On Monday (Oct. 27), Advertising Age published an interesting article on how Walmart bases marketing decisions on weather data.
I sat down with a couple of my co-workers (User Interaction Designer Tony Burbage and BI Developer Carole Cazenavette) to pick their brains about the process, and how iQmetrix BI could help our clients perform similar data analyses.
The idea of connecting weather to sales forecasting doesn’t seem that farfetched. Have you guys heard of this before and if so, which retailers were looking at it?
Tony: The idea of our new BI Day Report came up after looking at our calendar heat map visualizations. We have calendars that (based on intensity of color), show the relative amount of sales for each day over time. The darker the color, the more sales. The next step was trying to think of what people would be interested in when clicking on a particular day.
Looking at our heat map visualizations for sales, we wondered if sales fluctuated much based on good or bad weather.
We looked at sales, punched in hours, top sellers, and the regulars. Then we thought, what if bad (or good weather) would be a factor in sales? If it was a beautiful day out, people may be more inclined to visit a street-level store. If the weather is bad, perhaps a mall location. After interviewing some customers, we decided to add weather information to our new Day Report.
Carole: You can see in our Location Report (above - click to expand), there's a calendar heat map, and each day is a clickable link that redirects you to a Day Report.
Below (click to expand) is a Day Report for May 22, 2014, with a weather forecast of light showers -- sales went down compared to the yearly average on a Thursday.
Here's another Day Report (below - click to expand) from Sept. 19, 2014, a sunny day -- sales sent down compared to the yearly average on a Friday.
More specifically, have any of our RQ/BI clients actually built reports to cross-reference weather and sales? How easy is it to build such a report in BI?
Carole: For now, as far as we know, nobody built this type of report using iQmetrix BI.
For our weather/sales comparisons, we used historical data from a service called World Weather Online.
To develop this report, I am calling a webservice for weather historical data from World Weather Online. We've subscribed to this service for a year and we'll continue to do so, if our customers find this information useful. I didn't develop this report foreseeing customers would build their own reports based on this data, but we'll definitely keep it in mind for the next release.
The Ad Age article reported that Walmart switches out its displays based on weather forecasts (sunny, high of 75 = berries; sunny, high of 85 = salads; cloudy and 80 = steaks; expected high of 90 = burgers).
Clearly our clients don’t sell food, but beyond overall sales, would you guys predict any fluctuation over types of wireless sales based on weather?
Tony: I would say that right now, our weather reporting is more historical. So, if you see that sales were quite bad on a particular day, you can now drill into that day, and see if weather was potentially a factor. We are trying to find the little things we can report on, that help make sense as to why something happened. It's more of a "causation" than forecasting point of view.
Because wireless sales are so different from grocery sales, it's hard to determine specific sales trends. But Walmart's findings are thought provoking.
Carole: I would agree. Because wireless sales are so different from grocery sales, it's hard to say. Our goal is to at least give clients as much information as we can, based on a specific day of sales. For now we don't plan to trend sales based on the weather data, at least not until there is compelling data for our wireless retail clients.
But the findings Walmart has reported certainly are thought provoking!