With the disruption brought about by COVID-19, digital transformation is occurring at an accelerated pace. CPGs need to operate efficiently and intelligently, and they’re turning to advanced technology like artificial intelligence and machine learning to do so.
Even prior to the current state, CPG leaders have been leveraging the power of AI and machine learning to automate analyses, saving time and resources to improve their bottom lines. These existing use cases can act as models for CPGs that now have an urgent need to upgrade their technology.
This blog post will cover how CPGs use machine learning for:
Machine learning and AI are more than temporary trends; they’re fundamentally changing how CPGs understand their consumers.
Practically, machine learning algorithms can identify and tailor critical insights across different data sets, scenarios, and functions— meaning, the algorithm “understands” what information is most important and relevant for a category analysis whether the data represents cosmetics or food and bev.
To learn more about machine learning, check out this resource: Machine Learning in Business Intelligence Solves the Puzzle. Otherwise, let’s dive into the myriad of ways machine learning is used in data analysis.
1. CPGs use machine learning for brand health analysis.
Understanding brand health is critical for CPGs.
Just as data has become more complex in recent years, brand health too has evolved beyond the consumer experience to encompass a brand’s overall performance:
As revenue models and customer expectations continue to evolve rapidly, every aspect of a business can affect the brand—from logistics and inventory management to the in-store experience. As a result, organizations increasingly are considering the connection between their brands and their underlying business operations with a focus on how performance can affect brand health.
Source: https://deloitte.wsj.com/cmo/2018/01/11/assessing-brand-health-risk/
This evolution means there are an incredible amount of factors to consider when it comes to brand health. Yet, at the core of brand health analysis is the simple question: “how did my brand do?”
In order to answer this question, brand managers need to understand which factors are driving a brand’s performance.
Which metrics, if changed, would create the largest ripple effect? Is location critical to brand health, or does advertising channel selection matter more? Where should employees focus their time and energy for the best results?
These answers aren’t always intuitive. It’s not uncommon for CPGs to leave growth opportunities on the table because they’re acting on partial information or assumptions.
For example, a brand’s strong sales performance can mask losses like declining market share. Furthermore, it could be reasonable to attribute this strong performance to a sales value increase in the Southeast and subsequently double down on promotions in that region. Meanwhile, data indicates that an increase in brand penetration in the Northeast would offer the greatest growth opportunity, but this insight is obscured by numbers on the surface.
Some of the most intuitive and obvious assumptions can be the most dangerous, simply because it’s so challenging to look past them. Machine learning approaches brand health without preconceived notions or this kind of bias.
CPGs are using machine learning algorithms to parse data and identify brand health drivers.
Specifically, machine learning can:
- Analyze data intelligently and exhaustively, using brute force models to evaluate all possible data permutations and focusing attention on factors that matter the most.
- Perform contribution analysis, weighing and understanding how much each factor contributes to overall brand performance.
- Determine which data is relevant to brand health, without a biased approach.
2. CPGs use machine learning for market share research.
One of the most important metrics that CPGs must consider is market share, whether across a specific brand, segment, category, or industry.
For CPGs, market share insights indicate how brands perform against competitors. This context is critical to building a long-term growth strategy. The goal with comprehensive market share research is to leave no stone unturned and to fully understand where a brand fits into the larger landscape.
Machine learning is capable of performing this in-depth and complex research in seconds.
To do so, machine learning algorithms thoroughly cull through syndicated and first-party data sources, test every data combination to determine relationships, and generate insights.
A complete machine learning-generated market share report would include the following information:
- Market share performance — Year-over-year, has market share increased or decreased and by how many basis points?
- Market share momentum — How does recent movement in market share compare to the long-term trend?
- KPI drivers of market share growth — which key performance indicators (KPIs) are driving market share growth (or decline) the most, and to what degree?
- SWOT analysis— What are the brand’s Strengths, Weaknesses, Opportunities, and Threats based on its performance and that of its competitors?
With this information, CPG professionals can zero in on the factors that drive market share and make decisions based on the big picture.
What were the top gainers and decliners? Which contributors most determined market share performance? The answers to these questions can help teams take action with the greatest possible ROI.
What’s particularly compelling about machine learning in this instance is that its speed and efficiency enable immediate follow-up research.
When marketers learn that distribution contributed 250 basis points to the positive trend, they can quickly drill down into distribution and find out why. Essentially, machine learning enables more targeted, informed decision-making.
To learn more about market share research and brand health analysis for CPGs, check out this quick-read blog post. Or, watch this quick video demonstration to see this analysis in action.
3. CPGs are automating category analysis.
Syndicated data sources are an important investment for CPGs wanting to understand their performance, especially within the context of their competitors. To take advantage of these massive data sources, CPGs are using machine learning algorithms to analyze category data quickly and effectively.
Machine learning has advanced to the point that it can automate the bulk of what is largely a manual process. That means CPG professionals can start with a question (like “how did my category do?”) and get an answer in seconds without having to wrangle disparate data sources, formulate and test assumptions, pull data, and repeat ad nauseum.
Within the manual process, analysts spend so much time pulling data together that they tend to follow the same pathways when it comes to analysis. As a result, analysis can easily become biased; analysts don’t have time to look below the surface, so they stick with what’s worked in the past. Unfortunately, COVID-19 data is beyond historical precedent, and the old methods of analysis are quickly becoming obsolete (if not now, then in the next 5 years).
In contrast, machine learning is unbiased, approaching the data without a set pathway or preconceived notions.
Once the machine performs analysis, it generates insights and data visualizations that explain the drivers in a category. This analysis should, for example, identify how different brands contribute to the overall category and draw the user’s attention to the most relevant and important insights.
While data visualizations are familiar mainstays at many CPGs, insights, on the other hand, vary widely, and machine learning has significantly elevated the baseline expectation.
The value of insights is no longer limited to simply pointing out what’s happening in the data, i.e. “market share increased 50 basis points.” Now, insights can be structured in complete data narratives that explain analysis in natural language.
“You have that ‘a-ha’ moment; that’s really what an insight is, where you see things differently than what you saw before. And you can follow that train of thought without these interruptions,” states Laura Braunecker, Founder and Principal Consultant of Zeerio, in this Food Dive Playbook. “[CPGs] want their best and smartest talent to be more productive, to get the answers they need and to be proactive, not reactive.”
That’s the key with automated analysis — generating insights that lead to action. And generating insights quickly enough that action can be taken proactively, not reactively.
This automation means CPG professionals like brand managers, CMOs, finance teams, and sales teams can perform analysis themselves, quickly and without intervention from an analyst.
Of course, the potential applications of machine learning are nearly endless, and they can fulfill analytical requirements for a myriad of departments and functions.
A few additional examples of machine learning in action include:
- Scenario analysis — ask “what if” questions to see what happens when metrics are altered.
- Granular forecasting — predict revenue and profit performance.
- Opportunity analysis — identify and prioritize growth opportunities.
Each of the above examples are active use cases for machine learning.
Hopefully, these examples inspire ideas for how you can use machine learning to solve problems and drive growth at your organization.
Additional Resources
- CPG Analytics Guide: Learn everything you need to know about CPG analytics with this ultimate guide.
- Data Analytics and Machine Learning: Learn more about machine learning in this more technical overview of AI in analytics.
