I recently spoke at the MRMW Asia Pacific conference (2019) themed “Powering the future of insight automation”.
This is a transcript of my speech regarding my perspectives on data-driven insights.
Famous philosopher, Socrates said,
“True wisdom comes to each of us when we realize how little we understand about life, ourselves and the world around us.”
Through the human history, we have always been fascinated with the world around us. We questioned the who, what, why and how on life and business.
I do not profess to be an expert in the field of insights, but I am personally very curious with how people think and feel, and how these trigger what they say and do. Therefore, it is no surprise that my favorite subjects in school were consumer behavior and psychology.
Over the years, I had participated in different ways of mining insights and accumulated valuable hands-on experience. There are many things we can measure if we think hard enough, and many ways to do so.
My first big data venture
Things got really interesting for me when I joined the retail industry. I was instrumental in building the customer analytic platform with the grocery retailer. Our quest to understand our customers better drove us to look deeper into big data. We can pick up a lot of things by simply looking at what people are buying.
Many stakeholders in the organization were excited at prospects of using advanced analytics to understand purchase behavior. However, many of them do not know what to do with the data they are getting. Our analytic team received a lot of requests to churn out reports, with no specific business problems to solve.
In insights mining, most data analysts would advocate to first define the business problems that you are trying to solve. It should not be the other way around. Data insights are useless unless you know what to use it for and how to use it.
It takes creativity to define the problem, to steer the data mining. When done well, the insights can be used to support decision making.
I believe it is important to marry art with science where it comes to insights building.
The new human equation
With design thinking becoming more mainstream, many organizations are increasingly incorporating this technique of innovating into transformation journeys. It isn’t dissimilar to how creative industry generate ideas – just more scientific and methodical.
I love how it advocates starting with understanding latent needs, emotions and feelings of its potential customers. The principle of design thinking aligns with how I was trained in direct marketing, where I had to dissect what people think at a granular level.
The design thinking disciplines starts with uncovering inspirations, empathy study, followed by ideation and implementation.
I had personally utilized this in the ideation of digital transformation projects.
In empathy mapping, we observe our subjects (such as our consumers) in their environment – how they feel, think, say and do. We strive to understand what the consumers’ pains and gains are in getting their jobs done. These jobs can be functional, emotional or social.
In doing so, we can understand customers’ needs in real context.
Mapping customer journey, user experience and testing product usability
When I joined DENTSU, I introduced design thinking as a way to understand client’s business better. I believe that advertising agencies should start delivering creative business solutions, beyond branding and marketing campaigns.
Customer journey mapping
When we built customer journey maps for different industries and categories, we gained better understanding of unmet needs and underserved gaps. With this, we could start new conversations with clients and develop new solutions to solve their problems.
The methodology involved one-to-one conversations with people on different stages of customer journey.
We used the findings to map out
- the customer’s goals / jobs
- what they think
- what they feel (the highs and the lows)
- what they do
- what their pain points are
At honestbee, Delivery and shopping apps are used to enable the delivery and shopper contractors in the field. The apps helped them for order management, job task management, and delivery management.
The experience design team shadowed the shopper and delivery contractors in the field during their shifts. These direct interactions with the field workers helped us to acquire accurate contextual understanding of the usage context. We could improve work flows and user experience on the working apps with the insights gathered.
Likewise, for our consumer facing apps, we adopt usability test when we need to introduce features and experience. There is nothing like testing it with real people.
Real consumers are invited to try our prototype (app/web) products – each of them is given specific tasks and they are to think aloud as they experience the product, explain the rationale on their actions so that we can capture in-the-moment considerations.
It is common to track typical web/app usage data such as the time spent per session and numbers of sessions per transaction. They do not reveal the reasons for those usage behaviors.
Uncovering the motivations behind those actions calls for direct interactions to mine in-the-moment and contextual insights. These are not your conventional data.
Machine Learning X Human ingenuity
I learned a great deal from the data scientists at work on how we are applying machine learning.
A.I. for food recommendation
When we developed food recommendation engine on our app for food ordering, our data team fed patterns in past food ordering behaviors into the systems. Thereafter, the machine learning algorithm worked out complimentary dishes to go with an original dish selection.
The recommendation engine also learned by stringing together associated words. We had also used Wikipedia as a potential source for machine learning.
Typically, algorithm performance should improve as more data is being fed for machine learning. There is, however, a point of diminishing return where incremental data input does not guarantee increase in performance.
This is where we needed to inject human intervention and creativity, to dismantle and rebuild the algorithm. We engaged domain experts to get their inputs on cultural insights, country context, meal preferences, etc.
Google tags gone wild
An incidence in 2015 with google demonstrated the need for human intervention in machine learning.
We might have heard about the case of machine learning tag that an African face as gorilla. This is an example of why machine learning needs supervision to manage biases and interpret the data correctly.
Amazon A.I. hiring
In 2015, Amazon noticed a problem with its A.I. recruiting system. It was not rating candidates for software developer jobs and other technical posts in a gender-neutral way. This is because their computer models picked up patterns in resumes submitted over a 10-year period, which reflected a male dominance trend across the tech industry. As a result, it trained their computer models to vet applicants in a similar pattern.
Human intervention is needed from time to time to catch potential gender biases, and that applies to racial biases as well.
20th Century Fox experimented in data and creativity
I am fascinated by the potential of machine learning.
Beyond behavior observations, emotions and sentiments are captured into algorithm for performance and creative collaboration.
We are starting to see A.I. and robots rapidly automating Hollywood.
In 2016, 20th Century Fox worked with IBM Watson to design the trailer for the movie Morgan. IBM Watson utilized their experimental APIs and machine learning techniques to analyze hundreds of horror/thriller movie trailers. The research system helped them to understand what keeps audiences on the edge of their seats. The program helped to identify key moments which are used to compose the movie trailer.
It is amazing that A.I. can now recommend scripts to produce and what special effects to generate.
Despite of these, I personally believe that great storytelling still requires human creativity and imagination.
Indeed, machines excel in its capacity and speed to process data for insights.
As we evolve to automate our insight gathering, let’s not underestimate the importance of creativity, empathy and human interaction.
While A.I. is smart, I don’t think it could match or replace human creativity.
I like this quote on anthropology which reflects my position.
“Have the open-mindedness with which one must look and listen, record in astonishment and wonder that which one would not have been able to guess.”
We are in exciting time! Don’t you agree?
*This article was published on Linkedin on April 17, 2019.