashboards to a larger group in an organization can do more harm than good. This way, everyone draws the conclusion which fits his world.

The tendency is that “customer insights” is just running surveys.

It would be best if you staked out your terrain. Then it would help if you convinced others. This is called “marketing”.

Convincing others starts with empathy. It begins with understanding what key stakeholders genuinely care for. If it is “dollar”, give them the impact on profits.

Then educate on what you can deliver and why it requires marketing science to create useful customer insights.

A customer insights role needs an explicit internal marketing program. It is needed to bring insights into actions.

Building this marketing program starts with researching what is important to internal stakeholders. It continues with defining your brand. This is performed using a continuous content marketing and education program.

Name yourself to advocate for “how to turn data into insights” in your company.

If done right, you will be implicitely leading decision making.

# # #

What is your take? Which major trends did I miss?

Thanks so much,

Frank (frank@cx-ai.com)

--

--

Get the Medium app

A button that says 'Download on the App Store', and if clicked it will lead you to the iOS App store
A button that says 'Get it on, Google Play', and if clicked it will lead you to the Google Play store