The organization knew from the start it was undertaking a difficult task. To begin with, the subject matter was a high-stress topic (think $$$). Customers often ended up in the community when they were already frazzled and bewildered – and therefore not at their most pleasant. The community was a last resource after customers had been unable to find answers elsewhere or stressed after waiting in the customer support call queue.

The community staff themselves had a number of hurdles. The subject matter was vast and complex, making it challenging for a small team to address and secure other resources to assist with complex questions was also proving difficult.

In addition, team members frequently ran into questions that could not be answered due to company policy, leaving them feeling helpless and backed into a corner by frustrated customers. They felt like there was nothing they could do to make the customer experience more constructive.

Where Did We Start?

In order to determine the overall state of the community, we evaluated sentiment based on a sample of community activity.

At random, we selected 10% of the community topics each month and assigned a sentiment to each. Topics that included welcoming, constructive, or appreciative dialect were rated as positive. For example, discussions that included phrases such as “Thank you so much” or “This really helped”. We also looked for topics that included examples of hostile or unhelpful language such as “You are useless” or “I hate…” Everything else was classified as neutral. Neutral included nonaligned statements such as “I need to know” or “How do I…?”

This evaluation quickly confirmed our initial observations. The community had an unacceptable percentage of negative posts. If we were right that behavior mirroring was to blame, new members of the community were picking up and mimicking the hostile, unconstructive, and unwelcome behavior of existing members. In some cases, it was likely because it was the first thing they saw and it set the tone. In others, new users may have observed that angry topics saw more views and customer engagement.

Regardless of the cause, we knew a community where negative sentiment was 19 times higher than positive sentiment needed to change – and quickly.

The Principles at Play – Behavior Mirroring

What is behavior mirroring and how does this affect sentiment in communities? Do you ever notice how some of us tend to take on the quirks and mannerisms of others – or mimic each others words? This behavior is known as mirroring in which we subconsciously imitate expressions, speech, or actions of others. It is a thoroughly studied psychological concept, so we know that such behavior often manifests as a way to learn vicariously, build rapport and facilitate interactions in unfamiliar environments. Essentially, we are acting as chameleons – blending into our environment to observe, increase our relatability and keep safe. (Chartrand and Bargh, 1999)

Given that this behavior can be found in abundance in all sorts of daily interactions, it comes as no surprise that the “chameleon effect” is observed in online interactions as well. For example, we see mimicking in social networks with the rapid adoption of jargon as participants use hashtags and acronyms. It explains why acronyms such as “bae” (before anyone else), “smh” (shake my head), and #tbt (throwback Thursday) quickly spread and are adopted as common expressions with unprecedented speed online.

Communities also see mirroring, but the effect is more subtle. Have you noticed that once one person starts ranting, it feels like the rest of a group joins in? Or when one person posts a complaint, suddenly a dozen more appear? It is mirroring, and it becomes a feedback loop. The more people exhibit a behavior, the more it is copied.

But what if you could use mirroring to your advantage? That’s exactly what we did to turn this community around.

Plan of Action

In order to turn the sentiment of the unhappy community around, we knew we had to change the behavior of both customers and staff, which was no small task.

We developed a strategy comprised of three sections:

1. Community clean up
2. Guidelines & enforcement
3. Modeling constructive behavior

Each component of the community turnaround was designed to have an immediate impact on sentiment. Cleaning up the community would help prevent negative mirroring. Guidelines would help prevent abuse and empower community staff. Setting an example would begin to show members what behavior was expected of them and provide positive behaviors to mirror.

We also had a long-term strategy in mind. Everything we proposed and enacted would become part of the community maintenance plan, and would be carried out every single year.

Community Clean Up

By the time we were tasked with turning around the unhappy community, the active season for the community had already passed and our goals were entirely focused on making sure the next active season would go better.

Knowing this, we set out to clean up the community.

To start, we set up an “archive” category. This was a place for community staff to move topics and posts that shouldn’t be in the public setting but retained for future reference or metrics.

