Work-from-home and online shopping have fundamentally changed consumers' interactions with brick-and-mortar locations. Marketers, store network managers, and urban planners must better understand mobility patterns and predict online and in-store shopping behaviours. Where do people go to shop? Where do they go for services? How do online and in-store touchpoints interact? Today’s analytic tools seamlessly integrate more data, are more accessible, and are easier to use than ever before. They provide decision-makers with precise analytic capabilities and access to detailed store visits and purchase information in a privacy-compliant environment.
Enhancing the Tried and True Huff Gravity Model: A Modern Approach
The standard model for spatial analysis and visitation prediction is the Huff Gravity Model, originally proposed in 1963. Its primary function was to predict the probability of someone visiting a particular location—whether it be a store, national park, or any other physical place—based on factors such as distance and attractiveness. While groundbreaking at the time, the original Huff Model was limited by data availability and computing capacity. It didn’t account for the nuances of consumer behaviour, such as varying interpretations of attractiveness, ease of travel, e-commerce, or changing conditions like time and seasonality.
Arima’s data scientists have reimagined the Huff Model, incorporating these factors to provide more accurate predictions.
Arima’s Take: A More Dynamic Huff Model
The dynamic nature of Arima’s model allows for flexible predictions, enabling the tracking of how visitation and traffic fluctuate over weeks, months, or seasons, as well as in response to changing competitive conditions, such as the opening of a new store. For example, a beach resort may be far more attractive in the summer than in the winter, or store visitation may vary depending on the strength of competitors’ e-commerce platforms. The Dynamic Model captures these variables in ways the original Huff Model could not.
Arima’s model includes built-in features that describe site characteristics such as:
- Size
- Pricing
- Visitor demographics
- Type of amenities (e.g., restrooms)
- Time since last renovation
- Number and size of physical locations
It also captures visitor characteristics, such as:
- Does the location align with the consumer’s psychographics?
- Do they have specific product and brand preferences?
- How do they prefer to shop? How do they like to buy?
These factors are crucial for understanding what makes a location attractive to different individuals.
In the Dynamic Model, attractiveness is no longer a single number but a complex function of multiple factors, each weighted according to its importance in the decision-making process.
How Synthetic Data Enhances the Huff Model
Arima’s take on the Huff Model relies on synthetic data rather than first-party real data. Synthetic data is advantageous because:
- Enhanced Coverage: It fills gaps in real-world data, providing a more complete picture of consumer behaviours and market conditions.
- Privacy: It avoids issues related to personal data privacy.
- Flexibility: It allows for simulation and testing of various scenarios, which improves the accuracy of predictions.
By using synthetic data, Arima can offer more comprehensive and actionable insights than models that rely solely on first-party data.
Real-World Application: An Example
Let’s consider Jane, a hypothetical customer living near High Park.
She has four grocery buying options:
- Warehouse Grocery Store (30 minutes away)
- Local Convenience Store (15 minutes away)
- Local Specialty Grocery Store (3 minutes away)
- Online Delivery Service (10 clicks away)
.png)
Jane has access to a car, but only on weekends. She prefers larger stores, quieter shopping experiences and regularly orders heavier items online. She’s also influenced by ads and likes shopping in stores frequented by other mothers with kids like herself.
Using the Dynamic Huff Model, we can predict Jane’s likelihood of visiting different stores based on site properties and her preferences. The model predicts that Jane has a 36% chance of shopping online, a 32% chance of visiting Costco, a 24% chance of going to a specialty store, and an 8% chance of stopping by the local convenience store. It also gives store managers insights on how to increase her chances of visiting their store and whether moving to a different location could help attract more visits.
Arima’s tools, synthetic data, and dynamic modelling approach are essential for any organization that wants to optimize its sales networks and improve the likelihood of visitation and traffic.
Arima’s Dynamic Huff Model represents a significant leap forward in site visitation prediction. By accounting for a broader range of factors—including time, site properties, and e-commerce consumer preferences—it offers a more nuanced and accurate view of where people are likely to go.
Found this article interesting? Dive into the advanced version here!
.png)