How to master Marketing Mix Modeling in grocery retail?

Grocery retail is a highly competitive industry with a vast range of product categories and frequent discounts. The industry is also highly impacted by seasonality and external factors, such as weather and holidays, which can drastically affect sales. With so many variables to consider, grocery retailers must have a way to accurately measure the effectiveness of their marketing campaigns and optimize their budgets accordingly. This is where Marketing Mix Modeling (MMM) comes into play.

Why is it important to master MMM in grocery retail?

Mastering MMM is crucial for grocery retailers who want to stay ahead of the competition and maximize their profits. With so many different campaign types, store chains, product categories and marketing channels to consider, it can be difficult to know which campaigns are most effective and where to allocate your marketing budget. MMM provides a way to measure the impact of each marketing channel and optimize your spending to get the best possible return on investment.

Seasonality is another important factor in grocery retail, with sales often varying significantly throughout the year. Understanding these patterns and how they relate to your marketing campaigns can help you make better decisions and improve your overall performance. External factors such as weather and holidays also play a significant role in grocery sales, making it even more important to have a solid understanding of the impact of your marketing efforts.

Shortly, grocery retailers should be interested in mastering MMM because it can help them to:

• Accurately measure the effectiveness of marketing campaigns
• Optimize investments across store chains and campaign types
• Understand the impact of numerous product categories and discounts on sales
• Analyze external factors such as seasonality, weather, and holidays to improve decision-making
• Improve overall performance and stay ahead of the competition
• Maximize profits by making informed decisions about marketing spend

Collecting Data

MMM requires a lot of data and models to be effective. Grocery retailers need to gather data on sales, marketing campaigns, and external factors such as seasonality, weather, and holidays. This data can be collected from a variety of sources, including point-of-sale systems, marketing analytics platforms, and third-party data providers.

Once the data is collected, grocery retailers need to identify the key variables that are most important in driving sales. These variables can include marketing spend, product pricing, promotional activities, and more. The more accurate and comprehensive the data set, the more accurate and insightful the resulting MMM model will be.

Creating an accurate MMM model requires a sophisticated statistical modeling technique that can account for multiple variables and their interactions.

The modeled data is split to different combinations of product groups, customer segments and store chains:

For example:

• Daily sales of Fish product group to Platinum customers in Hypermarket chain
• Daily sales of Meats product group to Platinum customers in Hypermarket chain
• Daily sales of … product group to Platinum customers in Hypermarket chain
• …
• Daily sales of Fish product group to Gold customers in Hypermarket chain
• Daily sales of Meats product group to Gold customers in Hypermarket chain
• …
• Daily sales of Fish product group to Platinum customers in Supermarket chain
• Daily sales of Meats product group to Platinum customers in Supermarket chain
• …

E.g. 25 Product groups * 6 Customer segments * 3 Store chains = 450 modeled time series

3 years of data would have 450 * 3 * 365 = 490,000 modeled data points

In addition to data and modeling, accurate MMM also requires ongoing monitoring and refinement. As the grocery retail landscape evolves, so too must the MMM model. Regularly reviewing and updating the model can help ensure that it remains accurate and effective.

Here are the main benefits of using accurate data in your MMM model:

1. Better decision-making: Accurate MMM data provides valuable insights into the effectiveness of marketing campaigns, enabling grocery retailers to make more informed decisions about how to allocate their marketing budgets for maximum impact.
2. Improved ROI: By optimizing marketing spend across different store concepts, campaign types, channels and product categories, grocery retailers can improve their return on investment and maximize profits.
3. Deeper understanding of customer behavior: MMM data can provide insights into customer behavior and preferences, enabling grocery retailers to tailor their marketing efforts to better meet customer needs.
4. Ability to respond to market changes: Regularly monitoring and refining the MMM model can help grocery retailers stay ahead of market changes and adapt their marketing strategies accordingly.
5. Competitive advantage: Accurate MMM data gives grocery retailers a competitive advantage by enabling them to better understand their customers, optimize their marketing spend, and drive success in the highly competitive grocery retail industry.

Using product category level modeling

The only way grocery retailers can master MMM is by using product category level modeling.

Product category level modeling involves breaking down the MMM by individual product categories to gain a deeper understanding of the factors that drive sales within each category.

