Hi Tom,
Thanks for taking the time to provide us with this feedback, this is precisely the sort of information we are looking for to better understand how our customers use our tools.
I don't think you are doing any wrong , however there are some differences between the old search platform and the new one that can affect the results.
Some of these are just the nature of two different technologies, but some were conscious decisions for this first release.
This is a not a 'ship it and forget it' project, we will continue to review both specific customer feedbacks as well as performing data analysis to understand usage patterns, usage needs, content gaps and poor result relevancy as we go forward, and we will bring in improvements and tuning changes over time as we learn how the platform reacts to our user base.
Having said all of that, I'll try and address some of your specific points:
Product Family filtering vs specific Product filtering.
This was a conscious choice for this first release. There were a number of reason why we made this choice that I can't get into here in detail.
We understand that this would be a loss for a section of the user community that knows very specifically want they want to find, and we are still looking at how we might be able to provide options that work well for all of our customer base and all of our content.
This is one of the reasons why feedback such as this is so valuable. We can see what users do and don't do, but we don't always understand why and what they really wanted to achieve.
In the meantime, one option you can consider is including the specific product in the query. This won't act as a Boolean filter, however it should ensure that content that references that specific product ranks higher in the results than those that do not.
Number of results.
There are certain types of query for which the best results will be obtained by assuming the keywords are to be treated with a logical AND. You only get results that contain all the key terms
There are other types of query where a certain amount of 'fuzziness' is preferable, where you treat the keywords with a logical OR.
The reason you see a larger number of results is because the platform treated your query as a logical OR rather than a logical AND, so you got a much larger set of results. The results should, however, still be ranked such that those that have all or most of the terms higher than those without, and those with the keywords near each other, or in the Title and Body higher than those that did not.
This is one of those areas that we are looking at post go-live data to see whether the way we have it set up is working, or needs adjustment in the criteria used to determine when to use one approach vs another.
There is not currently any way for the end user to control this behaviour.
Google
I can't speak with any authority on how the Google platform works, however I would point out that a difference in result counts could be attributed to a combination of factors:
- Not all PTC content is indexed by Google. Even for articles, we specifically exclude some content.
- The difference in how the query is considered from a logical AND / OR perspective.
- I would have to assume that Google will only crawl the un-authenticated Article content. That is limited to the Article Title and Description only. The PTC search indexes the Cause and Resolution as well, so any given set of keywords can match a broader selection of Articles
Hope this helps answer some of your concerns.
If you have any follow-up questions, or further feedback or comments, please let us know.
