When the only product of a news organisation was print, engagement with readers was measured by how many people bought the newspaper or the amount of letters, calls or e-mails received in the newsroom.
In contrast, the many metrics available online can serve as an indicator of how people interact with a story and even the quality of the piece, and they can be used by journalists to inform their editorial decisions.
"Any metrics we let into the newsroom are important because we can do something about them.
"We can add subheadings, pictures or video and make sure we're selling a piece accurately, but those things don't necessarily have anything to do with the quality of the journalism," said Chris Moran, audience editor at the Guardian, speaking on a panel about the use of metrics in the newsroom at Sciences Po's NPDJ conference in Paris last Monday.
The Guardian's own analytics platform, Ophan, has a section that encourages journalists to reduce the amount of stories they post, rather than increase it only for the sake of traffic.
"It tells you how many different pieces you can share on a single topic during one day," Moran explained.
While page views and unique visitors are still important metrics, the word currently dominating newsroom conversations is 'impact' as a measure of quality, he said, and its definition can be relative. "It's a word that nobody can quite agree on, but that everyone wants on their side."
The Guardian is "data-informed, not data-led", said Moran, and the editorial team uses analytics to "understand the kind of journalism people might particularly want on different platforms".
For example on Facebook, the second biggest traffic referral at the outlet after Google, "you might need something simpler that hooks [readers] in and then leads them deeper into the story".
One of the questions Moran is frequently asked by Guardian editors is "when is the best time to launch a piece?", but there is no universal formula that can be applied.
The outlet's peak time, he said, is usually lunchtime, between noon and 2pm, but that does not mean all stories should be published at that time.
"There is no point in publishing five pieces of content about the same topic at once, because you simply cannot be top of Google Search on the same term at the same time.
"So my advice to journalists is, if you feel that you should launch it at this time, do it, see what happens, but make sure you learn from it next time."
Renée Kaplan, head of audience engagement at the Financial Times and also on the panel, said the FT is also data-informed, although in a different way – the outlet operates a paywall, which already requires some information about the reader at the sign-up stage.
The FT is currently developing Lantern, an analytics dashboard that will enable journalists to track how their stories are doing both in real-time and in the long term, drawing on data available across the organisation.
"The definition of success is different according to a story's category, its audience, the news agenda on the day and whether the piece works as multiplatform or not," she explained.
The FT relies on journalists' editorial instinct to commission pieces and not on analytics, but this data is used to optimise stories later on.
"You don't necessarily want people on certain platforms to interact with stories in the same ways, maybe you want them to share an article, like it or click it," Kaplan said.
"At the FT, we try to bring people back to the website through these platforms, by exposing them to enough of our content so that one day, they will want to pay for it."
"We used to think of social media as a promotional vehicle, but that is beginning to shift, and we're starting to treat it not as a distribution platform, but as a publishing platform."