Data sources and methodology
Where eForecast's scores come from, what they measure and what they do not.
Sources
| Source | Used for | License / attribution |
|---|---|---|
| Google Trends google.com/trends |
Monthly search interest from 2004 to today, and weekly for the last 5 years | Data source: Google Trends (https://www.google.com/trends). The scores shown are a processed, derived form of this data; they are not raw Google Trends data. eForecast has no partnership with Google and is not endorsed by Google. |
| Wikimedia page view statistics | A second measure of interest in the product topic | Wikimedia Foundation, CC0 |
| Official public holiday calendar | Religious and public holidays | python-holidays, MIT license |
| Special occasions | Mother's Day, Black Friday, Teachers' Day, etc. | Our own calendar |
What we measure and what we don't
- Scores measure search interest — not sales volume, price, competition or profit.
- Google Trends scales each term from 0 to 100 relative to its own peak; the absolute search volumes of two products cannot be compared directly. For this reason, "fill rate" (the share of non-zero periods) is used as an indirect indicator of volume.
- Probabilities such as "17 times in the last 19 years" are past frequencies, not model output.
- Common spikes that appear simultaneously across all terms in a country's data (breaks in Google's data collection) are detected and excluded from the calculation; this is indicated on screen.
- Weather effects are not currently used.