Skip to content

Instantly share code, notes, and snippets.

Show Gist options
  • Select an option

  • Save vmbrasseur/3db7b69c0a9e8a270382fac23c9fcf22 to your computer and use it in GitHub Desktop.

Select an option

Save vmbrasseur/3db7b69c0a9e8a270382fac23c9fcf22 to your computer and use it in GitHub Desktop.

Revisions

  1. vmbrasseur created this gist Apr 7, 2017.
    154 changes: 154 additions & 0 deletions gdc17-data_driven_community_management.md
    Original file line number Diff line number Diff line change
    @@ -0,0 +1,154 @@
    Community Summit

    Data-Driven Community Management

    * Data-driven CM bores people
    * So let's talk about fire
    * Post-fire, they try to figure out why it happened
    * Can they improve anything?
    * Was it a crime?
    * If something can be improved, is that actually done?
    * Codified by Law Enforcement Assistance Administration
    * How to spot arson
    * Started seeing arson everywhere
    * Experimented: actually arson?
    * "Where does the fire begin?"
    * If fire begins where it normally doesn't, it's probably arson
    * But < 6% of the investigator could locate where the fire began
    * So much of what happens in fire investigation was unscientific
    * Started a forensic approach, very scientific, # of arsons reported dropped
    * How much of CM are things that work & how much is cargo culting?
    * We have not even begin to figure out the potential of our data
    * If you collect the data, does it affect what you do in your community?
    * Find CMs are great at measuring things but not at determining insights & taking action based on it
    * We could be doing so much more here
    * We collect as much data as possible, but don't use it to change what we're doing
    * Measurement isn't as important as how we're going to take the insights we have & improve the outcomes for our companies & communities
    * Good at
    * Seeing what's popular & doing more of that for their community
    * Most community engagement is more about the people using the game than the game itself
    * De-lurking the lurkers
    * How do get them participating
    * Study from IBM: "lurker to leader"
    * Used to believe this wasn't possible
    * Now finding there are systematic ways to do this
    * Avoid novelty ideas
    * "Lurker week"
    * Small bump in participation & no long-term effects
    * Interviewed lurkers
    * Have nothing to offer
    * Have no experience
    * Have no time
    * Don't want friends/boss/family to see it
    * Started making asking questions the most valuable thing people can do in a community
    * Started looking for questions rather than answers
    * Lurkers started participating a lot more, and for the long term
    * STRESS THE VALUE OF QUESTIONS & make it safe for people to ask them
    * How to get newcomers to participate actively
    * Lots of visitors & bouncers (1 in 1000 will become active)
    * How to fix this
    * Don't just drop people in: Laser-focused in the action
    * Don't just leave it up to them or to chance
    * Provide direction/funnel
    * Cohort/analysis tool in Google Analytics
    * Don't just optimise for the first week of participation
    * Easy to get them to participate once; hard to make them participate in the long term
    * Long-term engagement involves psychology
    * Social-identity theory
    * The more we feel a group is successful, the more we participate
    * Provide history, top members of the community, what it takes to participate
    * Context helps people feel engaged & stick around
    * Add some friction to the entry process
    * ask people what they're interested in & provide them interesting content at the very start
    * Automation rules
    * After you join, you get a message from the community
    * After 1 day, 1 week, 1 participation, etc
    * Use this to help bolster social identity
    * In technical communities, encourage newcomers to make their requests as possible
    * Increases newcomer participation due to increased responses to requests
    * Even one response increases chance of participation again to 45% from 16%
    * They find that top 20 discussions in a community consistently end up with 50% of the page views
    * SEO to make discussions easier to find
    * Merge the good discussions into one discussion (if relevant to each other)
    * Can promote them elsewhere (Quora, etc)
    * Spend as much time improving your best discussions as creating new ones
    * Very good way to get people engaged & actively participating
    * Big announcements…often fall flat
    * Hard to get people to engage with them
    * Sticky posts are where announcements go to die & emails don't get read
    * How do you get people to open & read any type of message?
    * People don't progress through communities like we think they do
    * For instance, they often don't start on the home page
    * Figure out where people enter, then how to make the most of those entry points to promote announcements
    * Side-bars on blogs, etc.
    * Go to where the audience actually is, don't assume where they'll be
    * Sense of community
    * Building this
    * People who believe & identify as a member of the community
    * We assume that this just happens & is emergent
    * Yes, it's important that it happens, but we can help it along
    * It can happen, but it often leads to participation inequality
    * Core group & others don't feel they can break in
    * Virtual Sense of Community survey (can find this on the net)
    * Will tell you what you need to do to improve the sense of community
    * Composition of sense of community
    * Membership: people can identify each other
    * Rituals w/in the group
    * Anything which makes people feel a connection with each other
    * Unique shared experiences
    * Stronger sense of community from sharing these
    * Common symbol systems
    * Forum system should reflect how the community speaks & engages
    * Don't force a structure on them which doesn't reflect how they think of things
    * Use the same language as your audience
    * Increase the sense of influence
    * Those who participate are those who feel they can influence it
    * How can we increase this for everyone?
    * Give people an opportunity to show off (let them feel smart, clever, good)
    * Write about what people in the community are doing
    * Create a sense of success
    * How do you make the group feel successful?
    * Most communities don't have a clear common goal
    * Adding one bonds people more in the group
    * Having a shared emotional connection
    * Strong communities react the same ways to the same things
    * How can we encourage this?
    * Create opportunities for people to feel a specific way at a specific time
    * No practical use for sharing information, but bonds people into the group
    * So off-topic can be very good
    * Improve member satisfaction
    * Video gamers can get angrier than anyone he's ever seen
    * How do you make people feel more satisfied instead?
    * Looked at how to measure this
    * Responses from other community members are rated better than those from community managers
    * Turned out to be not that it was the community manager but that the community members were answer the questions quickly
    * Quick answers got a higher satisfaction rating
    * LIFO/FIFO in response methods
    * Generally found that LIFO got higher satisfaction rating
    * Response structure?
    * Direct answers get a higher rather than FAQ links
    * Even if the answer isn't as good as a FAQ or link to better answer
    * Wording?
    * Answers using validation, empathy, guidance get much higher satisfaction rating
    * Increase long term participation & happier community members
    * Are there questions which do better in the long term?
    * Yes. Questions beginning with "How" do much better in the long term
    * Take the time to develop these questions
    * What's the problem with increasing engagement?
    * It doesn't always equal more value
    * If you generate more engagement but it's not clear what the value is…then you'll get cut
    * Must connect engagement with value
    * FeverBee has a page "Classifying the Benefits of Community"
    * Very, very few communities are able to do this
    * Studying it is very difficult
    * Can figure out what metrics you need but must do it in a systematic way
    * Coursera course on introduction to probability & data
    * Outsource it to a data scientist
    * Explain the question, hand them a bunch of data, they can get you answers you wouldn't find
    * Hire someone else to do it
    * PhD grad students are great for this
    * Create a custom dashboard for the metrics & outcomes you specifically need for your community

    * Q&A

    @richmillington