Social media crisis management: Be sincere, and verify.

adecco_justthegraph Social Media

Ian Lurie Jul 17 2013

My rule in a social media crisis: Apologize, but verify.

If your brand screws up:

  1. Remind yourself that you do not get to decide what an apology is. The audience does.
  2. Apologize sincerely and promptly.
  3. Start monitoring response sentiment.
  4. If you don’t see a recovery within hours, try again, with feeling.

In real life, failing to apologize makes you look small/lame/insincere. Social media has a longer memory, and a wider reach. Mess up and fail to apologize there and it could take a long, long time to recover your reputation. So get it right, fast.

Adecco was kind enough to demonstrate this theory. They tiptoed to the edge when they stole a blogger’s brand, concept and campaign for their own, trademarked it, and got caught. They lept, screaming with laughter, into the social media meat grinder when they issued statements that said “We’re sorry. Sorta.”

Adecco's apology

Adecco apologizes. Minus the apology part.

What happened

On June 20th 2013, the story broke that Adecco’s new ‘Around the World In 80 Jobs’ campaign was stolen property. Turner, an independent blogger, had been running a site at aroundtheworldin80jobs.com for years. Adecco not only lifted the idea, they also claimed the trademark.

Turner was not amused.

Now, I really doubt Adecco hired an agency by saying “Hey, guys, can you find a hapless blogger, steal their entire brand and use it for our campaign? Thanks!” And I’ll assume no one at Adecco seriously thought stealing the concept was OK. Once they found out, they wanted to make it right.

That means they could have had all of this handled by the end of the day on June 21st. One sincere apology, and a revamp of the campaign, and all would have been well. Instead, the crisis dragged on for five days, and the damage to their reputation continues. What could Adecco have done differently? What could you do if you ever get caught up in a crisis-nado like this?

Apologize, for real

Adecco’s first ‘apology’ reads like a carefully-worded corporate message that avoids admitting liability. What is does not sound like is an apology.

They’re sorry “for some recent negative comments…” Okaaayyyy. This kind of language is a huge problem, because a mob of angry bloggers will jump all over it. In Adecco’s case, they did, and overall Twitter sentiment around the Adecco brand stayed in the negatives:

Even the little lift you see in the graph isn’t genuine – sentiment appears to improve because the initial hail of angry tweets slowed down a bit. There was no positive response to their apology, though.

If they’d been tracking response to their first, half-hearted apology, they could have avoided long-term damage to their brand.

Measure response

Which brings me to the second must-have: Sentiment analysis. This is a relatively new technology that uses language processing to score statements from +1 (really good) to -1 (really bad). If the poop hits the fan, you need to start tracking the sentiment of any tweets, Facebook posts or other narrative beyond your control. Then, as you issue statements or apologies, measure response. If Adecco had done that, they might have known their initial ‘apology’ fell flat before five days passed.

Measuring sentiment is like looking someone in the eye and asking “Are we ok? What else can I do to fix this?” It’s your way to make sure you’ve patched things up.

Remember who’s in charge

When a crisis like this happens, and you’re in the wrong, the mob is in charge of your brand. Period. You have no options until you apologize, and they accept your apology. Then you’re back in charge.

Ignore this principle, go silent and sulk, and you end up making the damage to your brand more severe and longer-lasting. June 26th, Adecco reached a settlement with Turner. The result didn’t help brand sentiment, though:

Because at that point, an angry mob of bloggers (including me!) had been in charge of their brand for days. We increased tweet volume for the Adecco Brand 20x, and all of that increased volume was negative. It’s hard to get over that.

Adecco: Still recovering, weeks later

Sentiment started rising again around July 12, 2013. But it’s not a real recovery, for two reasons:

  1. They’re far more vulnerable to new crises. July 10 and 11, news broke that Adecco is part of an antitrust probe in France. That drove sentiment down, briefly, but shows how quickly a single news headline can affect their brand.
  2. Sentiment appears to be rising now because there are a lot fewer people talking about Adecco. Twitter posts mentioning Adecco fell 90%. Job opening announcements have posted to Twitter at a constant rate for months. With other chatter dying down, it looks like sentiment is improving. Is it, really?

Lack of visibility/discussion isn’t the same as good reputation. Who knows when Adecco will get an opportunity to rebuild their image among the social media geeks. They’ll need to, though: Adecco is a staffing agency. I imagine a lot of their potential employees (and employers) are on Twitter and Facebook.

Lessons learned

  1. If you’re a brand – an individual famous within a niche, or an organization – and you make a mistake, own it. Apologize. Do it clearly. Don’t attempt to divert blame, avoid admitting you did something wrong, or otherwise sound insincere.
  2. The audience, which is now full of angry people, must accept the apology. Only that will defuse the crisis and pull you out of the reputation nose dive.
  3. The good news: In a crisis, a little sentiment analysis goes a long way. You get lots of second chances. Track audience response, analyze it, and then adjust your message accordingly. That will prevent a scratch on your reputation from becoming a scar.

Closing note: How we measured it

Measuring sentiment around a brand is a lot easier than it used to be. Here’s how we did it:

  1. We archived all tweets mentioning Adecco from June 20th to July 15th, 2013.
  2. We measured the sentiment of each tweet using the Python Natural Language Toolkit.
  3. Finally, we hand-checked the sentiment data, because computers are notoriously bad at understanding when “bad” is bad versus good.

For tweets before June 20th, we used Twitter.com and Topsy.

Then we exported the result to Excel, ran a pivot table to get averages by day, and turned it into this narrative/graphic/graph:

Embed it:

Portent

3 Comments

  1. Wow Ian, great post.

    I will really like to know more about how you used Python NLTK to process the content end extract information about sentiment (I’m a pythonista too).

    • That’s my next blog post, I think.

  2. Slip-ups do happen! The companies who make the quickest recoveries are those who quickly address the issue and take full responsibility for their mistake.

Comments are closed.