When making sense of wave-on-wave changes, always consider the following:
Could seasonal variation be having an impact?
If there’s a shift in shopping patterns between two waves, it could be influenced by one wave being just before Christmas and the other just after.
Are there local market factors at play?
We see some of the biggest jumps in data when a country’s government intervenes to allow/prohibit a particular behavior or service. For example, when a social media service gets restricted in a particular country, there’s often a noticeable impact on global usage - particularly if that country has a large population. In other words, don't assume a global trend is true of all markets you have selected - it could be being driven by just one or two of them.
Have there been any changes to the question, or any other aspect of the survey set-up?
Changes in question format or wording can have an impact on responses. For example, when we add more answer options to a question, we sometimes see decreases in the existing options as respondents identify more strongly with the newer options and select those instead. Changes like these will be noted in the question details on our platform.
Is the change meaningful?
While it’s tempting to see changes of 1-2 percentage points as meaningful, particularly when looking at more niche audiences, they could fall within the margin of error (making them statistically insignificant).
For example, if a country had a online population of 3,000,000 and a sample size of 1,500, the margin of error would be 2.5% (assuming a confidence level of 95%). Therefore, a wave-on-wave change smaller than this would be statistically insignificant. However, in other markets, smaller changes could be significant; the US has an online population of c. 200,000,000 and a quarterly sample size of 25,000 in GWI Core, meaning the margin of error is less than 1%.
In short, you should consider both the universe size and the sample size when interpreting wave-on-wave changes. Although you can find margin of error calculators online to make precise calculations, avoiding reading too much into changes of 1-2 percentage points when working with smaller sample sizes is a good starting point and sufficient for most analyses.