Predicting target traffic. How do we analyse the weather?

Objective

To analyse the impact of the weather on shopping centre footfall, generating targeted traffic to the outlet.

Context

Every retail chain and shopping centre knows how weather affects customer traffic, but few of them measure and analyse with precision how one factor correlates with and depends on the other. There are even fewer of those who incorporate such variables into their development and promotion plans, as they are seen as poorly predictable and virtually impossible to analyse. Here is an overview of how such data can be incorporated into development plans.

Solution

When we solve network optimisation tasks, select optimal locations or do predictive traffic analysis and calculate the organisation's potential financial results, we pay attention to all kinds of digital data available to us. Usually, this includes social and demographic parameters that are mandatory for the calculations, transport and other urban infrastructure characteristics, as well as the customer's internal data.

Our approach is: if there is something extra to learn, it should be learned. Anything that could be algorithmized and enriched with additional data should be digitized and added to the calculations. There are parameters which, to a greater or lesser extent, are usually intuitively understandable. For example, everyone understands that shoppers go shopping, to cafés or the cinema differently depending on the weather. In warm weather, urban dwellers are more likely to go to a park, while in windy and cold weather, they are more likely to warm up in a café.

We do not know how to forecast the weather, but we do have detailed retro data for the last few years, and we can build models for different scenarios. If we are talking about a monthly forecast, we can assume, as an example, three probabilities: whether or not it will snow in November, whether there will be frost (temperature data in general) and whether it will be windy. Roughly speaking: the weather will be good, bad or neutral. Each scenario calls for the customer to have a different product stock, marketing plan, promotion plan, change the way they work with the product range, advertising, bonus programmes, etc. One simple example: an advertisement for hot drinks with honey and ginger will draw attention to the catering outlet in cold, windy weather, but will rather evoke little emotion in warm, sunny and windless weather. Together with dozens of other variables that are involved in the calculations, weather data is an important part of the traffic attraction strategy. If we are talking about shopping centres, with the weather factor in mind, different offers from different tenants can serve as an additional catalyst for increasing traffic to the shopping centre as a whole. If they can be combined into a consistent marketing strategy or programme for a shopping centre that is also positioning itself and promoting itself on the market, the effect can be synergetic. Notable examples are Black Friday sale periods, Christmas and New Year offers and other festive promotions. If a shopping centre offers a potential customer a hot drink special, discounts on coats, scarves, and umbrellas, the cinema is showing something like "Autumn in New York" or a hyped premiere, tenant cafes offer an autumn menu, and design shops, interior and home textile shops offer promotions on plaids and fireplaces for the home and countryside cottages, all of this will convey the shopping centre's consistent message that "we are warm, we will keep you warm and fed and offer you discounted goods that you need the most right now, given the weather out there." This is just one example involving the autumn cold, but we have a pronounced seasonality across much of the country, and the weather affects sales all year round to a lesser or greater extent. Instead of using this factor "intuitively", you can build it directly into your action algorithm, marketing plans and projections of financial results depending on the scenario being implemented.

We can make projections based on required and weather-dependent actions using retro data on the weather. Such simulated quarterly projections make it possible to manage not just 3 options but 27 combinations of monthly weather for three months ahead. At the same time, the actual forecast reduces the number of variants and triggers a certain algorithm, affecting calculations of, for example, traffic, and the popularity of certain goods that depends on seasonality and weather. The criteria for “bad” and “good” are different for each project. In the case of a shopping centre, cold weather in autumn is good weather because it brings audiences to the shops and cafés. Good, warm weather, on the other hand, draws people away from the shopping centre to the streets and parks; this option triggers a different action scenario. It should be emphasized here that it is about taking action, not about passively waiting for the cold weather and for the traffic to come back. If there is a model already built, this is simply another part of a clear and understandable plan.

Result

The main result that the customer gets from the analysis of additional data, such as (but not limited to) weather, is predictability and an assessment of the impact on the final financial results. The more factors and attributes become part of the calculations and are loaded into analytical models, the more accurate the revenue projections and the more reality ceases to be a "risk factor" or a white spot on the map and becomes another option in the company's overall growth and profit increase plan.