If you're in either the Monitored Keywords or the Keyword Visits Estimates modules (or other places where we show Search Volume data), you may find that some of the look quite odd (especially if you're used to seeing Search Volume numbers around to the nearest 10 or 100).
However, there's a very good reason for this - some of our Search Volume numbers require us to forecast them for the period you're looking at. If you're looking at periods like 'Weekly' or 'Last 30 Days' then we will be using forecasted Search Volumes. Why? Because Adwords takes some time to update its Search Volumes, so there's a gap in the data we import and we need to fill that gap.
We do this rather cleverly, by using a fairly complicated algorithm to assess the pattern of the Search Volume over the preceding period.
Essentially, the algorithm seeks to answer four questions:
- Is the Search Volume over the previous 12 months linear (i.e. a straight line)?
- Is the Search Volume cyclical (seasonal)?
- Was there anything particular about this keyword which suggests it a ‘fad’ keyword (i.e. keywords that come in and out of fashion or have a limited ‘lifespan’ - e.g. in relating to particular trending topics or news stories)?
- Does it not fall into any of the above 3 categories?
#1 Keywords Search Volumes
This is perhaps the easiest to calculate as we can just 'continue the straight line', which will be a linear average of previous monthly Search Volumes.
#2 Cyclical Search Volumes
This is a little more complicated as we have to judge variations from a certain co-efficient and if we determine that the keyword is displaying evidence of being seasonal or cyclical, the algorithm will then check previous months' volumes in order to work out whether the current season or year is likely to be worse or better than the previous one. Clever, huh?
#3 'Fad' Keywords
If the algorithm notices that the Search Volumes over time suggest the keyword is going 'out of fashion' then we will take this into account in predicting subsequent volumes.
This a 'catch all' option. If the keyword (and remember, we're analysing every single one) doesn't match the above 3 rules, then it falls into this last category, in which case we will simply take the previous three months' average and use that.