At last week’s annual Omniture Summit, there was a lot of buzz about social media and Twitter specifically.  In my SiteCatalyst Power User session, one of the things I covered was an idea about how you can leverage Omniture SiteCatalyst to monitor your company’s brand reputation in tools like Twitter.  This concept seemed to really resonate with the Summit audience, so much so, that I was given an opportunity to share it at the closing session.  The following will describe the concept in greater detail for those who could not attend my Summit session.  DISCLAIMER: What I presented is a “proof of concept” and is no way a formal introduction of a new Omniture product.

Business Scenario
So this idea started with me thinking about how cool Twitter is and how it could be used for marketing purposes.  As an Omniture Twitter ambassador (Omni_Man), I see people talking about Omniture all of the time on Twitter.  Sometimes this chatter is positive, sometimes it is negative.  I usually try and send my colleagues at Omniture “tweets” that I think might be relevant to them, but this can be very time consuming.  So I said to myself, “SiteCatalyst has a Data Insertion API that is used to inject non-website data into SiteCatalyst and Twitter has an API associated with its search.twitter.com website, so if you put the two together, why couldn’t you pass Twitter information into SiteCatalyst?”  Doing this would allow you to do many cool things which I will describe below.  I enlisted the help of one of our Omniture Consulting geniuses and within 24 hours, we had a working prototype.  The following examples of this functionality will use Comcast (who I co-presented with) as an example, but these represent test data and is not meant to imply that Comcast is using this functionality.  In my session, I posed the following hypothetical business scenario:  You are the web analytics manager at Comcast and your CMO returns from an executive retreat where he/she has learned all about Twitter and believes that Comcast needs to do everything it can to monitor what its customers are saying on social media sites.  The CMO calls an emergency “all hands” Marketing/PR meeting and demands to know the following:

  1. How often Comcast is mentioned on tools like Twitter
  2. If there is ever a spike (positive or negative) in brand-related terms (in a week, day or even hourly)
  3. Who are the people most often mentioning Comcast on social media tools and who are they communicating with the most?
  4. When are people on social media tools mentioning key Comcast product/service features that Product Managers should know about

So at this point, the CMO turns to you and asks if there is anything you could do to help… What would you say?  Not sure?  Let’s tackle them one at a time…

Monitoring Brand Comments
As stated earlier, the key to this solution is leveraging the SiteCatalyst Data Insertion API and the Twitter API.  For those of you not familiar with the SiteCatalyst Data Insertion API, it is used to send data to SiteCatalyst when a JavaScript tag is not an option.  By combining this with the search.twitter.com API, we can set a SiteCatalyst Success Event for every “Brand Tweet” by pumping the results of a “Comcast” Twitter search into SiteCatalyst.  This allows us to see a metric chart of “Brand Twitter Comments” so we can track it by month, week, day or hour.  However, why stop there?  SiteCatalyst has a built-in Alert feature that allows you to be notified via e-mail or mobile device when a Success Event metric hits a threshold or changes more than a specified percentage.  Why not take advantage of this feature and send yourself (or others) an Alert when your brand is mentioned 25% more this hour than last hour or decreases significantly day to day?  This would allow you to stay on top of what is going on in Twitter without having to constantly monitor Twitter every day/hour!  Below is a screen shot where you can see the “Brand Twitter Comments” Success Event and an Alert related to it being set:

Who’s Tweeting?
The next thing your CMO wanted to know who are the most active Twitter users that are tweeting about your brand.  Are there some really good brand advocates out there?  Are there people who are repeatedly bashing your brand?  Are any of your employees “going rogue” and confusing the marketplace with mixed messages?  Again using the API, it is possible to extract the Twitter user name associated with every tweet.  In our proof of concept we did our best to extract the author and the recipient, with the latter being more difficult since there are times when there is no recipient or multiple recipients (we are still working on this).  However, by placing both in separate Conversion Variables (eVars), we could breakdown the “Brand Tweet Comments” Success Event metric by author to see who is twittering about your brand the most.  We decided to take this one step further by creating a Conversion Subrelation between the author and the recipient so you could break one down by the other (note that if there is no recipient we used “[No Recipient]“).  This allowed us to see who was tweeting with each other the most often.  I imagine that this could be useful to see what types of people have formed virtual communities and some companies might consider contacting the key members of this virtual community to gather product feedback/suggestions or to leverage them for brand promotion.  You could also use SAINT Classifications to group Authors into meaningful buckets once you knew who they were (i.e. Customers, Vendors, etc…).  The following is a screen shot of the subrelation report we created:

Mining Important Product Keywords
The final thing the CMO tasked us with was discovering when, in addition to our brand name, social media users were mentioning specific keywords that are product or service related.  For example, if someone “tweets” about Comcast and in the same tweet mentions “speed” it is likely that this tweet is related to high-speed internet access and could be interesting to the Internet product manager.  Alternatively, if a “tweet” mentions Comcast and also mentions “Tivo” or “DVR” it is likely they are expressing an opinion in the digital TV recording arena that would interest the associated product manager.  So you have millions of opportunities to read what your customers are saying, but who wants to scan through all of those “tweets” to find the relevant ones, especially if this has to be done manually?

