Global Technology is rethinking its internal reporting to optimize decision-making.
Subramanian Krishnan, Senior Director, Global Technology Data and Analytics Akshay Sahni, Director, Global Technology Data and Analytics
Brian Stroner, Data Engineer, Global Technology Data & Analytics
At McDonald’s, we strive to provide the business with a comprehensive overview of sales drivers, so they can make informed decisions to optimize spend/investment.
Over the past year, we’ve overhauled our old process for generating executive reports to make them faster, automated, and accurate, and created a platform to transform the way we interact with our data. Using advanced regression models, we can isolate the role of each individual sales driver, as well as dynamically predict and plan future sales scenarios. This is our first foray into centralized, standardized sales breakdown and forecasting. The product will continue to mature and evolve throughout our journey.
The current executive reporting process is individualistic, where actual sales and projections are grouped into categories that are not mutually exclusive (eg, channel, pricing, media). It does not take into account the interactions Between each of these parts, which is necessary to give a complete picture.
In addition to providing a holistic view, we also wanted to automate data collection that enables reporting. Completing executive-level reports for all countries can take analysts hours of manual calculations, and we wanted to make sure the tools we provide help make the Business Insights team’s tasks more efficient and actionable. .
Since McDonald’s operates globally with nearly 40,000 locations in over 100 countries, some countries may require additional nuance. Due to local variations in menu items, promotional materials and prices, we found that countries prepared their local reports using different methodologies, hence the need for a dynamic and versatile tool to take into account these differences.
We’ve built an always-on insights platform with self-healing model governance accessible to analysts and business leaders, through a lightweight, web-based user interface.
The platform was designed with the following technology principles in mind:
The platform architecture includes three key levels:
- Level of data collection and processing
- Modeling level
- User interface (UI) level
under the hood
Level of data collection and processing
Our multi-step engineering solution combines the power of data engineering, machine learning, and user experience to transform the way our business understands overall sales performance. The starting point for effective and accurate information is timely and reliable input data. Our data collection and transformation layer is automated to obtain periodic data from various data sources and incorporates data validations and data quality thresholds applied before the data is transformed and aggregated for modeling.
We feed internal and external data into a Least squares regression model to map multiple menu categories to their most apparent business drivers. We employ a variety of tactics to ensure that the model is a good fit for our variables before finally exposing the results to the interface.
For a holistic view of what influences a sale, we take a multidimensional approach and consider:
- Incremental sales driven by marketing interventions, consumer accessibility, etc.
- Simple factors, such as the expected sales we would make regardless of any outside factors, if the store was just open for the day.
- External factors, such as holidays, unemployment rates, COVID restrictions that may have been present at the time the sales data was collected.
This information provides a more complete story and accounts for any fluctuations that might otherwise remain unexplained.
After ensuring the quantity and quality of data, we use data from past years to build our model. We archive data that fails quality checks, and a custom notification system built into the tool alerts relevant parties to a potential data breach that can cause delays.
After ingesting and transforming the internal and external data points, we are ready to integrate them into the model. This high-level modeling algorithm gathers the final information that is exposed to the tool. Variables are filtered through millions of models to determine how these seemingly unrelated factors actually influence each other, with some having more influence than others.
We create many combinations and only choose the ones that make sense and tell a meaningful story. We expose the values on which the model is safest through the final data displays.
In our model, we perform several insightful transformations on our variables for feature engineering to reflect customer response. Let’s review a few of them:
Adstock effect – the effect of marketing activities over time on sales or brand health – and captures how advertising builds and decays in consumer markets.
Gap — measures the lag in the impact of a marketing activity, reflecting a structural lag in consumer response to advertising.
Diminishing returns — Captures consumers’ non-linear response to marketing. The more a consumer sees a TV ad, the less incremental effect it has
Our models are trained to determine each driver’s effect on sales without influence. This results in a Anchor model from which we iterate for future measurements. We use common model fitting methods to ensure that our model results on real data are consistent with the sample data we used in our model. Then we adjust the statistical results with gradient descent, which basically means that we help the model to perform at the most optimal level.
User interface level
Once the data is rechecked for quality and tested for the correct model, it is ready for UI exposure. Models are loaded and exposed through a web application based on the React Framework.
As new data becomes available, the tool periodically updates automatically for the latest results. At its core, it has three main features: executive reporting, sales breakdown, and scenario forecasting. With multiple views, the user can easily access a market’s key performance indicators and determine each driver’s unique impact on sales.
The tool allows users to interact with data in a way that surpasses the capabilities available on most standard data and analysis tools today. The information is readily available to the user when loading the application. They can also use features such as trends, comparisons, custom sales forecasting scenarios, one-click upload to editable presentations, etc., enabling our analysts and business leaders to drive results.