The inside story of enterprise search supporting Covid-19 information requirements in a multi-national company


Over the last few years Paul Cleverley and Simon Burnett (Robert Gordon University) have set a very standard for enterprise search research. Mind you, there is so little published that they almost have the field to themselves! You can catch up on all of Paul’s research on his blog.

Last week Paul and Simon, together with Fionnuala Cousins, published a paper that makes a significant contribution to understanding search processes inside the enterprise. The paper, entitled The impact of Covid-19 on enterprise search is open access in the Journal of Information Science so I am not going to try and summarise the paper. There is no excuse for you not reading it – slowly!

The study was conducted in a large knowledge-intensive multinational corporation with over 100,000 staff, the majority of which were office based and located around the world. The search index contained several hundred million items including Intranet web pages, office documents and discussion threads. That size of index is not uncommon, and one of the challenges in enterprise search research is managing the implications of these index sizes and the number of employees undertaking search tasks.

The two objectives of the research project, which took place between the end of January and the beginning of May 2020, were

  • To what extent has COVID-19 affected enterprise search usage?
  • What insights can be inferred from COVID-19-related search queries in an enterprise?

In the course of the research over 2.5 million queries were analysed over a roughly four-month period. Search query volumes per day, number of unique users per day and the average number of queries per user per day were selected as the core parameters. There were 2,500 COVID-19 related query strings identified in the query analysis. In my view this demonstrates the importance of clustering and categorizing search query terms rather than producing query count/rank charts on the basis of single terms.

Figure 8 in the paper shows that explicit COVID-19 search queries prior to 16 March were almost exclusively single words. Subsequently, multi-word queries (up to seven-word queries) became equally as important by volume as intents narrowed.  From this the authors infer there is a difference between how people searched explicitly for COVID-19 information running up to lockdown versus post-lockdown.  They consider that it shows a transition from broad, single-word exploratory like explicit COVID-19 queries which may be driven by intents such as ‘reassure me’ and ‘educate me’ to narrower task-driven queries related to safety, business impact (such as ‘covid impact to supply chain’), strategy, policy and response.

Hopefully there will never be another pandemic but the authors highlight the importance of search analytics in tracking changes in search query patterns in response to any major external or internal event. I was working for the IMF in Washington on 9/11 and post the event I had access to the search logs of both the IMF and the World Bank but never thought to ask permission to retain them for research. What a missed opportunity! 

In the paper the authors raise a number of issues where they have had to make some assumptions about the implications of the results. As a result, there are a number of open issues just waiting for the research community to realise the importance of undertaking similar projects in other organisations.

Martin White