This got me thinking about SiteCatalyst’s search functionality.  If we had all of the “tweets” in SiteCatalyst, you could perform a keyword search and let SiteCatalyst find all of the comments that mention a specific keyword.  For example, let’s imagine we use the Data Insertion API to pass all Comcast “tweets” to a Conversion Variable (eVar) and then conduct a search for the phrase “Tivo.”  SiteCatalyst would isolate those “tweets” and then you can bookmark that report and schedule it to be e-mailed to the appropriate product managers at whatever time interval you desire (hourly, daily, weekly, monthly, etc…).  This way, no one at your organization would ever have to use or look at Twitter, but instead, the information they need to see would be pushed to them automatically.  Best of all, there is no limit on how many searches and bookmarked reports you can create so you can create hundreds of different keyword searches and send them to different groups of people using Publishing Lists.  The following screen shot provides an example that shows “tweets” having been sent into SiteCatalyst and a user entering the phrase “Tivo” in the search box and a highlighting of one “tweet” that would be found:

As you can see, if you know a lot about SiteCatalyst including API’s, Conversion Variables, Subrelations, Searching, Bookmarked Reports, etc… you would be able to amaze your CMO by answering all of his/her questions and be a rock star!

Track Multiple Brands
Another concept related to this that we have explored is the idea of tracking multiple brands.  There is no reason why Comcast, in this example, could not also capture “tweets” about its competitors or subsidiary brands to see them side by side.  This would require the use of an additional eVar or potentially some additional Success Events, but we got this working in our prototype.

Next Steps
As mentioned previously, all of this was done as a proof of concept, but as you can see, the concept has great potential.  We at Omniture are going to explore this topic more and hope that you do the same.  We hope to add more information about this to the Developer Connection.  We will also continue to explore new ideas related to this, but I encourage you to leave comments here with your ideas on how this concept can be extended.

If you want to learn the implementation details of this solution, please refer to the following technical blog post: http://blogs.omniture.com/2009/02/24/implementing-twitter-data-tracking-in-omniture-sitecatalyst/

Have a question about anything related to Omniture SiteCatalyst?  Is there something on your website that you would like to report on, but don’t know how?  Do you have any tips or best practices you want to share?  If so, please leave a comment here or send me an e-mail at insidesitecatalyst@omniture.com and I will do my best to answer it right here on the blog so everyone can learn! (Don’t worry – I won’t use your name or company name!).  If you are on Twitter, you can follow me at http://twitter.com/Omni_man.

Learn more about Omniture Consulting
Learn more about Omniture University
27 comments
myforexnow
myforexnow

Great info, but, is it updated? I saw is from 2009. Thanks for the articles

BUNANE
BUNANE

Interesting!I was hoping you might be able to add some additional detail as to what the [No Recipient] rows mean in the context of the reports above.Like it~

Adelard Gasana
Adelard Gasana

I really love omniture. It's really easy to use, and integrating it is a simply one, two, three!

SEO Professional
SEO Professional

Adam, any idea as to when this will be intergrated into Site Catalyst? Also, would this concept expand into other social API platforms like Facebook etc.?

Rudi Shumpert
Rudi Shumpert

D-Rod, The pull from the twitter api should be returning an element "source" for each tweet. You could filter out anything sent by your api -Rudi

D-Rod
D-Rod

Adam, the Twitter insertion API (the one used to post Tweets to Twitter from outside of Twitter.com) allows us to pass a "source", which is where the Tweet came from. Here at MLB.com, we have our own Twitter app and would love to see how many Tweets are coming directly from our app compared to all Tweets. I would assume this is something the OMTR Twitter API could query no?

Twitter Search
Twitter Search

thats great that you are talking about the twitter api,a good example of searching with the twitter api is on twiogle.com because you can search on twitter and google at the same time.

Ben Gaines
Ben Gaines

Jim: Yes, you can timestamp the tweets if you're using a separate report suite with timestamping enabled; ClientCare can turn this on for you. Note that timestamping must be done in a separate report suite, as timestamped hits cannot be mixed with non-timestamped hits. This would be done by taking the <published> element value out of each entry returned by the Twitter Search API and putting it into the <timestamp> element in your Omniture Data Insertion API post.

Jim Hazen
Jim Hazen

Hi Adam, We actually have started this up but wondered if there was a way to get a timestamp for the tweets using the API?

Jason
Jason

I've put together a fully configurable Twitter-Omniture plug-in based on Adam's design. The plug-in is written in PHP and comes with a config file that lets easily define your Omniture data (connect information, variables used) and your Twitter Search criteria. http://emptymind.org/?page_id=69

Praveen Pandey
Praveen Pandey

Awesome concept, moreover the presentation is quite interesting in the wake of "Social" buzz

vanscoy
vanscoy

Elegant solution. I love it when platforms work together like this. Of course, I'll tweet this article. Even better, I'll actually bring it up face-to-face to challenge my colleagues. Thanks, Adam.

