Does weather affect conversion rate?

The Inspiration

A few months ago I was talking with a customer and friend and he told me that "the weekend's sales were very low because of the good weather""! I asked him what he meant and he said that "when the weather is good, people get out, so they are not in front of their computers to place orders". I didn't pay much attention to it, but as he was talking and explaining I wondered if we could measure and certify what my friend had as an intuition.

The measurement

So we set up a script that stores on the server that the site resides the weather conditions and temperature every five minutes. Then we added a custom variable in Google Analytics that stores the weather conditions and temperature for each visit. The result was that after a few days we could have a report on sales and conversion rate depending on weather conditions like the following:

also, for the temperature (the numbers are degrees Celsius):

These were fun, but they don't allow for any clear conclusions. The sample is too small.

After a few months

This was forgotten and I remembered it a few days ago so I opened my friend's Google Analytics account and created ​​the following report:

The analysis

The above is interesting! The sample is large enough for some conclusions. If we compare the conversion rate for the fair weather with 1.48% with the non-fair weather average conversion rate 1.69% we get a decrease of 12.4% on sales when the weather is good!

Confidence Level

I loaded these numbers on an A/B testing confidence level calculator that shows whether these figures are statistically significant or a result of randomness. The result is a 95% confidence level. Not bad. And even better, if you do the same comparison between the fair weather and then partly cloudy and mostly cloudy, the difference increases to 13.8% and a degree of confidence to 96.7%! So it seems that maybe my friend's intuition is right...

Imponderables

These findings, however, again can not be considered safe, as at least 2 imponderable factors that are not measured are involved in the results. One factor is that the weather and temperature variables we used were for Athens, but it's likely that site visitors from other parts of Greece were not in the same weather conditions and temperature. To have a proper measurement, we should geolocate each visitor's IP (with the inherent inaccuracy of this method) and then send to Google Analytics the weather for this area. This would dramatically increase the setup complexity of the experiment though and would get us out of the "a for fun measurement" area. The second factor that may be invalidates the measurement is that the number of days with fair weather may not be evenly distributed in time. The experiment was running from February to July, so maybe the fair-weather days were more frequent as the summer approached. So the conversion rate may be influenced by factors other than the weather as time passes.

Finally, while the original idea of the experiment was that "the fair weather, kept people from sitting in front of their PC's in order to shop", the above measurement ultimately did not relate to how traffic increases or decreases based on the weather, but to how the conversion rate increases or decreases based on the weather. So the question should be whether the weather changes the mood of visitors to this site to shop.

Possible explanations

Possible explanations that have passed through our minds as we discussed the above at Netstudio were:

• 1. The weather is good, so even if users are in front of the PC and visit the site, they have their minds to go for a ride, not to shop.
• 2. The weather is bad, so visitors shop in order to cheer up.

So what?

Apart from the interpretations that could be given, the most important is how this could lead to action, assuming the weather does affect the conversion rate; Do not forget that we always look for "actionable data", data that help us make decisions about actions that will bring improvement.

One idea would be that if the conversion rate drops in good weather, then maybe when the weather is good we should reduce the AdWords spending in order to maximize ROI. Still, for such a rationale to be complete, we would have to check how the weather affects the "Days to Purchase", the time that elapses from the first visit to the order completion. And maybe we should even watch the weather forecast.

One idea would be that if the conversion rate drops in good weather, then maybe when the weather is good we should reduce the AdWords spending in order to maximize ROI. Still, for such a rationale to be complete, we would have to check how the weather affects the "Days to Purchase", the time that elapses from the first visit to the order completion. And maybe we should even watch the weather forecast.

Comment

Have you noticed similar variations of your sales or your site's conversion rate depending on weather or other external factors? What actions did you take? What other factors could ruin the validity of these conclusions? Comment below.