Virtual assistant salute works in several channels - voice interface, haptic interface, text chat. Also, on a hypothetical level, it is possible to use a pre-recorded voice message.
The task of this project was to find out how our users switching between different channels within Virtual Assistant in order to develop more convenient interaction scenarios.

My role: VUI Researcher, methodology designer, data analyst
Credits: Product team of Sber Bot
Platform: SberBank mobile app (Actions on Apple and Android plateform)

In our experiments, we needed an urgent resolution scenario. We decided to use the scenario of blocking the card as a result of fraudulent actions and its subsequent issuance.
We decided to interview people who had to solve a similar problem in different environments and circumstances - "it was very noisy, their hands were busy, it was not comfortable to talk in the presence of other people", etc.
Then we processed the interview data and verified it with an online survey. From the collected data, we built an alluvial chart that demonstrates how people switch between different Virtual Assistant tools, taking into account our user segments.

• In depth interviews with 8 respomndents
• First iteration of Channel Migration Diagram
• Quantitative research using Oprosso
• Final iteration of Channel Migration Diagram and analysis

NOTE: Some images are low-res or blurred and cannot be expanded for confidentiality purposes.

We developed a matrix of cases where a person has to switch between Virtual Assistant tools. During the interviews, we identified clients' real-life experiences, pains, and the communication tools they used to solve them.
The research was conducted as an in-depth interview with 8 respondents of Salute Virtual Assistant users from different segments.

We interpreted the results of the first in-depth interview into the Cross-Channel Migration Matrix. The chart shows Which tools users will choose when faced with different types of interference and obstacles.
Visual analytics allowed us to draw our first top-level conclusions and tie the qualitative research data to them.

The left side of the diagram shows various cases and obstacles: bad hearing, bad screen, too many people around, and so on. On the right are the tools that users prefer to use depending on the context. Also, the research board snippet includes contextual statements from users that allow for a better understanding of the trigger and motivation.

To better understand what tools and scenarios the user would use in the context of the obstacles and triggers, we put together a tool matrix showing all the possible solutions, both through the visual interface and through the voice tools.

In the final phase of the project, we supplemented our research with analytics data and an unmoderated survey of 70 respondents. This allowed us to build an Alluvial Chart of client switching between Virtual Assistant channels. We also overlaid the chart with qualitative interview data at touchpoints to better understand not only what users do in certain circumstances, but also why they do so.

The diagram demonstrates how users of two genders and three age segments will switch between different tools or remain in the pre-game interaction channel in the context of changing circumstances. The vertical line shows the different channels - voice, text, vidget, etc. Horizontally, different situations and challenges.