Commercial Search Intent

One of the questions presented to me recently was how far Google accounts for search intent in presenting search results to users.

As we know, users search with different intent, either because they are at different points in their buying process (researching options, researching products, comparing prices, or looking at websites to buy from) or because they are simply looking for information as opposed to buying.

For the purposes of this discussion, I will define “high commercial intent” as being equal to a higher likelihood of users actually looking for a site to buy from. Likewise, “low commercial intent” I will interpret as users only looking for information.

Investigating Commercial Intent in Organic Web Searches

Using Microsoft adCentre Lab’s online commercial intention detection tool (4/11/10 – appears to be currently unavailable), I carried out a simple experiment using 8 keyword phrases to investigate whether the commercial intent presented by Google.co.nz top search results actually matches the commercial intent that is likely to trigger the keyword search in the first place.

The results are as follows:

Google Organic Search - Search Intent vs. Commercial Intent

Google Organic Search - Search Intent vs. Commercial Intent

The four graphs in the first row use keywords that most likely are not too commercial, whilst the four graphs in the second row use keywords that are likely to be more commercial.

With the first three graphs in each row, the red and yellow lines are not too far apart:

  • Where the search query (red) commercial intent is clearly defined, the average site commercial intent (yellow) is always on the same half of the vertical axis.
  • Likewise, when the search query commercial intent is ambiguous (gran turismo 5, avatar dvd), the average site commercial intent hovers around the halfway mark of the vertical axis (0.5 probability) due to Google presenting sites with a diverse range of commercial intent (Query Deserves Diversity?)

However, the last graphs in each row look rather anomalous:

Life Insurance”:

  • Personally, I think the search query commercial intent should not have been that high (the Microsoft tool gives a 0.90 commercial intent probability; something which I disagree with). Could this mean that this is an anomaly and the Microsoft detection tool is…wrong? (* gasp *)
  • Google seems to agree with me. The resulting yellow line has a very low commercial intent due to most of the blue lines also showing very low commercial intent.
  • Therefore, if we assume the query’s assessment was wrong (i.e., we assume that the “life insurance” query should have a very low commercial intent), the adjusted result would be perfectly in-sync with the other graphs as analysed above.

“Buy avatar dvd”:

I agree with the assessment tool’s assessment, from the query to the websites in the search results. Furthermore, Google provided a variety of websites in the results, including those with very low commercial intent:

As such, the average score becomes very confused and thus not in-sync with the query’s assessed commercial intent. In short, I’m not entirely sure what is happening with Google here for this particular query. Comments welcome.

Commercial Intent Factoring – Conclusions

In a nutshell, I think it is a very valid to assume that Google seriously (albeit not heavily) factors in the commercial intent of the web pages in its organic web rankings.

However, considering that even when the search intent is clearly defined, we still see 1 – 3 websites at the opposite end of the spectrum within the top 9 results. Google therefore also tries to inject the QDD factor (Query Deserves Diversity) into the search results at the same time.

Implications for Search Engine Optimisation

With SEO, the implication would be whether we are satisfied with only being associated with a certain level of commercial intent, or whether we should try to capture both ends of the search funnel at the same time.

There are instances where it would be worth trying to capture as many users as possible, by utilising a site structure and SEO content architecture that can be associated with both types of search intent (non-commercial as well as highly commercial).

As for this research, in itself it can be improved with a wider set of search queries.