Web Notifications, Part 5: Live Polling

Bringing users into the conversation with real time questions and answers.

Photo by Daniel Sorabji for AFP/Getty Images

The live polling notification format is also our most recent which we debuted during the EU referendum experiment. In that event, live polling offered users a way to respond to a question in a notification.

We sent three polls during the course of the EU referendum, the first time as a technical test, the second to gauge user opinion, and the third time to provide a way to close out the experiment after a surprising outcome.

The flow of each poll varied depending on what information we were trying to gather. Our first test was simple, with two options: choose A or choose B.

We sent this to users just as polls were closing in the UK

After users voted, we sent back the following immediately:

We sent this “Hang on” as a stop-gap measure, in part because we didn’t know how quickly votes would come in, and we didn’t want a user to see an unrepresentative result hanging out on their screen for a long time. We let potential voters have 10 minutes to vote before returning results.

The body copy of the results notification was pulled from the live blog. From this first test we learned that within 10 minutes we could get close to 1500 voters — so in short, it worked. We adapted the format for our next test.

Shortly after the polls closed in the UK we sent our next live polling notification, which combined interactive elements. First, we wanted to reach a select audience of those who had voted. We had considered sending this poll only to those in the UK, but considering the Guardian’s audience, which contains many expats who may have voted absentee, we wanted to allow anyone signed up for the full experiment to try it.

The first notification asked users if they had voted and attempted to teach users how to use the action buttons to vote.

Based on the answer returned, the user then received one of two responses.

If a user said yes, they received this notification:

Tapping the left action button recorded a vote for the economy, and tapping on the right action button began a “carousel” of options to reveal the other issues highlighted in the notification body. For example, if the user clicked “No, Next issue” they would be shown “Immigration” in the left action button, then “Security” and so on, repeating from the beginning after a user tapped through all of the options.

In the notification copy right under the title, we told users they had until 10:30pm BST to vote. After logging a vote, they then immediately received a live updating result of which issue was leading.

Automatic updates were sent intermittently, as 50, 100, 200 etc. votes came in, before the final result was sent.

If the user responded “no” to the first question “Did you vote?” in our poll, they were not asked to vote on which issue mattered most, but received updates as results came in until the final result was pushed out half an hour later.

The following morning New York time, we sent our final poll. Initially we had not planned to end the EU referendum experiment with a poll, but rather with a set of quotes from world leaders reacting to the result. However, with the unexpected “Leave” result, events were still moving quickly and we had difficulty making our initial concept relevant.

Instead, we crafted a final poll to follow up on the one we had sent as polls closed. We asked users what, given the results, was now their primary issue of concern.

As before, we used the carousel in the action buttons to navigate between options and gave them until a set time to vote.

At the end of that time, we called an end to the poll and delivered a final result.

Live polling is a great way to get real-time reactions from users. The format could be used to generate a sense of what users want to see more of, add color to entertainment coverage, or, as we tried to use them during the EU referendum, give users a place to express an opinion and see how they relate to their peers.

It’s still early, but here’s what we’ve learned from live polling notifications:

  • Users will vote immediately. This may seem obvious, but it was a real consideration. We weren’t sure how long users would let the notification sit on the lockscreen and thus were not positive that starting an automatically returning response was possible or advisable. From the three times we have run polls, we’ve seen that the vast majority of the voting comes in within the first 5 minutes of voting and lags after. If you try them, key the rate at which you send updates to how many votes have come in, in order to reflect the dynamism of these early minutes.
  • There are a limited number of useable lines in the notification body before it cuts off. Similar to the character limit, a notification can only be 9 lines long before the text gets cut off, even while expanded. For others looking to run polls, this means you need to consider the number of answers you offer.
  • Users seemed to understand the UI of how to operate the carousel of options. From the data returned on the issues poll, we can see that the fourth answer in the carousel was the second most popular, suggesting that users understood how to move through the options.
  • Randomize the options to reduce bias in the data. Perhaps in contrast to the previous point above, the first option was the most popular, suggesting there were a fair number of users who accidentally tapped on the first button or did not understand the UI. When we next run a poll, we’ll try randomizing the buttons to further test out this assumption, and probably work to further refine the UI.
  • Users were interested in polls, but did not necessarily rate them high in value. In the survey we sent after the EU referendum experiment, users marked the polls as the format they would be least likely to sign up for again. However, it may be likely that users are less eager to sign up for a notification format (a poll) when it isn’t associated with a topic they’re interested in. It’s early, and as we further develop our poll copy and formatting we need to keep testing polls in various ways to see if users find them relevant and compelling opportunities for interaction.

More in this series about web notifications: