Humpty Dumpty opens a new shop,
Humpty Dumpty market-researches all props.
But all the great STM and forecasting methodologies,
Could not save Humpty from losing the monies.
This rather lame limerick encapsulates the dilemma of the marketer trying to grow business through new products. As per Bussgang and Clemens (HBR, November 2018), the marketer wants:
- Faster time to market
- More experiments
- Fix errors faster
In contrast, what conventional MR forecasting offers:
- Months of project run time
- Demands for detailed spreadsheets about marketing plans by every month
- Stringent, inflexible testing protocols.
- High costs
Even after all this, the forecasts still come with an error range of +/-20%.
These challenges result from conventional research that universally follows the same approach:
- Expose concept one at a time – except in real life, people never make choices in isolation
- Ask intention to purchase or equivalent measure – except people do not always do what they claim to intend
- Individual responses are uninfluenced by what others say – except in real life people influence people
- Model, benchmark vs. norms – except most normative databases comprise largely of concepts/products which were never launched. Creation and maintenance of these databases is a slow and expensive exercise
- New to the world concepts without database benchmarks are difficult to forecast
There are variations to the above, of course, but at heart, they are just that – variations with fundamentals remaining unchanged.
We challenged ourselves to dismantle each of the above forecasting research tenets; what did we have to lose, anyway? At the worst, we would go back to what exists currently. But what if we succeeded?
Very deliberately, we created a design that eliminates the lab-like components of conventional approach and pushed it towards realism:
- Get the individual to ‘show, not tell’: let them practice their intent instead of stating it, in a life-like environment for closer to in-market behavior information:
- Real market simulation: mimic shopping experience – preview a new store stocked with new and existing products. Test option(s) is called out as ‘new’
- Competitive context: they make their choice vs. in-market options, by adding first to their wishlist and then the shopping cart. Each individual sees a different combination(s) of products, facilitated through an experimental design that turns each combination into a choice task. Every individual definitely sees the test concept(s); the rest of the concepts are rotated.
- Word of mouth: they see what others say before making their choice
- Feedback: they writes their product review, with star ratings as they would on any eCommerce site. They give feedback for the test concept and one other competition concept chosen (in order of priority) from her purchase cart, wishlist and the set they were exposed to
- To get an understanding of image perceptions, Intuitive Association (IA) based tradeoffs between the two chosen concepts – test and competition – conducted. In brief, the approach is a fusion of response latency (time span between a stimulus and response) and timed paired trade-offs.
- Modeling the choice: additions to purchase basket are converted into utilities and subsequently preference shares for test and existing options.
- Self-calibrated share: preference for existing offers is calibrated to match market share; the same calibration factor is applied to test concept preference for final projections – no norms needed.
Proof of Concept
The approach has already been executed for widely different categories – from personal care to AI devices to food & beverage to new age sensorial experiences.
Our first study was with a disruptive and new-to-the-market idea in a niche category. The results were available in less than two weeks at a fraction of the cost. The product since then has been launched in the market and has achieved 80%+ accuracy only based on a concept.
For a leading smart device company, we have used the approach to forecast volumes for multiple products, including the launch of their flagship brand in local Indian languages.
This Sounds Interesting, But Can We:
- Do this for products that mainly sell offline and not on eCommerce platforms? Yes, we can. The principles remain the same. Ecomm is merely representative of one retail environment.
- Provide people with product experience along with concept exposure? Yes, we can, though this is more relevant for FMCG products. The way we do this is as follows:
- Present a bouquet of products to the individual to test – either through in-home placement or on the spot, depending on the category and sample availability
- This bouquet would include the test option. It ensures that the individual makes their own choice of what they want to use and how without the overdue and artificial focus on the test option.
- Post-placement, the individual gives feedback on each of the products.
- The feedback is used to check dissonance vs. concept and accordingly calibrate the basis of the forecasts for the concept.
- The model, thus, is more a computational model for volumes and less of a conventional trial-repeat model. For the marketer also, tracking volumes is far easier than estimating on-market trial and repeat for launches.
- Build awareness and distribution assumptions in the forecast? Yes, we can. You just need to provide those to us.
The question for readers of this post is: can you put your money where your intention for behavioral testing is?