Query Formulation in Microsoft Teams Mobile

The pandemic has changed the way we work - we work on laptops, tablets and mobiles. Everywhere and anywhere. With the COVID-19 outbreak, Teams hit 145 million daily active users in early 2021. These users together generate 1 billion search queries every day. Users reflected that the search journey was one of the biggest factors impacting the overall impression of the Teams user experience, alongside video and audio quality.

To help users perform their searches quickly and effortlessly, Query Formulation (QF) capability was rolled out for search function in early 2021. This user research was to understand user's searching behaviours with QF on mobiles, and to validate the new UX models for the next phase of QF. The 2-week user research consisted of diary studies, usability tests and competitor analysis.




User Research



*Due to Non-Disclosure Agreement, this summary discussed the methodology of the research without touching upon the research outcomes.
Journaling: users logging what they just searched by screen recording or writing


People use search everyday with their smartphones, but they may not pay attention on how the search bar interact with them at each of the 2 steps involved - "Zero Input" and "Result Page".

Zero Input is when users click or tap the search bar before inputting anything. The most common thing that people would see at this step is their recent search history. After users have input the keywords and proceeded, they will be directed to the "Result Page", which displays everything that the search engine could find from the external or internal database.

Query formulation (QF) is the stage in-between - the interactive information access process. In laymen terms, the formulation of "Search Suggestions". For example, typing "Hap" will prompt query suggestions such as "Happy" or "Happy Birthday". Depending on the user needs and robustness of search engine, QF could have greater capabilities. Some QFs do not only show query suggestions, but also entities like websites or files, something that users can consume immediately without making further search.

Key steps in search

Comparisons between QF and result page

Dilemma: make a better QF, or a better result page?

QF provides the most immediate response to users in a search journey. When QF is getting more mature, more traffic may be diverted to QF, as users may have already fulfilled their need from the search suggestions. But QF can never replace result page. If more features are added to QF, it will greatly affect the entire search performance, because the search engine tries to front-load everything for you when you input a new character. So when Microsoft Teams was planning for the next search experience enhancement, there was a heated debate on where the equilibrium laid.

The hurdles faced in the pursuit of this equilibrium came from the lack of data, due to Microsoft's vigorous data protection policies. There was no way for product managers and designers to obtain empirical data on user's queries, and understand how QF works with users for mapping enhancement solutions.

The dilemma was in front of us but the necessary data was not. We needed a way to overcome the hurdles and understand user's behaviours in QF in Teams Mobile, and collect feedback on different new UX models of Zero Input and QF.

User research design

It is technically challenging to understand user behaviour (what and how they search) through moderated testing, as search is a very short and spontaneous action. We need to observe user behaviour in their real-life context. So we engaged interviewees to log their search events on a daily basis for a week. After analysing the journal record, interviewees were invited to a follow-up usability test, with further follow-up questions to their logs.

Before the start of the diary studies and usability test, we did the competitor analysis to gain insight into what QF other service providers offered when it came to search. It also helped me as the interviewer to deepen understanding on the market trend.

Screen captures of competitor analysis

Qualitative research part 1 - diary studies

Having invited 10 users via usertesting.com, I gave them a 15-minute briefing on the diary exercise with them while ensuring that they understood the requirements.

Interviwees were then asked to record how they search on Teams for a week. They needed to record their intent, the query they used and how they navigated. They could choose to think out loud through screen recording, or wrote it down. Eventually, more than 80 logs were collected for analysis.

A 15-minute call to familiarise users on journaling their search events

Qualitative research part 2 - usability test

We then invited 6 of them to join the 2nd part of the research, which was a more in-depth usability test. Interviewees were asked to complete 3 tasks using our prototype, in which some new search features were embedded. We then evaluated the models from the perspectives of discoverability (visibility of the product features), readability (intuitiveness to understand) and usability (ease of use).

Part 2 of research: usability test on new prototype

Design impact

The user research result was shared with the core team of Microsoft Search Assistant Intelligence division. With a more precise understanding on the user behaviour on QF, product managers and designers could make better judgement throughout the design process.

Based on the positive feedback on certain designs, we had a clear direction to switch focus in exploring alternative designs going forward.

Other work