Mi Algorithm update September 2020
Part of the make-up of our proprietary dataset is taken from Google Keyword Planner. In Google Keyword Planner they have what they call ‘Close Variants’ which allows search terms to match to searches that are similar, but not identical, to the targeted search term.
Close variant search terms
Google Keyword Planner in the past has told us what the ‘root’ search term was. For example, if you entered ‘ankle boots for women’ Google would return ‘womens ankle boots’ and we would know the two terms are variations of each other.
In Pi Market Intelligence, in order to avoid double-counting search volume, we’d ignore the variation if the root search term was also in the calculations. The trade-off was the search engine results for those search terms was also ignored.
Recently, the Google Keyword Planner User Interface changed fairly subtly, however, the change meant that we could no longer determine the ‘root’ search term. Thus, all search term variations had the potential to be double-counted in Mi.
This prompted an Algorithm update.
Market Intelligence’s first step when processing is to now to match the market data profile of similar search terms. The search volume, CPC, and competition profiles are identical for ‘ankle boots for women’ ‘ankle boots women’ ‘womens ankle boots’. We know the average search volume is 27,340.
Instead of saying one of those terms has a value of 27,340, we now divide the value by the number of terms matched by the system. In this case, 3, giving each term the value of 9,113. Furthermore, we now include the search engine results of all three variations.
In the past, we think Google, and the Mi algorithm, has judged variations too harshly. When you look at those three words on Google trends they’re fairly similar over time and all have value.
If the workspaces that you use for Mi have minimal search terms with matched search volume profile, then you won’t see any difference in your Mi outputs.
If the workspaces that you use for Mi contain search terms with matched search volume profiles but have very similar search engine results, you will not see much difference to your Mi outputs.
You will see the biggest difference in Mi if your workspaces have plenty of matched search volume profiles AND search results differ greatly.
- More positions are considered when the Mi is run, improving the accuracy of the share of voice.
- Overinflation of search volume will not occur
- Search terms with clearly different intent that Google incorrectly marks as variations are now included.
- For example, ‘Wavy Hair’ and ‘Wavy Haircuts’ are about the same subject matter but clearly have different searcher intent.
Advanced SERP Feature selection
You can now be more advanced with your SERP Feature selection when building Market Intelligence and Visibility Indexes. This feature appears during the setup wizard and allows you to pick as many or as few SERP Features as you need to analyse.
Keep in mind that this will only remove the sites that are within the SERP Features from the analysis. The value of sites positions on the SERP doesn’t change. For example:
If you don’t include text ads and are looking for the value of piedpiper.com, the value of piedpiper.com would remain as position 4.
You can now be very precise about the players for particular features. If you need to see who owns Answer Cards across a market you can do so.