Re: [SearchCoP] Acceptable % of no results searches?
- Hi Mary Jo - I think Chirag has it right - that it shouldn't be too
much of a focus.
I can share that in several different intranet search engines I've
worked with the % of searches returning zero results hovered
consistently in the range of 5% to 8% of searches. This range has not
I've also dug into the details of the searches in these situations and
found that they are almost always well into the "long tail" - the most
common terms are rarely used more than once or twice (in a set of tens
of thousands of distinct terms). When looking at the most common
among these (as common as they get, anyway), they tend to include
misspellings which commonly I find trigger the spelling correction of
the search engine, which offers the "did you mean X?" link. In other
words, the most common zero returns are easily corrected by the
Another factor to consider is to make sure you understand the full
context of the user's search. If your engine supports context
sensitive search ("search within this category or space") or faceted
search, just the search terms by themselves returning zero results
doesn't mean anything, so you need to have a way to keep track of the
other factors used for each search.
I would also agree with Chirag that it's important to try to normalize
the terms to find patterns, though that is a significant challenge to
do repeatedly over time, I find.
Interestingly, I get more feedback in the other direction in terms of
results - users voice the opinion that they get too much back from
their searches. I don't see large result sets to be an issue in and
of themselves if the relevant stuff is at the top (the fact that there
are 10,000 results is irrelevant if what the user is looking for is on
the first page. Of course, relevancy is its own challenge! :)
I also have gotten feedback along the lines of, "It doesn't matter
what search terms I use - I always get results even if I use
completely made up words!" That's not literally true, but I have seen
some, um, interesting examples of odd terms resulting in search
results, but this is more a reflection of the very broad corpus of
content being indexed than anything with the engine (online
discussions, for example).
I hope this is useful for you.
On Fri, Jun 4, 2010 at 2:38 AM, Chirag Gandhi <c_gandhi@...> wrote:
> Dear Mary Jo,
> We have started advising clients not to get too hung up about how many 0 results appear, but rather take a more qualitative approach and assess if these searches are really something which they would like to respond to. At the end of the day it is a value engineering decision.
> We present this data as a long-tail and suggest that the client really focus on the top 20% of graph. One caveat is that the data needs to be normalized. We had one particularly difficult case because of a strange product name which was leading to many variations - you need to make sure that these variations are normalized.
> From: Mrs Mary Jo <jborrell@...>
> To: SearchCoP@yahoogroups.com
> Sent: Thu, 3 June, 2010 22:42:30
> Subject: [SearchCoP] Acceptable % of no results searches?
> I'm wondering if any of you have an 'acceptable' percentage of no results searches that you use as a goal and what that might be. I'm specifically interested in site search, but examples from intranet search would be interesting as well.
> It seems that there will always be searches that don't match content (i.e. someone always searches for a curse word, just for kicks) but I'm wondering if there is a threshold that is more-or-less acceptable.