In terms of what to do instead, at least on the search end of things, we have lots of high level things to work on, specifically: * Integrating page view statistics into the scoring * Calculating page rank of the wikipedia's and integrating that into scoring * Extending the relevancy lab MVP to be more useful * Integrating the above scoring adjustments into the replacement completion suggester All of these have a good bit of work that can be done without pushing any brand new code to mediawiki prod in december. The php end of things is, comparatively, the easy part. Research still needs to be done on how to weight page view statistics, how to weight page rank, how to move this data from hadoop to the elasticsearch cluster, etc. etc. We can easily fill a month with this work and be able to test how it improves the search results in january. On Tue, Nov 3, 2015 at 3:42 PM, Chad Horohoe <chorohoe(a)wikimedia.org> wrote: