Blogs

Consumer market analysis: Tips for stronger predictive modeling

Consumer Products Writer

https://kapost-files-prod.s3.amazonaws.com/uploads/direct/1455573470-30-1971/what-if-blog.jpg

One of the biggest challenges that leaders in the consumer packaged goods (CPG) industry face is lack of predictability. Even with in-depth planning and market testing, there's no guarantee that a product will succeed once it reaches store shelves. Brands need a way to outsmart these uncertainties, and big data fuels consumer market analysis capabilities that brands need to launch high-performing products.

"What-if" scenario analysis is a technique that's gaining popularity. The idea is simple: Rather than making general forecasts, brands can tailor their analyses to specific situations and contexts. As a result, CPG brands are able to modify their actions in a more focused manner.

The following are three tips to keep in mind when designing your brand's predictive analysis engine.

1. Ensure that your data-modeling foundation is airtight

There are several prerequisites that predictive analytics programs need. For effective customer market analysis, data must highlight the right market demand signals. Before building a predictive analytics layer, take the time to streamline your data collection process and automate it as much as possible.

Through automation, teams can free up bandwidth to introduce a predictive analytics layer. As Nari Visawanthan, the vice president of product management for integrated business planning at River Logic, points out, companies very rarely have "a single version of the truth" when conducting sales and operations planning. Data science, as a result, is often a matter of interpretation. Your analysts need to build the best possible models for capturing and acting on market demand signals if your brand is going to create the most accurate predictions.

Visawanthan recommends asking the following questions when developing your "what-if" scenario analysis:

  • Do you have mechanisms in place to model operational and financial data simultaneously?
  • Is it possible to perform optimizations across strategic, tactical and operational time horizons?
  • Is there a way to rank-order and prioritize certain demand signals above others?

Exploring these questions will ensure that your predictive modeling foundation is versatile, intelligent and flexible enough to capture the maximum number of explanations for trends that you're seeing in your market.

2. Focus on a specific challenge

The complexity of predictive modeling is often challenging for CPG leaders. There are many moving parts from the production stage through supply chain and point of sale, and the thought of modeling a product's entire life cycle may seem daunting, if not impossible.

Instead of tackling the brand's entire universe of possibilities and predictions, analytics should focus on a few important signals and outcomes. For instance, brands can use "what-if" models to calculate estimated yields of each product in terms of supply chain performance as it relates to sales. As Cleverism points out, demand-sensing tools have the potential to decrease inventory challenges in a range of industries, particularly CPG. Early demand-sensing tests at Unilever reflected a 25 percent decrease in forecast error and a 40 percent increase in demand forecast levels over seven days, according to Supply Chain 24/7.

This technique will streamline the demand signal forecasting process. After compartmentalizing the different aspects of your supply chain and marketing initiatives, you can use data to develop a story around these interconnected complexities, one pain point or open business question at a time.

3. Identify the right variables

Data science is like a blank canvas. When designing your predictive models, there are many factors to consider. How do you prioritize which variables will make their way into your models?

A recent survey from Nielsen Catalina, summarized by Warc, makes the argument that businesses should give careful thought to this aspect of model design. According to the report, engaging shoppers based on models that predict their propensity to buy can result in trial rates that are five times more effective.

The bottom line? In addition to modeling demographic variables, it's important to know what shoppers find important and why. This mind-set will improve the accuracy of your consumer market analysis by focusing on factors that deliver the highest impact. With these predictive modeling capabilities, organizations can more accurately interpret demand signals in the CPG market.

Hone real-time data to optimize predictive modeling. Explore IBM's Consumer Products industry solutions page.