Trevor, I really appreciate your active engagement on this issue. And I would have expected results that matched your experience. I'm not sure if this is, as Marketo support indicated, "especially true when using 1000+ different words/character strings," but we're seeing (and support is confirming) things that fall outside the bounds of that experience.
I'm not sure if this is, as Marketo support indicated, "especially true when using 1000+ different words/character strings," but we're seeing (and support is confirming) things that fall outside the bounds of that experience.
Why would it be any more true with a larger number of matchable entries? Either each entry is broken up into individual matching words or it's not. This should be true of a list of 2 or a list of 2000. Any design that actually changes the behavior based on the length of the list is very, very, broken.
Robert, I don't want you to feel ganged up on (think I can speak for Trevor and Dan as well)... if anything, we mean to gang up on support, not on you.
There's def'ly something weird about their deeply researched (?) conclusion that the [contains] operator works in a way that is counter to our observations over several years, and counter to our double-checks today.
If there is indeed some undocumented point at which a list-of-search-strings transforms into a list-of-lists-of-search-strings, that's pretty close to catastrophic. Perhaps it is so -- if you can give us an exact repro case we're happy to test it in our instances -- but that should be framed as a longstanding bug. Look at my example: how could "United Kingdom" [contain] the ostensible string "United States"... and not have that be a broken feature?
Sandy, I definitely don't feel ganged up on, and very much appreciate the lively and helpful exchange that the question prompted. I've gone back to the original support case seeking clarification, as I'd like there to be a different (or at least clearer) answer.
If Marketo doesn't weigh in directly on this discussion, I'll report back anything I hear. Thanks!
Test lead:
Smart List with "United" plus other text in a matching entry, but not "United Kingdom":
Empty result:
Change Smart List to include exact case-sensitive match on Country:
Result:
Change Smart List to include partial match on Country:
Result:
Change Smart List to include an entry that is a superset of the Country:
Empty result, superset is not a match:
Unfortunately, Support is providing you with incorrect information. There must be some other factor that is causing this issue (such as the number of values used in the filter). To prove this, I created a very simple smart list:
...which returned 122 records.
Then I added to the string:
And this returned 0 records. If what Support is telling you is true, this second filter should have provided at least 122 records (with the additional string "logic" to evaluate).
Hey Robert,
My guess is that some of the 1,000+ values are pulling in things that are unexpected and should not be included in a contains filter. For example, I have tried filtering for job titles that contain CRO (for Chief Revenue Officer) and received job titles like "Microsoft Architect" and "Cross-Channel Marketing Manager"
I've found that it is best to use a combination of "Is", "Contains", and "Starts with" filters to get the correct list. For the CRO example, we would use something like:
starts with "Chief Rev"
OR
is "CRO"
OR
contains "Revenue Off"
Thanks!
Okay, support was able to confirm that their original answer was incorrect, and that you guys (thanks Josh Hill, Dan Stevens, Trevor Parsell and Sanford Whiteman!) were right in your faith in the correct operation of the 'contains' filter. What they discovered on further analysis was that with a list of 1000+ values to check against, the filter wasn't failing but was rather falling victim to unintended (and hard to spot) consequences of certain values in a list that long. Just as two examples, 'contains "HP"' also finds the company "Highpointe", and 'contains "CA"' also finds any company with "PharmaceutiCAls" in its name.
Thanks for the update -- good to know we haven't all lost it.