• Peerindex is ‘taxonomy’ based scoring – it provides details of the things about which a user has relevance as well as the score
• Klout provides scoring to 2 decimal places; Peerindex provides it to the nearest integer only (no decimal places)
• Klout now provides information about “TOPICS”
• I am not a developer, so accept the rudimentary nature of this study.  The Paessler monitoring software made this task very easy; but not scientific!
• I have noticed occasional errors from the API’s in both cases; both suddenly reporting scores may off the median mark. Does this signal danger for a contact centre routing strategy? Could the returned score be wrong? How often does the API return bad median scores?
• The average response time for both API’s is about 2.8 seconds from request to response.  Does this benchmark remain true over a month long study?

Methodology

I tend to tweet 6 to 8 times a day of my own volition. I also have several automated Twitter, Facebook and LinkedIn feeds and Google News RSS feeds on my chosen subjects.  I always tweet about my core subjects and maintain consistency, except for location based social-media updates based on Foursquare.  The automated updates to my social-media accounts will continue for a period X, at which point I will turn them off.  What affect will this have?  I also tweet regularly with links to blog content such as this study or any number of specialist themes in customer services, contact center solutions, customer experience and customer effort.  I will attempt to rally several bursts of increased and very intense activity to monitor underlying change in scoring.

• Peerindex reported a consistent score via its API of 27 from 02:55am (GMT) on to 07:25am (GMT) on the 6th December 2011. For this 4.5 hour period the API remained constant at 27. Is this a hint that one of the connected ‘measured’ networks was offline and the PeerIndex API built scores based only on the remaining monitored social network activity?
• After 20 days of running the API I see more or less nightly drops to a baseline score of 27 – but it quickly recovers to 55 typically within the hour

Average API response times

• Klout API – 3.216 seconds per lookup
• PeerIndex API – 3.122 seconds per lookup

Raw Result Set

If anyone would like a copy of the raw study data in either HTML or PDF format then please let me know @dave_t_pilot

Summary

I am pleased by the high level of availability of both API’s. Klout appears to be performing marginally better, and I also like the fact that I get scores to 2 decimal places.  If I was a user particularly looking to track my social network activity against an ‘improvement’ in score I’d be more likely to use Klout – because it gives me more or less real-time feedback on the nature of activities that are assessed.

Next Study Phase

The next phase of the analysis will be to manipulate my on line behaviour to see how quickly the results appear to change in accordance with the activity. No conclusions against the study can be drawn against his first 15 day run-cycle.  It is merely to establish what the baseline is against which manipulation of on line behaviour may be assessed.

Social Activity Increase Test – First Milestone Trigger

I have been carefully increasing social media activity on Facebook, LinkedIn and Twitter over the last 4 or 5 days.  I have also used some automation tools to feed posts from the Blog directly to the social streams at regular but not heavy intervals.  I have noted a 10% increase in my @Peerindex score from its flat baseline of 48 to 53.  Should I now expect to see an appropriate increase in the @Klout score? I expect so.  I will maintain a watch on the scoring to see if the same or similar increase occurs.

Results of this phase, Dec 18th

Despite increasing remarkably the rate of posting on Twitter as primary channel, Facebook as secondary and LinkedIn as 3rd the only difference occurred in the PeerIndex score which jumped from 49 to 55 more or less overnight.  Changes in Klout only occured at 0.0X digit – in other words; remarkably changing online behaviour made no difference to the reported Klout score.  Even heavy engagement did not appear to make any difference to the returned klout score.  Periods of heavy engagement seem to be rewarded with a klout improvement of hundredths of a percentage point.

Therefore, I see no result change in Klout.  This suggests an update to the score happens once every X weeks.  PeerIndex appears to have quickly picked up the change in social engagement.  Now I am confused? Why is my Klout score moving in 10th of points seemingly randomly each day?  Klout seems to have taken absolutely no heed of the increased frequency, engagements, converstations and immersive social activity.

Next Phase – commencing December 18th

I have removed the following networks from my PeerIndex and/or Klout scoring authorisations.  In both cases, the only remaining social network is Twitter.

• LinkedIn
• Foursquare
• YouTube
• Instagram
• Facebook

Let’s see which of the two engines picks up on this change first and to what extent does removing all the other tied networks impact the reported score.

Baseline on start of test;

• Klout – 46.64
• PeerIndex – 55

Update – December 19th

It took just under 24 hours for @Klout to re-assess the influence score based on removing all Social Networks except Twitter to its new level of 42.43.  Therefore, the total decay in removing all other networks was only 4 points. This tends to suggest a heavy weighting on Twitter; which is correct, as it is my primary social network. It has had a much more marked impact on True Reach, Amplification and Network scores.

• Klout True Reach decayed from 641 to 533 people
• Klout Amplification Score decayed from 17% to 9%
• Klout Network Score decayed from 30 to 22

I will update this post when I notice any decay in the PeerIndex score.  The main control panel on PeerIndex is still stating that it was last updated 3 days ago.  Let’s see how long it takes to update.  I am now intrigued why the @peerindex score has remained completely static for almost a month; despite heavy social engagement.  As a baseline my total following has remained within plus or minus 10% of the starting figure at 930 followers. Given I tweet and talk only about a pretty restricted range of ‘on-topic’ subjects in the Customer Services, Customer Experience and Social Media genres the audience is quite restricted.

I also note that the XML – API is not returning accurate scores on @Klout.  3 hours after the online Score was updated by a forced refresh of the browser the Klout API is still reporting the old score.  I am now looking for 2 things;

• How long will it take @Klout to update the API feed (This is a key one, because it is the score we use in the strategies when we route social media interactions to customer services agents)
• How long will it take @Peerindex to update the score in any form from its static 55

14:45 British Time, the Peerindex score changed! It didn’t go down! It went up! The score is now hourly reporting 56; despite having disconnected all the networks authorised for its algorithm 24 hours ago.  I can only assume that this is a result of longer term monitoring!

Christmas Break 2011

I am going to return both influence engines to their standard baselines and connected networks and enjoy a couple of weeks off over Christmas.  In the meantime the engines will be reporting their scores on this blog. See you next year!