Driving adoption of data to build better products with voice interfaces

Product teams need to continuously monitor their product usage data to understand the user pain points and iteratively improve the product experience. As a product manager, you need to have tools that improve visibility, make it easier for the team to comprehend, and derive actionable insights from the product data. A few months ago, I wrote an article that outlined the process of building tools to enable that for your team. I’d highly recommend reading that article as well, but here is the summary:

1. Identify your product assumptions. Create a list of data points required to prove or disprove those assumptions.
2. Ensure that the above data points are being recorded.
3. Build APIs to query data and model it in the required format.
4. Share product usage data with the team and help them derive insights out of it.
5. Connect those APIs with visualization tools and then create widgets and dashboards as per need.

Now that you have a working dashboard, you’d expect everyone on the team to open it up frequently and use relevant data for their work. But does that really happen? Does the team start consuming data for their everyday decision making? Nope, not really.

Here are some of the challenges I faced when I was trying to drive adoption for data dashboards inside my team:

1. People heavily rely on their intuition for making decisions and do not refer to data even though it is available. This is normal human behavior!
2. During meetings, data was often needed to propel the conversation forward but people would proceed with behind-the-envelope calculations for the sake of everyone’s time and avoiding friction to access data sitting inside some view of the dashboard.
3. Over a period of time, each function of the product had its own dashboard. For e.g. engineering, dev-ops, sales, and marketing had their own dashboards running. This created data silos and no one had the bird’s eye view of whether the product was heading in the right direction. Yes, dashboard fatigue is real!
4. And, we humans tend to ask questions and validate hypotheses with our colleagues and trust them rather than looking for facts and information ourselves.

Welcome Alexa Voice Assistant

In order to sustain the data-driven culture and cater to all the natural human tendencies, I realized that voice interfaces were the way to go forward. I chose Amazon Alexa to create my own skill and plug them with the APIs which were already created to query the data for the dashboards.

I created an MVP

Within a week, I had an MVP ready which could answer questions like:

• How many users signed up in the last 7 days?
• How many users bought a subscription in the last one month?
• How many users have downloaded and signed up from the mobile app?

I took feedback from different team members and created a list of questions that everyone wanted to be answered on a day-to-day basis. I slowly started connecting APIs from different data sources with this one skill.

I ensured that the MVP served all the product functions

After a couple of weeks of further development, the skill was ready to answer questions like:

• What is the uptime of the product in the last 1 month? — DevOps
• What is the mean response time of our APIs? — Engineering
• How many unique user sessions did we have on our website in the last month? — Marketing
• What is the dollar value of our current sales pipeline? — Sales
• How many users have completed the user funnel? — Product

All of this took roughly 3–4 weeks, given the fact that all of the APIs were available to query the required data. Once the skill was ready, we bought Alexa devices for each of our meeting rooms and common areas.

Benefits of using the voice interface

With Alexa device placed in different locations inside the office, I ensured that:

1. No meeting or conversation was ever paused due to the lack of a data point.
2. Reduce the friction of opening a dashboard and getting data as soon as a question popped up in someone’s head. Essentially, cater to the innate human tendency of asking questions.
3. Get rid of all the different data dashboards and create a unified interface to access all the data irrespective of the product function. (Though, we were not able to completely get rid of the dashboards as there were always complicated queries and analysis that had to be performed which couldn’t be done through a voice interface.)

I continuously improved the skill

Through logs, I closely monitored what were the top questions that the team members were asking and the skill did not have an answer to or wasn’t responding to as per need. I kept adding those questions to the skill to ensure that team members kept coming back to the voice assistant for all their data needs.

Over a period of months, the team not only had access to data but were using it for their day-to-day needs of understanding how our users were using the product. With this arsenal, the team was able to improve the user conversion funnel from 10% to more than 50% over the period of 6 months!

Have any thoughts, comments, or feedback? Feel free to share it below :)