As a Midwesterner, we like to joke that “if you don’t like the weather, give it five minutes.” From sunny skies to snow showers, the weather here is, to say the very least, an extreme variable.
However, no matter where you’re located, weather plays a factor in our everyday lives … including our shopping habits — both online and brick and mortar. To keep up with these fluctuations, retailers are turning to weather intelligence to predict retail sales.
I know what you might be thinking — the weatherperson is NEVER right! How can we depend on this data? In all reality, historical data shows that 5-day forecasts are 95% accurate, while 7-10-day forecasts are around 80%. A bit more predictable than your uncle’s bad hip, right?
“Over 90% of a business’s annual weather-driven sales come from day-to-day changes in temperature and precipitation that influence consumer shopping patterns and behaviors.”
— National Retail Federation, Five Myths About Weather & Its Impact on Retail
Plan for the Weather
A weather forecast not only helps you decide whether or not you’ll need an umbrella or flip flops, it helps retailers determine:
- The prioritization of shipments of stock to stores
- Product promotions and markdown strategies and timings
- Appropriately staffing stores
James Smith, Founder, Demand Data, a retail focused analytics company that helps retail partners better plan their inventory and staffing needs for their locations, shares his experience utilizing weather intelligence.
“Our clients are looking to be more proactive with data and ensure that their resources (stock and personnel) are deployed forward in the most efficient manner possible.”
— James Smith, Founder, Demand Data
Demand Data uses Alteryx to model a number of known factors (including day of week, season, time of month, holidays, etc.) and now uses real-time weather forecasts to predict changes in sales figures in stores — including sales figures across key categories.
“We model all fixed factors, like holidays, weekends, or seasonalities, and add in variable factors, like the weather. From here, we’re able to create predictions for products over the next 14 days based on live weather forecasts. In turn, these predictions influence business processes.”
— James Smith, Founder, Demand Data
This process involves several steps,
including:
- Collecting historical weather data from online sources
- Geospatial mapping of weather data to store locations
- Normalizing sales data across stores and categories
- Building a predictive model (linear regression) based on weather (conditions, including maximum and minimum temperature)
- Collecting live daily weather forecasts for the next 14 days
- Publishing a conditions favorability report across stores and categories
Demand Data used Alteryx Designer to build the workflows and Alteryx Server is running on a regular basis to
provide updated patterns. Some of Demand Data’s key steps:
- Data is primarily housed in Microsoft Azure, Snowflake is the main repository for sales data and Alteryx Server is deployed in Azure.
- Sales data is coming from Snowflake, we rely on the In-DB tools to process billions of rows of scan level data.
- Weather data history is also coming from Snowflake. Data is available in JSON which has historical conditions every three hours from over 20,000 data points.
- Weather data forecasts are provided via an API from an online source (JSON format). This data provides 14-day forward predictions of minimum and maximum temperatures.
- Tableau dashboards provide executives with high level visibility of weather conditions in different regions. The dashboards also show high potential categories (like ice cream before a heat wave, or umbrellas before a rainstorm.)
- Individual category managers can use Alteryx Gallery analytic apps to download weather-based factors for their categories and use these outputs to adjust their short-term inventory and promotion allocations.
- Regional managers can use Alteryx Gallery to send reports of locations that are expecting an exceptionally high or exceptionally low amount of traffic in the next several days. These are exception reports which they can use to adjust staffing plans.
On the Radar
Within the first few weeks of using their weather prediction model, Demand Data customers have seen
several key benefits, including:
- Deferment of a planned markdown of winter stock days before a major cold front
- Reduction of the stock out in leisure locations during a high temperature/good weather public holiday by planning an extra midday delivery in key locations (No running out of hamburgers and charcoal!)
- Adjusting the space allocation of summer products in stores in more temperate locations
“Efficiencies that I always knew existed are becoming solvable through the availability of data. We are also becoming less and less limited by how data needs to be organized to extract value from it.”
— James Smith, Founder, Demand Data
Watch This Next.
Tap the combined power of automated analytics and real-time
cross-data sharing to create validated solution
frameworks.