Participation is showing up. Engagement is caring enough to come back. A like, a comment, a completed survey or a new community member is participation: someone did the minimum the interface asked of them. Engagement is what people do when nobody asks and nobody pays. They return on their own, contribute something you didn’t request, stay for years and bring their friends.
Most of what brands and researchers report as engagement is participation with a better job title. “Engagement rate” usually means likes, comments and shares divided by followers, or by reach, depending on who built the slide. Community dashboards count members. Research reports count completes. Web analytics count time on page. All of these show that something happened. None of them tells you whether anyone cared.
The difference matters because the participation numbers are easy to inflate, and the platforms that invented them know it better than anyone. Here is what the evidence says, and how I’d tell the real thing from the imitation.
What does engagement actually mean?
The word has been stretched so far that it now covers a reflexive tap on a phone screen. I’d use a stricter definition: engagement is people giving you something that costs them, without being prompted or paid. Time, ideas, reputation and money all count. A thumb on a screen barely does.
Academic marketing had a more demanding picture long before social media. In a 2001 paper in the Journal of Consumer Research, Albert Muñiz Jr. and Thomas O’Guinn studied fans of the Ford Bronco, the Macintosh and Saab, and defined a brand community by three markers: shared consciousness, rituals and traditions, and a sense of moral responsibility toward other members. Look at what those are. Members recognize each other, keep habits together and help each other. You won’t find any of the three on a social media dashboard.
Now compare that with what a like is worth. Leslie John, Oliver Emrich, Sunil Gupta and Michael Norton ran five experiments and two meta-analyses with more than 14,000 people for a 2017 paper in the Journal of Marketing Research (John et al., 2017). Getting people to like a brand on Facebook did not change their attitudes toward it or what they bought. What predicted both was how fond they were of the brand before they clicked. The like is a symptom of affection that already existed. Collecting more of them doesn’t create more of it.
Why is engagement rate a bad number?
Start with the denominator. Divide interactions by followers and a page with a large, bored audience looks dead. Divide by reach or impressions and the same page produces a different number, often a kinder one. There is no standard formula, so every team gets to pick the one that flatters it, and a rising trend line on the slide is sometimes just a new formula.
Then look at the numerator. When researchers at the Ehrenberg-Bass Institute studied the Facebook pages of big brands, they found that around 0.4% of a brand’s fans interacted with it in a typical week. A brand with a million fans was talking to a few thousand people, most of them presumably customers already.
Even the platforms’ own counters can be wrong. In September 2016 Facebook admitted that for two years it had overstated the average time people spent watching videos, by 60 to 80% according to an estimate from Publicis Media reported by The Wall Street Journal. The calculation had quietly left out every view shorter than three seconds.
Research has its own version. A survey complete is a record that someone reached the last page. It says nothing about whether they read the questions, and in my experience the people who make it to the end of a long questionnaire are disproportionately the ones being paid in points to get there. I’ve written about what that does to the data in You killed market research, and about who is really answering opt-in polls in Movin’ on up.
Time on page is no better. In 2014 Tony Haile, then chief executive of the analytics company Chartbeat, wrote in Time that across 2 billion visits, 55% of visitors spent fewer than 15 seconds actively on a page. Looking at 10,000 articles that had been shared on social media, he found no relationship between how often a piece was shared and how much attention readers gave it. Plenty of people share what they haven’t read. They also leave tabs open while they make dinner.
Goodhart’s law and the engagement number
The British economist Charles Goodhart made the original observation in a 1975 paper on monetary policy: a statistical regularity tends to collapse once someone puts pressure on it for control purposes (Goodhart’s law). The version everyone quotes comes from a 1997 paper by the anthropologist Marilyn Strathern on audits in British universities: “When a measure becomes a target, it ceases to be a good measure.”
Engagement is the textbook case. Once a social media manager’s bonus depends on engagement rate, the cheapest way to raise it is to ask for it. Comment YES if you agree. Tag a friend who needs to hear this. Double-tap if you’re a Monday person. Each of those adds interactions and subtracts meaning, and the dashboard can’t tell the difference.
Platforms fall into the same trap, which should worry anyone who treats their numbers as neutral. Internal documents reported by The Washington Post in 2021 showed that from 2017, Facebook’s ranking system counted each emoji reaction, including “angry,” as worth five likes. The company’s own data scientists later found that angry reactions were much more common on posts carrying misinformation and toxic content. By September 2020 the weight of “angry” had been cut to zero. Reward a reaction and you get more of whatever provokes it.
What is engagement bait, and why do platforms demote it?
