Achieving Advanced Sales Breakdown at McDonald’s | Tech in the World | McDonald’s Tech Blog | Nov 2022

Global Technology rethinks its internal business reporting to optimize decision making.

go through
Subramanian Krishnan, Senior Manager, Global Technical Data and Analysis Akshay Sahni, Director, Global Technical Data and Analysis
Brian Stronger, Data Engineer, Global Technical Data & Analytics

At McDonald’s, we strive to provide businesses with a complete snapshot of their sales drivers so they can make informed decisions to optimize spend/investment.

Over the past year, we’ve revisited the traditional process of generating executive reports to make them faster, automated, and accurate, and created a platform to change the way we interact with data. With advanced regression models, we can isolate the role of each sales driver and dynamically predict and plan future sales scenarios. This is our first attempt at centralized and standardized sales breakdown and forecasting. The product will continue to mature and evolve during our journey.

The current executive reporting process is individualistic, where sales actuals and forecasts are grouped into non-mutually exclusive categories (eg, channel, pricing, media).This does not account for interactions between For each of these sections, a complete picture needs to be provided.

In addition to providing a holistic view, we also wanted to automate data collection for reporting. Completing executive-level reporting for all countries can require analysts to spend a lot of time doing manual calculations, and we wanted to make sure we provided the tools to help make the Business Insights team more efficient and easier to operate.

Since McDonald’s operates globally with nearly 40,000 locations in more than 100 countries, some countries may require additional nuances. Due to local differences in menu items, promotional materials, and pricing, we found that countries used different methods to prepare local reports, so a dynamic and versatile tool was needed to account for these differences.

We built an always-on insights platform with self-healing model governance, accessible to analysts and business leaders alike through a lightweight web-based user interface.

The platform was built with the following technical principles in mind:

The platform architecture consists of three key layers:

  1. Data collection and transformation layer
  2. shape tiger
  3. User Interface (UI) Layer

under the hood
Data collection and transformation layer
Our multi-step engineering solutions combine 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 insights is timely and reliable input data. Our data collection and transformation layer automatically fetches periodic data from a variety of data sources and applies built-in data validation and data quality thresholds before data is transformed and aggregated for modeling.

shape tiger
We input internal and external data Least Squares Regression Model Map several menu categories to their most obvious business drivers. We employ various strategies to ensure that the model fits our variables well and accurately before finally exposing the results to the interface.

To fully understand the factors affecting sales, we take a multi-faceted approach and consider:

  • Incremental sales driven by marketing interventions, consumer accessibility, etc.
  • Simple factors like the expected sales we would make without taking any external factors into account, if the store were only open that day.
  • External factors that may exist during the collection of sales data, such as holidays, unemployment, COVID restrictions.

This information provides a more complete story and explains any fluctuations that may not be explained.

After ensuring the quantity and quality of the data, we use data from the past few years to build our model. We archive data that fails QA, and a custom notification system embedded in the tool alerts affected parties to potential data flaws that can cause delays.

After ingesting and transforming the internal and external data points, we are ready to feed them into the model. This advanced modeling algorithm gathers the final insights exposed to the tool. These variables are filtered through millions of models to determine how these seemingly independent factors actually affect each other, some more than others.

We create many combinations and keep only those that make sense and tell a meaningful story. We expose the most confident values ​​of the model through the final data presentation.

In our model, we perform some insightful transformations on the variables of feature engineering to reflect customer responses. Let’s take a look at a few of them:

Creative performance — the impact of marketing campaigns on sales or brand health over time — and capture how advertising builds and decays in consumer markets.

laugh – Latency in measuring the impact of a marketing campaign, reflecting the structural delay in consumer response to advertising.

diminishing returns — Capture non-linear consumer responses to marketing.The more consumers watch TV ads, the smaller the incremental effect

Our model is trained to determine the impact of each driver on sales without impact.This leads to a Anchor Model We iterate on it for future measurements. We use common model fitting methods to ensure that our model’s results on real data are consistent with the sample data we used in our model. We then use gradient descent to adjust the statistics, which essentially means that we are helping the model perform at an optimal level.

User Interface Tiger
After rechecking the quality of the data and testing against the correct model, the UI exposure is ready. Models are loaded and exposed through web-based applications built on the React framework.

The tool automatically refreshes periodically to get the latest results as new data becomes available. At its core it has three main functions: executive reporting, sales breakdown and scenario forecasting. With multiple views, users can easily access key performance indicators in the marketplace and determine the exclusive impact of each driver on sales.

The tool allows users to interact with data in ways that go beyond the capabilities available to most standard data and analysis tools today. Once the app is loaded, users can easily gain insights. They also have access to trends, comparisons, sales forecasting custom scenarios, one-click download to editable presentations, and more, enabling our analysts and business leaders to drive results.

Source link