Even when there is no data available on which promotions have featured in which media, we still recommend product group level modeling because:

1. Granted discounts

Granted discounts are discounts that are offered by grocery retailers to their customers on certain products or product groups. These discounts are typically advertised through promotional materials such as flyers, emails, or in-store signage. The retailer sets the discount rate, which is the percentage by which the regular price is reduced.

The purpose of offering granted discounts is to increase sales of certain products or product groups, which can help the retailer to meet sales targets, clear out inventory, or increase customer loyalty. Granted discounts can be offered on a regular basis, such as weekly or monthly promotions, or on a seasonal basis, such as holiday sales.

When discounts and promotions are accurately allocated to each product group, they can be just as effective in driving sales as media. In our experience, these tactics have a significant impact on sales. While it may not be possible to precisely allocate media spend to individual product groups, accurately allocating discounts and promotions can provide a signal that is just as strong. This can greatly benefit the MMM model by providing valuable insights into the performance of different product groups and enabling grocery retailers to optimize their marketing strategies accordingly.

Furthermore, discounts can be allocated to any dimension that exists in the sales data. If there are significant differences in granted discounts across different sales channels or geographic areas, we recommend splitting the modeled data into these dimensions as well. However, the main value of accurate MMM data is gained through product group-based modeling.

Our aim is to have a situation where promotion and media investment levels (i.e., features) differ over time and are not fully correlated. When we apply models on a product group level, we observe significant variance over time in granted discounts and media features. This is particularly helpful in finding the right level of promotion price elasticity per category.

2. Promotion price elasticities

Promotion price elasticity is a measure of how sensitive consumer demand is to changes in price during a promotion. It is a key factor in understanding the impact of discounts and promotions on sales.

A positive price elasticity indicates that as the price decreases, demand for the product increases. For example, if a 20% discount leads to a 10% increase in sales, the price elasticity is positive. A high positive price elasticity means that consumers are very responsive to price changes, and a small decrease in price can lead to a large increase in demand.

On the other hand, a negative price elasticity means that as the price decreases, demand for the product decreases as well. This can happen if consumers perceive a low price as an indication of low quality, or if the product is seen as a luxury item that loses its appeal when its price is reduced.

Assuming that granted discounts have been allocated by product group, we have observed that the promotion price elasticity of demand can vary significantly across different categories. For example, a 20% discount might cause sales of Salmon to increase by +200%, whereas the same discount could result in only a +10% increase in sales of Cucumbers. If granted discounts are not analyzed by product group, this important explanatory factor can be missed in the model.

3. Seasonalities

Seasonality refers to the recurring patterns in demand for products or services that are influenced by changes in the time of year or the weather. In grocery retail, different product groups may exhibit different seasonal patterns based on factors such as holidays, weather conditions, and cultural events.

For example, fresh produce such as fruits and vegetables may exhibit strong seasonal patterns based on growing seasons and availability. Certain fruits and vegetables may be more popular during the summer months, while others may be more popular during the winter. Similarly, seasonal holidays such as Easter or Christmas can drive increased demand for certain products such as lamb, ham, or holiday-themed treats.

By deconstructing the data and analyzing seasonalities by product group, we can better identify and understand these patterns. This leads to improved accuracy in our baseline sales estimates, which in turn helps to optimize our forecasting and decision-making.

4. External Variables

External variables refer to factors outside the company's control that can impact sales, such as weather, economic conditions, or public health crises like COVID-19.

Temperature is a good example of an external variable that can impact grocery retail sales. For instance, during hot weather, sales of cold beverages, ice cream, and other summer-related products tend to increase. Conversely, during cold weather, sales of hot beverages, soups, and other winter-related products tend to increase.

Another example is the COVID-19 pandemic, which has had a significant impact on the grocery retail industry. The pandemic has changed consumer behavior, with many people preferring to stay at home and cook more meals. It has also led to supply chain disruptions and increased costs for grocery retailers.

In MMM, it is important to incorporate external variables like temperature and COVID-19 in the analysis to accurately predict and optimize sales. By understanding how these variables impact sales and how they interact with other variables like promotions and media, grocery retailers can make informed decisions about their marketing strategies and maximize their profits.

Curious to learn more? Book a demo.