James Gurd
James Gurd

Hi Adam Loved the post, really interesting reading. I'm trying to get my head round how our Clients can put together a business case for social media to demonstrate tangible benefits to the doubters. Tracking Twitter brand activity and user comments would be excellent. Like others I would welcome a "how to" guide to digest. I've posted a link to this blog on my twitter account because it will be of interest to a wider audience. http://twitter.com/einbusiness_JG thanks james

Sebastian
Sebastian

That's really great stuff! Great job Omniture!!!!

Kristi Barrow
Kristi Barrow

Adam, thanks for the great post. Your description of the CMO back from executive camp scenario was spot on! This happens all the time (twitter, facebook and once upon a time Google!) and it's great to have you guys pro-actively working on solutions. I am looking forward to implementing this kind of feature for our clients and will be reading with interest the implementation details. thanks! Kristi

Ben Gaines
Ben Gaines

John: You make an excellent point. For this exact reason, in my post on implementing Adam’s Twitter tracking solution, I recommend using a separate report suite to track Twitter action. It is downright inaccurate to lump “tweets” together with the actual page views occurring on your site—they should never be mixed under any circumstances, really (unless you need to inflate your PV/visit/visitor data, I suppose). If you haven’t seen it yet, check out my blog post at http://blogs.omniture.com/2009/02/24/implementing-twitter-data-tracking-in-omniture-sitecatalyst.

John Nguyen
John Nguyen

We developed a similar Twitter app over a month ago but found that, unless Omniture provides a way in the Data Insertion API to flag something as "external" (this came up in Brett Error's feedback session), you will see an inflation in overall page-views/visits (the Data Insertion API requires that you provide a page URL and/or page Name--you can also provide a unique visitor ID which should prevent a huge spike in visitors but again, page views go up every time). If you're tracking many keywords/brands, you might see a skew in your traffic data. If you're using calculated metrics, this can throw off your numbers. Ideally, Omniture should provide a way to insert "external" data which can then be correlated to things like traffic and conversions. As of right now, if you look at a graph of page views vs. tweets, you're going to see an increase in traffic as the number of tweets go up, which might not necessarily be true!

Adam Greco
Adam Greco

Pere- Another idea would be to use the SAINT API... Adam

Adam Greco
Adam Greco

James - Great point. We are looking into sentiment. Our hope is that partners help us extend this. An auto-classification feature would be ideal, but has been a long time request. I like the idea of a VISTA rule based upon hundreds of keywords... Adam

Adam Greco
Adam Greco

Jordan - The "how-to" is coming in the next day or two. Look for a follow-up post by one of our more technical folks... Adam

Adam Greco
Adam Greco

Greg - In this example "[No Recipient} is when a user simply "tweets" without targeting it at someone specific. We are looking into cases where there are multiple recipients, but that is a bit trickier. Stay tuned. Adam

Adam Greco
Adam Greco

James - Great point. Measuring sentiment would be a more advanced solution, but I could see looking for specific keywords and using a VISTA rule. We have dreamed of a SAINT "auto-classification" feature for years but no luck as of yet. There is always SAINT, but that would be somewhat unbearable. Perhaps Omniture or partner developers can solve that one... Adam

Jordan
Jordan

So did I misunderstand at the closing session that were was a "how-to" coming shortly for users to implement? I've been sold on the need for this, just need the how-to to implement.

pere rovira
pere rovira

awesome post, adam! such a simple solution, and yet so powerful. thanks for letting me know, will write about it on the blog this week when i get a chance. i guess you could also apply classifications to the tweets, to see the relative impact of the diverse brands of Comcast, though it could be quite consuming... actually, now that I think of, it would be really cool if you could make a search on sitecatalyst, and then apply a classification to all of the search results. cheers pere

Greg Asman
Greg Asman

Adam, Great post and great presentation last week. I was hoping you might be able to add some additional detail as to what the [No Recipient] rows mean in the context of the reports above. They appear to mean that they were general tweets sent out by the "From" and the others were either replies to the individuals listed or tweets that included their name. It might be nice to be able to distinguish between the two. Thanks for the clarification in advance! Great meeting you! Greg

James Dutton
James Dutton

Adam, Quite interesting - will you be posting sample code to the developer website so others can replicate and build on this? Some great proofs of concept explored here, with some interesting ways to have users adopt some of (I must assume from your previous posts) those under used features of Site Catalyst. I'm curious to learn more about how you would plan on addressing sentiment, you mention "Sometimes this chatter is positive, sometimes it is negative." but don't suggest any ways to track and monitor sentiment; something that is always top of mind when I am reviewing social media analysis tools. Cheers, J