Product Usage Insights with AI

[Big Ideas Series] Leveraging AI's ability to gain insights to test products and understand consumer behaviour with an objective, data-backed approach.

Editor’s Note: This post is part of our Big Ideas series, a column highlighting the innovative thinking and thought leadership at IIeX events around the world. Kushank Poddar will be speaking at IIeX North America (June 11-13 in Atlanta). If you liked this article, you’ll LOVE IIeX North America. Click here to learn more.

How do you test new products? The most direct way is to “ASK”. If the product is at a concept stage, you ASK people for their feedback.  If you want to test prototypes, you ASK people to use them and give feedback. The process of asking is direct and simple, but it comes with some well-known challenges in tow.

‘Asking’ only scratches the surface

When being asked, respondents often get influenced by the questions, they may not be good at articulating their thought process or more importantly, they may not even know what is really driving their opinion. Yes, there is solid science behind this last point. Studies conducted by Harvard Business School professor Gerald Zaltman says that 95% of our purchase decision making takes place in the subconscious mind. It’s safe to say that our subconscious feelings play a big role in why we like a certain product. Asking people for feedback can only scratch the surface and hence ‘implicit’ research methods are needed to get a deeper understanding.

Human Observation is subjective and doesn’t scale

A useful implicit research approach for testing products is Human Observation. Typically, skilled ethnographers and qualitative researchers keenly observe how people use products and provide big-picture answers to what makes a consumer like or dislike a product. This approach delves into an understanding of human psychology and can be very useful. However, given the manual approach – this exercise is time-consuming, subjective and not scalable beyond small research participant sample sizes (usually ~20). It is intrusive in nature and may bias the usage behaviour of the respondent. Also, this approach requires senior experienced researchers to “shadow” respondents, making the entire exercise prohibitively expensive. We propose a new research approach that helps researchers dig deeper while addressing the above concerns.

The Machine Observation Approach

The recent advancement in AI has now given machines the ability to see, and this ability is getting stronger every day. By leveraging this new technology, we propose a product testing approach where observation is automatically done by machines.

This approach uses AI to analyse videos of people using different products. The AI breaks down the product-using journey into small actions or components, and it records second-by-second time series data for these actions. In the example video of the beard trimmer below, you can see that the AI system is accurately recording metrics related to time spent in trimming different sections of the research participant’s face.

The AI analysis can be extended to record accurate data following key metrics related to beard trimming:

  • Total time spent in trimming.
  • The number of times excess hair was cleaned by each participant from the trimmer.
  • The number of trimming strokes and stroke types.
  • Facial emotions and alignment during the entire process.
  • Sensitive areas where the participant came closer to the mirror for trimming.
  • The angle at which trimmer is held, or point of contact (middle, corners).
  • The density of beard and beard style before and after using the product.
  • Trimming length/precision and power settings used (IoT).
  • and much more…

This AI-based analysis of product usage videos creates a treasure of data that can be mined to understand user behaviour and explain why they like or dislike certain products. With some training, AI algorithms can also be used to implicitly predict user feedback without explicitly asking. This is an objective, data-backed and quantitative approach for product testing that can be easily scaled across a statistically significant quant sample size (i.e. ~80-250 participants). And this can be used for a wide variety of use cases – like testing beauty products, ice creams, lawn mowers, automobiles and more.

The right answer is Man + Machine

So we conclude by again asking – How should you test new products? By asking consumers, by human observation or by machine observation? At Karna AI, we believe there is no right answer to this question. These three research approaches are complementary and address different sections of the pie. We don’t believe that machine observation will replace other approaches. But what we do believe is that adding machine observation to your research mix can unearth hidden insights that cannot be derived from other approaches. In today’s hyper-competitive markets, this extra edge could mean the difference between winning and losing.

Have some questions or interested in knowing more? Come and hear my talk: See the Unseen – Observing How Consumers Behave Using AI on Monday, June 12 at 4:40 PM at IIeX North America.

Please share...

Join the conversation