A year ago everyone was talking about the impact of the Google exit from GSA supply in 2018 but since then there has been very little comment. Probably many organisations were waiting to see what Google would offer as an on-premise replacement for its search appliance but that wait has been in vain. This raises the question about what Plan B should be. I know of organisations that have decided to extend the use of its SharePoint investment but have found that that the skills needed to are not immediately available. Search Technologies are presenting Elastic as an option. There is no doubt that Elastic can match the capabilities of Google but moving to an open source option raises some strategic issues for IT strategy and for the level of internal support needed not just to implement an open source option but to continue to maintain it.

Another factor to put into the mix is how long the selection, installation, testing and implementation are going to take. In its 2017 Magic Quadrant on Insight Engines Gartner makes a number of comments on the length of time it should take to implement a search application that suggest Gartner has never actually been directly involved in one. One of the few blog posts on search implementation comes from Perficient, which in 2015 set out the elements of a GSA implementation by project phase. This shows a 12-14 week period from the time the GSA purchase had been agreed. However you need to add on the time for the selection and procurement process, which for a search application tends never to be straightforward. So that could well be a 28 week start-to-finish project, which means if you start today it could be working by Christmas 2017.

In reality things are not that simple. If you are going to replace a GSA are you just going for a like-for-like replacement or an application that will offer all the ‘insight’ capabilities that Gartner suggest are now business critical? Does your organisation have a search strategy that defines what will be needed over the next 24-36 months? Replacing a search application also throws up all sorts of interesting challenges. One UK organisation found that the first crawl by a new search engine found a security mapping flaw in one particular collection and in the initial testing presented users with a significant amount of highly confidential information. That set the entire project back around six months as all the content in the errant server had to be reviewed and new security permissions sorted out. In its profile of Sinequa Gartner issues a caution that the deployment process was longer than the average across all participating vendors and then goes on to say that Sinequa clients are more likely to have larger-scale implementations than most other vendors with more than 10 types of document repositories. Gartner seems not to appreciate that these two issues are connected.

My experience on a range of search projects over the last decade is that a rough estimate of balance of time needed for a successful implementation is 30% (3 months) on the initial specification and procurement, 20% (2 months) on the installation and base configuration, 20% (2 months) on the User Acceptance Testing and 30% (3 months) on optimising the search performance when rolled out across the organisation. That of course assumes the search team has the resources to work at this speed. So if you are running a GSA and are looking for a 2018 search re-launch you need to start the process in the next couple of months and hope that nothing unusual arises when the results of the first crawl are assessed.

[This post was revised on 12 April 2017]

Martin White