43
Fueling the Data Drive: Facebook & Twitter Data Mining Solutions for Digital Communicators Adrian J. Ebsary @AJEbsary

Twitter and Facebook Data Mining Solutions

Embed Size (px)

Citation preview

Page 1: Twitter and Facebook Data Mining Solutions

Fueling the Data Drive: Facebook & Twitter Data Mining Solutions for Digital Communicators

Adrian J. Ebsary

@AJEbsary

Page 2: Twitter and Facebook Data Mining Solutions

The Content Marketing Process

Page 3: Twitter and Facebook Data Mining Solutions

Starting your archives

• Never trust your data to a third party

• Own your own Excel (.CSV) files

• Don’t settle for time-slice dashboards

Page 4: Twitter and Facebook Data Mining Solutions

Starting your archives

• Keep regular export files on recurring basis (monthly)

• If possible, segregate data by keyword/keyword families

• Master archives can get bulky• May need to separate over time periods (yearly)

• Might be errors in combining files or deletion of duplicates

• If size is an issue, keep daily over weekly/monthly data

Page 5: Twitter and Facebook Data Mining Solutions

Excel with Excel

• Plenty of free resources to learn• Google: “learn how to use Excel”

Page 6: Twitter and Facebook Data Mining Solutions

Excel with Excel

• Plenty of free resources to learn• Google: “learn how to use Excel”

• Understand IF statements

• FIND, SEARCH, MID, LEFT, RIGHT, LEN• Find text by characters, position or string length

• CONCATENATE• Paste text together

• Excel time vs. Unix Time• Unix time easier to use for mathematical operations

• Unix: Number of seconds since 1st of January, 1970

Page 7: Twitter and Facebook Data Mining Solutions

Data by network by ease-of-access

• Facebook• No detailed data on user-generated content

• Powerful generic interaction and reach analytics

• Completely free, from the source (Facebook Insights)

• Third-party offerings (usually) offer little more

Page 8: Twitter and Facebook Data Mining Solutions

Data by network by ease-of-access

• Google+• Integrate Google+ Pages into Google Analytics

Page 9: Twitter and Facebook Data Mining Solutions

Data by network by ease-of-access

• Twitter (Tweets by Keyword)• Build your own app or…

Page 10: Twitter and Facebook Data Mining Solutions

Data by network by ease-of-access

• Money, money, money• dev.twitter.com/programs/twitter-certified-products

• More options, more networks, more $$$

Page 11: Twitter and Facebook Data Mining Solutions

Data by network by ease-of-access

• Twitter (Account interaction data)• Want Twitter data? Buy a $10 ad.

Page 12: Twitter and Facebook Data Mining Solutions

Data by network by ease-of-access

• Klout• Regular algorithm changes = little value beyond bragging rights

• Multiple, regular scoring changes confuse scoring

• Lack of transparency surrounding algorithm

• One number to rule them all?• Multiple social networks simplified to single logarithmic scale

• Benchmarking a moving target

Page 13: Twitter and Facebook Data Mining Solutions

Facebook Data

Page 14: Twitter and Facebook Data Mining Solutions

Facebook: NewsFeed Algorithm

Page 15: Twitter and Facebook Data Mining Solutions

Facebook: NewsFeed Algorithm

• Negative Feedback kills post reach (anti-weight)• Hide post

• Unfollow page/person

• “I don’t want to see this”

• Report as spam

Page 16: Twitter and Facebook Data Mining Solutions

Facebook: NewsFeed Algorithm

• The vanishing ‘Virality’ score• Efficiency of attention consumption > overall reach

• Virality: Total engagements / Total reach x 100%• Engagements = Comments + Likes + Clicks

• Every pageload that results in no engagement = lost affinity• Boring content cuts your future reach potential

Page 17: Twitter and Facebook Data Mining Solutions

Facebook Insights Dashboard: Like Spikes

• Watch for like spikes and identify source• Correlate with posts or events for additional insights

Page 18: Twitter and Facebook Data Mining Solutions

Facebook Insights Dashboard:Like Spikes

• Watch for like spikes and identify source• Correlate with posts or events for additional insights

Page 19: Twitter and Facebook Data Mining Solutions

Facebook Insights Dashboard:Unlikeable days

• Separating posts by at least one day clarifies unlike spikes• Anecdotally, unlikes often correlate with high like spikes

• Low relative number of daily likes with high unlikes indicates highly ineffective content

Page 20: Twitter and Facebook Data Mining Solutions

Facebook Page-Specific Data: Virality

• Virality = Engaged users / Total Reach * 100%• More accurate: Total ORGANIC Reach

• Paid reach has less impact on affinity score

Page 21: Twitter and Facebook Data Mining Solutions

Facebook Page-Specific Data: Virality

• Virality = Engaged users / Total Reach * 100%• More accurate: Total ORGANIC Reach

• Paid reach has less impact on affinity score

• Consumers vs. Engaged users?• http://www.jonloomer.com/2013/03/11/facebook-insights-consumer-vs-engaged-

user/

• Consumer = interacted with your posts

• Engaged user = interacted with your posts OR your page

• Best: Daily Page Consumptions / Daily Organic Reach x 100%

Page 22: Twitter and Facebook Data Mining Solutions

Facebook Page-Specific Data: Unlikeable days

• Negative feedback• Hidden from dashboard – need to download!

