“Thick” or qualitative data (1) helps organizations and their AI-solutions make better decisions by including insights about the context, nuances and influences that leads to customer’s behaviors and decisions.
But “thick” data sounds expensive, difficult and complicated to collect and use.
It isn’t.
In fact collecting “thick” data means doing almost the same as we are already doing. Only asking different questions.
First: What is “thick” data
According to Gilbert Ryle (2) “thick” compared to “thin” data is:
- “thin” : includes surface-level observations of behavior
- “thick”: adds context to such behavior
Put even simpler: “thick” data is what helps us answer “why” people are doing what they are doing while “thin” data only documents what has happened.
In an oversimplification (below) we could say “thick” data is the yellow data on the left helping us understand what can influence people, while channel performance and sales data are “thin” data on the right documenting what people are doing.
Why should we care about “thick” data
AI-engines use both thin and thick data. The former is used to look for patterns and probabilities, while the latter helps the AI better understand context, nuance, culture, rituals etc.
According to a search on Perplexity (3): “thick data significantly enhances AI’s ability to understand and respond to complex human behaviors and motivations, leading to more nuanced, personalized, and effective performance across various applications.”
Ethnographer Tricia Wang (4) argues that over-reliance on “thin” data was part of Nokia’s downfall. Their own “thick” data was clearly suggesting that even low-income individuals would invest in smartphones, while Nokia’s quantitative data was telling it (what it wanted to hear), that the group would keep buying dumb phones.
How do we get “thick” data
Thick data is readily available almost everywhere, especially online. We only need to ask for it.
Which is kind of the problem: we don’t. For no apparent reason?
But the value is there, for anyone who wants it. Right in front of us.
These are just a few examples of questions we’ve answered by capturing “thicker” data about the customer:
- What job [need] are customers trying to solve which leads them to us?
- How do customers identify themselves?
- Are customers looking for products or insights? (Where are they in their job-map / journey)?
- What is the perspective / lens /context through which customers understand your offer?
- What else influences customers decision making?
- What language (terminology) do customers use?
- How do customers measure success?
- Are we helping customers achieve this success?
This is how you get “thick” data online (one generalized approach):
- Collect assumptions by asking your team “what has to be true?” in relation to what you are trying to achieve (can be anything from strategy to creative concepts or execution).
- Choose the assumptions that you want to test and articulate them into a clear experimentation statement (if x happens then y) identifying where and how you will run the experiment including how to measure it and what success looks like (5).
- Create the experiment, run it, analyze and adjust.
Now scale this from a novel to a system wide approach
We talk about “thin” or quantitative data being at scale, but are we resigned to qualitative data having to be expensive, slow, complicated and not to scale? If so this is only because we’ve decided to think it like this.
We can ask any question and design any interaction to capture its answers. So why not scale up the “thick” data production? Breaking from the boring data everyone else has (which we collect and make sense of in the same way and with the same algorithms … what’s the fun in that?(6))
Asking “thick” questions should be as natural as asking “thin” questions (e.g. open-rates and conversion funnels). Why not?
We need to establish new design or experience principles that put learning about our customers (not our channels) at the forefront, make available compliant “learning modules” and objects as part of our design and experience systems. And we need to create demand for them by having our teams ask for better questions, not better answers.
This would produce a constant flow of “thick” data from across the ecosystem feeding into the AI models in real-time and at scale.
Competitive advantage
“Thin”-questions are thinking inside a box inside a box. “Why”-questions are thicker, they help us understand the customer when they are with us and when they are not.
“Thin” questions are conventional and conformative. They make us the same as everyone else. Everyone collects the same data, asking the same questions using the same algorithms .
“Thick” questions are competitive advantage and the best question is “why not”?
Sources:
(1). Pratibha Kumari J., What is Thick Data, https://www.linkedin.com/pulse/what-thick-data-pratibha-kumari-jha/
(2). Wikipedia, Thick description, https://en.wikipedia.org/wiki/Thick_description
(3). Search related to “thick” on Perplexity, https://www.perplexity.ai/search/in-general-terms-whoch-data-do-4myjkNOXSPWBSNYfyv8iyw#1
(4). Tricia Wang, The human insights missing from big data, https://www.youtube.com/watch?v=pk35J2u8KqY
(5). Helge Tennø, Business experimentation, https://medium.com/everything-new-is-dangerous/business-experimentation-f5620919f209
(6). Helge Tennø, Customer as Competitive Advantage, https://uxdesign.cc/customer-as-competitive-advantage-19a6ede62852
How to close the customer-data-gap was originally published in UX Collective on Medium, where people are continuing the conversation by highlighting and responding to this story.