Guidelines & Enforcement

The structure of the unhappy community was not our only concern. The community staff was growing weary and frustrated with the unwelcome behavior of many users, but they didn’t feel like there was anything they could do about it. We knew we had to give them tools and approaches that allowed them to proactively create a constructive environment.

Our first order of business was to rewrite the community guidelines. Guidelines are often viewed as a secondary task in a community, but research and practice has shown that well formulated guidelines can be the difference between mayhem and harmony. After evaluating months’ worth of bad behavior, moderation flags, and complaints in the unhappy community, we developed a list of specific rules and very clear consequences. The impact was twofold. The rules informed visitors and members what behaviors were expected and those that would not be tolerated. They also provided community staff the justification needed to say “No, you can’t do that here.”

With these new rules, we also developed processes for community staff to follow:

• First- and second-time offenders were informed of their transgressions, but not necessarily banned from the community until their third offense
• Intentional offenders were subject to immediate banning.
• Unpleasant but not clear rule breakers were assigned to a “troll” group for further observation.

In addition, we tackled the “smut” filter or blacklist. New unacceptable words were added to the existing list for automatic removal. (We won’t list them here, but you could probably guess a few.) We also added alerts for words that could be problematic, but were open to manual interpretation such as “I’m having a hell of a time installing

” or “I’ve searched every damn place, but I can’t find…” These alerts let community staff know a post needed to be evaluated and potentially edited depending on the intended sentiment.

Even with these tools, some community staff still hesitated as concerns about moderation versus censoring arose. In order to empower community staff to moderate freely but prevent blatant censoring, we developed three rules to keep in mind:

• Angry posts are allowed as long as they do not break any other rules regarding harassment, unhelpfulness, or swearing. Posts aimed at the company as a whole or a specific product are allowed.
• Even if valid and not breaking any existing rules, angry posts or complaints aimed at a specific user or employee are not allowed. These posts are archived and the poster is informed of how to get their concerns or complaints to the right person or department so that action can be taken.
• If a topic becomes a long threaded rant, users are told that the topic has exhausted itself and will be locked from future contributions. The topic is archived at a later date unless the thread has clear value.

Once we started cleaning up, there were many questions as to why we did not remove all the negativity. The answer was simple: Do you trust a product with only positive feedback? Of course not. Allowing a healthy amount of honest negativity helped customers see that we were making a sincere effort to be trustworthy.

Outcome

By the time the second season arrived, the previously unhappy community had a different feel to it that was far more welcoming, constructive, and interactive. It wasn’t perfect, but that was never the intention. We simply wanted members to feel comfortable asking questions, suggesting improvements, and letting the community staff know what was needed.

More than anything, we wanted new members to understand how to interact with each other and the company. With positive and neutral examples front and center, the “chameleon effect” was evident. For example, it was more common to see “Hope this helped” and “Thank you for your assistance” exchanges throughout the community. Good questions and answers almost always had likes. Members were clearly mirroring the behavior of others, but now it was constructive and resourceful. Positive sentiment rose, and we measured a stunning and significant decrease in negative sentiment.

We also observed some indirect effects on the community. We saw more questions and fewer statements – a welcome change for community staff as they had more opportunities to respond to customers. We also began to notice clear “super users” emerging, which we attributed to the change in sentiment and engagement. These super users provided clear value to the community, and much-needed support for the community team, and we made sure they were appreciated.

By the end of the second season, we had achieved our goal of creating a constructive tone in the community – goodbye unhappy community! The effect of that was a massive increase in member registration and volume far beyond what we believed likely.

Today, the community is thriving.

It continues to maintain a primarily neutral sentiment with rapidly growing registrations and healthy volume. Most impressively, the community staff has observed a maturation, from a community dominated by redundant questions and one-time visitors to one with high search rates and impressive successful search results. In short, what was originally an unfriendly, unhappy community has transformed into a functioning, highly-effective, and healthy community.

Connect with community peers in our free facebook or linkedin groups or in our private peer community, TheCR Network.