• Identify affinity-killing days to find posts in need of improving

• Look for high unlikes + negative feedback on single day

Page 23: Twitter and Facebook Data Mining Solutions

Facebook Page-Specific Data: Unlikeable days

• Negative feedback• Account for reach

• http://simplymeasured.com/blog/2013/05/30/negative-feedback-on-facebook-what-is-it-and-when-you-should-worry/

Page 24: Twitter and Facebook Data Mining Solutions

Facebook Post-Specific Data

• Use to complement page-level analysis

• Easier to visualize impact using daily metrics from page-level data• Graphing with posts as x-axis may obscure smaller data points, hide multi-day effects

• Best: tag posts with the date page-leve data, date as x-axis

Page 25: Twitter and Facebook Data Mining Solutions

Twitter Data: Give ‘em your card

• Use $10 to buy ad, get permanent access to analytics.Twitter.com• 30 day window on follows, unfollows, mentions

Page 26: Twitter and Facebook Data Mining Solutions

Twitter Data: Give ‘em your card

• Use $10 to buy ad, get permanent access to analytics.Twitter.com• 30 day window on follows, unfollows, mentions

• 90 days of your tweets (or 500 tweets) .CSV download• ID

• Time sent

• Faves

• Retweets

• Replies

• Text

Page 27: Twitter and Facebook Data Mining Solutions

Twitter Data: Give ‘em your card

• Use $10 to buy ad, get permanent access to analytics.Twitter.com• 30 day window on follows, unfollows, mentions

• 90 days of your tweets (or 500 tweets) .CSV download

• “Request your archive”• Detailed data on all the tweets sent from your account

• No engagement insights, only text, time, etc.

Page 28: Twitter and Facebook Data Mining Solutions

Twitter Data: Followers, following?

• Twitsprout.com! Free for three Twitter accounts• Data begins from sign-up date:

• Total tweets sent

• # of followers

• # following

• Export as .CSV!

Page 29: Twitter and Facebook Data Mining Solutions

Twitter Data: Keyword-based collection

• Hootsuite Archives• Best deal: 10,000 total tweet archive

• Can delete archives and restart at any point• (effectively limitless for low-medium rate keywords)

• Cons: Needs some massaging to work in Excel

• Contains data on:• Tweet text

• Sending username

• Time sent

• Language

• Tweet ID (link to tweet)

Page 30: Twitter and Facebook Data Mining Solutions

Twitter Data: Keyword-based collection

• Hootsuite Archives• Step 1: Export & convert to Google spreadsheet format

Page 31: Twitter and Facebook Data Mining Solutions

Twitter Data: Keyword-based collection

• Hootsuite Archives• Step 1: Export & convert to Google spreadsheet format

• Step 2: Download as an Excel file (or .CSV)

Page 32: Twitter and Facebook Data Mining Solutions

Twitter Data: Keyword-based collection

• Hootsuite Archives: Basic Manipulations• Rebuilding the link to a single tweet

• =CONCATENATE("http://twitter.com/",C2,"/status/",D2)

• C column = “from user”

• D column = “id”

Page 33: Twitter and Facebook Data Mining Solutions

Twitter Data: Keyword-based collection

• Hootsuite Archives: Basic Manipulations• Building a master archive with overlapping keywords

• Step 1: Concatenate the text and the sending user

• Prevents loss of multiple RTs

• =CONCATENATE("@",C2,":"," ",A2)

• Produces: Username: Tweet text

• Step 2: Use Excel’s ‘Remove Duplicates’ Function

Page 34: Twitter and Facebook Data Mining Solutions

Twitter Data: Keyword-based collection

• Hootsuite Archives: Basic Manipulations• Given time in two formats: text & Unix time

• Unix time: Seconds since Jan 1st, 1970, excluding leap seconds (easier to use for math)

• Convert Unix to Excel time (Serial date) using this formula• Column M contains Unix time

• =(M2/86400)+25569+(-5/24)

• Will look like: 41353.0116

• Format Cells for ‘Date’, = March 20, 2013

Page 35: Twitter and Facebook Data Mining Solutions

Twitter Data: Keyword-based collection

• Hootsuite Archives: Basic Manipulations• Counting number of tweets per day

• Assume tweet dates in serial format are Column M

• Create column with desired date range in serial format (N)

• =COUNTIFS(M:M, ">" & N2, M:M, "<" & N3)

Page 36: Twitter and Facebook Data Mining Solutions

Twitter Visualization: Netlytic.org

• Free software created by an academic lab• Creates visualizations in Gephi with no coding knowledge needed

Page 37: Twitter and Facebook Data Mining Solutions

Twitter Visualization: Netlytic.org

• Convert Excel file back to .CSV & upload to Netlytic.org• Will not need to ‘Clean data’

Page 38: Twitter and Facebook Data Mining Solutions

Twitter Visualization: Netlytic.org

• Select field containing tweet text only• Do not use concatenated username + tweet text

Page 39: Twitter and Facebook Data Mining Solutions

Twitter Visualization: Netlytic.org

• Get rid of keywords you do not want for text analysis

Page 40: Twitter and Facebook Data Mining Solutions

Twitter Visualization: Netlytic.org

• Keywords visualized by usage over time

Page 41: Twitter and Facebook Data Mining Solutions

Twitter Visualization: Netlytic.org

• Proceed to mention network analysis• Ignore chain network analysis

Page 42: Twitter and Facebook Data Mining Solutions

Twitter Visualization: Netlytic.org

• Also interactive visualization for more styling

Page 43: Twitter and Facebook Data Mining Solutions

AdrianEbsary.com