What we learned from our 3 most popular election features
They helped us break the record for the most visitors in one day.
The day after the 2017 local Danish election turned out to be the biggest in our history at tv2.dk with about 1.3 million visitors. That’s about a third more than on a regular well visited day (Denmark only has about 5.5 million inhabitants).
As the digital projects manager I was in charge of planning and executing the election that taught os something profound.
I am presenting our biggest learning from that election along with the three most popular features at this year’s ONA conference in Austin, TX. This is basically a short version of the presentation.
Bonus: I’ll also cover two digital features that didn’t do as well as planned.
The 3rd most popular feature: The mayoral overview
Everything was more os less settled a week before the election. All the biggest features were being tested, the battle plan had been made and everybody was ready for election night.
But then a reporter asked a question:
“How are we going to give the users an overview of who will be the next mayor?”
A simple question we hadn’t really thought about or discussed at all at the many, many meetings leading up to the election even though the fight for the golden mayor chains is bloodier than House of Cards.
We had less than a week until election night and all the developers at the CMS team were busy building the election results page. Luckily the Editorial Development Team was able to come up with a simple solution.
Team lead Anders Bergmann combined two simple tools on a blank page to create this continuously updated overview of the new mayors filling the 98 seats all around Denmark.
Datawrapper was used to give a quick view of how well the many parties had done compared to the last election four years earlier.
Our in-house “Mosaic Tool” was used to do the mosaic of the many faces, where the users could press a face and expand to read a bit more about the new mayors.
The reporters had to 1) update the Datawrapper visualisation, 2) fetch the right photo from our database with photos almost all the candidates, 3) run it through a simple photoshop template before 3) uploading it to the Mosaic Tool and 4) then finally updating the page to load the new info.
A hacky and laborious workflow but it worked. The readers kept coming back over the next few days and spent a considerable amount of time on the page.
The 2nd most popular feature: The results page
Just about every news site in Denmark had some sort of live-updated results page showing the distribution of the votes in the different municipalities. But ours was the only one that actually worked on mobile.
It seems banal to say, but the competitor’s results pages were optimised for desktop viewing or simply broke down under the extreme load on election night.
Election result pages are notoriously hard to do (people are still talking about what happened at USA Today many years ago). When you combine live updating data with massive spikes in traffic you have a recipe for disaster and a good reason why something like this wasn’t handled by our small Editorial Development Team but expertly done by the big development department.
The results page was simple, easy to use and had rock solid performance. But this time we added a small stroke of genius that I believe made the users stay longer on the page.
To show the users that the page was being live-updated with the latest numbers we added a small animated countdown, when the numbers were still coming in.
The most popular feature: The candidate test
TV 2 was one of the first Danish media companies to build what is now a staple feature in all Nordic elections. Finnish YLE did the first candidate test sometime in the 90's.
The premise is simple: Rate a bunch of political statements on how much you agree with them and we will tell you which candidates in your area that you agree with the most.
Megafon did the necessary polling, wrote the algorithm and helped os formulate the questions that Paqle distributed to the more than the 10,000 eligible candidates.
But our version, that the CMS and API/Server teams made with our regional affiliates, differs a bit from the others’:
- It is more personal. Firstly, the questions are based on what a representative portion of the Danish population has told us matters the most when they are voting. Secondly, the users can choose what subjects matters the most to them personally and get more questions on those.
- It is more local. We added more than 300 local questions to the test. The questions were done by our regional affiliates’ political editors and reporters that know what matters in their area.
It made a difference with the readers. They were especially impressed with our local questions that made the test much more relevant to them.
The test was taken almost 2 million times, peaking on election day.
The Key Learning
I believe that the three digital features above contributed to the record-breaking succes of the election coverage because of one thing:
The three digital features all addressed really simple questions that most voters had.
- Who should I vote for? The candidate test alone could not give you the answer but it was an easy and fun way of discovering new candidates that you might not have known of.
- Who is winning? The results page was a simple and easy place to find the latest election results no matter where you where. Even on your mobile.
- Who will be my mayor? The continuously updated mayor overview gave you a quick answer to that question and reason to check back later if the politicians were still negotiating.
Also two things that didn’t become the hits we had hoped for
The Editorial Development team also built two more digital features for the election that had decent numbers but never took off as much as we had hoped for.
Community Radar
Inspired by a Finish project and with the help of a bunch of experts, we collected 31 datasets from all the 98 municipalities and rated them in four categories. The result was impressive but the stories not as surprising as we had hoped for. It might have bordered on data porn.
Three things that can affect your election
Again we worked with the local political reporters and editors of our affiliates to point out three things that might affect the election in their municipality. Unfortunately the reader interest was limited. But it turned out it was a very helpful tool for the national repprofessionorters to get an idea about what was going on around the country.
Many things could have been improved, but with that said, I believe both the features were trying to answer smaller or less defined questions than the three succesful features did.
They also showed what is possible when developers and journalist join forces. I’ll be studying the relationship between the two very different professions this year at MIT and Harvard.