Engagement bait is the industry’s name for simply asking. In December 2017 Facebook announced that it would demote posts that ask people for likes, shares, comments and tags to inflate their reach, because users said they disliked that kind of spammy post. It trained a machine learning model on hundreds of thousands of posts to spot them, promised harsher demotions for Pages that did it repeatedly and exempted genuine requests for help, such as missing child reports or fundraising. Over the next two years it extended the rule to more languages, to comments and to video. Engagement bait is still listed in Meta’s Content Distribution Guidelines as content that gets reduced distribution.
YouTube had learned the same lesson earlier. In 2012 it changed how search ranked videos, putting the weight on watch time instead of clicks, to reward engaging videos that keep viewers watching.
Think about what that means. The companies that built these counters, and make their money from them, decided that clicks and reactions were too easy to game to be trusted on their own. Many brands still put the same counters at the top of the monthly report. Bait has the same flaw as ads that treat customers like idiots: it assumes people won’t notice they are being worked.
Why do brand communities become ghost towns?
A community’s member count is participation in its purest form. It measures sign-ups. In 2006 the usability researcher Jakob Nielsen described participation inequality: in most online communities about 90% of users only read, 9% contribute now and then, and 1% produce most of the content. On Wikipedia the split was steeper still, with active contributors making up about 0.2% of its US visitors at the time.
So a branded community of 50,000 members, launched with a sweepstakes and an email blast, is in my experience a few hundred regulars at best, and a good share of the rest came for the prize. The press release announces the 50,000. The moderator knows the few hundred by name.
The most expensive ghost town of all may be Google+. In October 2018, seven years after launch, Google announced it would close the consumer version and admitted that 90% of Google+ user sessions lasted less than five seconds. Nine sessions in ten ended before anyone could have read a post.
Real communities look different. LEGO Ideas asks fans to design a set and then gather 10,000 supporters before the company will even review it. That is months of effort for an uncertain reward, by the designer and by everyone who bothers to sign up and back the project. People do it because they care about the thing itself.
How do you measure real engagement?
You can’t buy it with a prompt, so stop measuring the things a prompt can buy. These are the signals I trust:
- Unprompted contribution: reviews, forum answers, bug reports, fan designs and suggestions nobody asked for. Count what arrives without a campaign behind it.
- Return visits without a reminder: people who come back from a bookmark or a typed address, not from your newsletter or a push notification.
- Retention: the share of people still around after six or twelve months. Cohorts are harder to dress up than totals.
- Recommendation: actual referrals, meaning who came because a friend sent them. A survey question about whether someone might recommend you is a different and much weaker thing.
- Time given when nothing is paid: open answers that run to three paragraphs, customers who volunteer for a follow-up interview, people who turn up to an event without a voucher.
One test sorts most metrics. If the number goes up when you add a prize or a “like this if” prompt, it measures participation. If it goes up only when the product or the conversation gets better, it may be measuring engagement.
My take
Participation isn’t worthless. It’s the doorway, and nobody gets engaged without first showing up. The mistake is reporting the doorway as the house.
I think brands and researchers keep doing it for ordinary reasons. Participation is easy to count, it rises every month if you push it, and nobody gets fired for a number the whole industry uses. Engagement is slower, smaller and less flattering, and it has an awkward requirement: you have to listen. The brands with real engagement answer the people who write to them, change something because of what they heard and say so. That can’t be scheduled in a content calendar.
Renaming participation as engagement is the measurement version of a transformation that changes only the label. If your dashboard says engagement is up 40% and nobody in the company can name a customer who did something they weren’t asked to do, believe the customers.
Questions people ask
What is the difference between participation and engagement?
Participation is showing up: a like, a click, a completed survey. Engagement is what people do when nobody prompts or pays them, such as coming back on their own, contributing something of their own and recommending you to friends.
What is engagement bait?
Meta’s term for posts that ask people to like, comment, share or tag in order to inflate reach. Facebook began demoting such posts in December 2017, and engagement bait is still on Meta’s list of content that gets reduced distribution.
What is Goodhart’s law?
The idea that a number stops being a good measure once people are rewarded for hitting it. The economist Charles Goodhart made the point about monetary policy in 1975. The best-known wording comes from a 1997 paper by the anthropologist Marilyn Strathern: ‘When a measure becomes a target, it ceases to be a good measure.’
How do you measure real engagement?
Count behavior that costs people something and isn’t prompted: contributions nobody asked for, return visits without a reminder, retention over months and real referrals. If a number rises when you add a prize or a ‘like this if’ prompt, it measures participation.