A minimalist desktop application to download data from your ViralHeat.com monitoring profiles.   The application lets you select which profile to download, and which channel (microblogs, websites, facebook, videos).  Mentions are downloaded to text files specified by the user, compatible for text mining.

You can download the free application from versiontracker.com, or from this link: viralheat-downloader

Recently, a colleague of mine (Nathan Gilliatt) posted a great article about a crucial and usually absent skill in social media analysis: boolean logic expertise (Can Analytics be Taught?)  The crux of his article - at least from my standpoint - is that the quality of an analytics effort relies upon the ability of the analyst to define his search parameters, and those parameters are usually defined using some interface to boolean logic. Boolean logic is a systematic way of defining complex, “and:or” descriptions, but it is generally only taught as part of computer science, mathematics, and engineering disciplines.

I have used, demoed and toured most of the popular social media analytics platforms out there, and pretty much every system offers some interface, “wizard”, or helpful manual staff-time solution to help would-be analysts overcome their inability to define their search and monitoring profiles in a useful albeit a priori way.  The different platforms are definitely not equal in how well they accomplish this task.  Poorly defined profiles end up measuring and generating great visualizations for, to a greater or lesser degree, irrelevant content.

Hey, We’ve Already Solved This Problem!

However, it has struck me that setting up analytics profiles like these are really problems of classification: are the social media items being examined in Category A (relevant to my brand) or Category B (not relevant to my brand)? It seems an obvious solution that most future analytics platforms will improve the quality of their search profiles through machine learning tools like Naive Bayesian classifiers. The process might look like this:

1. Define initial search profile using boolean logic or a logic wizard.
2. The analyst vets initial results, confirming or denying results as being correctly relevant to the brand in question.
3. Once a sufficient training set has been built (1000+ posts?), the system switches to “full automatic” classification
4. (optional) the classifier can continue to be trained over time, or sensitivity tests can be run to understand which terms are most significant.

A well-trained Bayesian (or other) classifier can reach incredible accuracy rates, 99% and higher.  But here’s the kicker: successfully defining a search profile this way will itself produce some deep insight about your brand’s reputation and about the things people associate with your brand! In other words, the classifier works well because it has come up with an accurate “definition” of a brand based upon the keywords that it has trained itself upon.

Coming Soon To A Theater Near You

I’m certain there’s a few analytics systems providers out there that already have a head start on this; I recall that at least Biz360’s dashboard implemented  machine learning tactics to improve analysis of industry-specific lexicons and jargon.  Given the continuation of Moore’s law and its extension to general technology growth, we can expect that features (like Biz360’s) of high-end systems will become increasingly available in middle-market and entry-level analytics solutions.

Since I’m using these tools to customize and create professional-quality graph diagrams, using data auto-generated from another application, it means I’m reviewing these tools from a pretty specialized perspective. However, maybe my thoughts can help guide your own, or at least let you know about some of the great tools out there.

yEd

My current favorite still remains yEd from yWorks. It’s completely free for non-commercial, and cross-platform.

Strengths: It offers an amazingly comprehensive set of layout options and import options from numerous graphing languages (GML, GraphML, yWorks’ proprietary YGF, XML, GedCom, TGF, and GraphViz .DOT with a little work). It provides a lot of options for node / connection graphics, shapes, fills, and even has some node auto-sizing and auto-coloring algorithms based upon network metrics like centrality. Via SVG, it can export fully vectorized images for publication or import into your Adobe suite. It’s free, did I mention that? It has a customized dock that will let you create a window of your favorite layout settings for quick re-application, and every layout method has lots of options for you to tweak.

I’m particularly fond of the annular (circular) and force-directed layouts, although the software’s ability to layout connections in a variety of methods (curved, organic, orthogonal, circuit-style) is pretty impressive too. You can interface directly with the nodes/connections to manually touch up your layout, and have yEd perform layouts on only sections of the entire graph (although I have yet to be happy with the results of this option).

yEd also lets you export to a small number of other graphics formats, giving it some functionality for use with other graphics programs.  It also has a few random network generation algorithms, a lot of editing / formatting toolsets (rotations, fit-to-label, grids and aligning) and even some basic analytical tools (centralities, clustering).

Weaknesses: It’s a java-based app, so it’s interface looks and feels clunky. The app never remembers my preferences, so it’s not native and tends to eat a lot of memory. There are unintuitive UI elements; I constantly end up accidentally creating new nodes when all I wanted to do was move a node or add a connection. While the software has a random graph generator (grid, hierarchical, planar), the planar networks it creates are pretty sad and aren’t complete (connected) even if you select the “complete” box; graph generation algorithms are well documented, so it’s disappointing they can’t do better here.

Omnigraffle

I’ve used Omnigraffle in the past for its integration with Applescript, and it definitely has the best OSX GUI integration of the bunch. The $200 professional version has additional options for subgroups and group layouts, as well as SVG export.

Strengths: OmniGraffle offers a wide variety of image export options including EPS, PDF, SVG, and a variety of bitmap image formats too; so, publication quality export is definitely available. While the suite of layout methods is limited (force-directed, radial, hierarchical, and some sort of force-directed circular), the algorithms do a decent job of making pretty graphs. There’s a lot of control over graphics and shapes for your nodes and connections, including shadowing, gradient fills, and 3D effects which are quite nice.

Weaknesses: Like yEd, its GUI can be unintuitive at times and requires a learning curve. It has few graph import options, but they do exist; I work with the GraphViz .DOT format to import into Omnigraffle successfully, and will probably experiment with OmniGraffle’s own “gdiagramstyle” format since it’s a pretty structured XML based format that seems to support complex features like opacities. OmniGraffle only saves to its own graph formats.

GraphViz

GraphViz is sort of the big name in the area of network diagramming, and you’ll find a lot of academic work that used GraphViz visualizations. It’s open source, and you have to love that. It has versions compiled for all major platforms. However, it really fails for me in terms of flexibility and ease-of-use.
Strengths: It has a solid suite of layout algorithms.

GraphViz natively exports to something like 40 formats, including PDF, SVG, and EPS vector formats. So, whatever you can generate, you can use in high-resolution publications or scale up as large as you like without losing detail.

Weaknesses: If I said the other tools had unintuitive interfaces, it’s not by comparison to GraphViz. You can’t manually interact with nodes or connections, and all of your editing options are limited to a single tool window with “graph, node, connection” settings shown in one long, unexplained list. GraphViz is very limited in terms of graphics; no 3D features to speak of, and all graphs end up looking somewhat boxy and flat. The toolbar does have built-in “what is this?” descriptions, but they often aren’t very helpful, eg: “Packmode: This indicates the granularity and method used for packing (cf. packMode). Note that defining packmode will automatically turn on packing as though one had set pack=true.” After playing around with it for 20 minutes, I still couldn’t make a complex graph I imported layout using a force-directed algorithm. I also wasn’t able to create nodes or connections manually.

AiSee

AiSee is another cross-platform tool that looks to have the broad capabilities of GraphViz; it has an impressive gallery of different layout types. Once imported, you have some interaction capability with nodes and edges. Overall, I found it more intuitive and useful than GraphViz, especially the Layout menu was easy to interact with. However, it was far less easy to use than yEd or OmniGraffle.  Like Tom Sawyer Software, AiSee seems focused on providing graph visualization libraries for integration into commercial software libraries.
Strengths: AiSee offers an unusual set of network navigation tools, such as “show neighbors”. It has a reasonable selection of output formats, including PDF, SVG, and EPS vector graphics and the usual bitmap image options. The software is free for non-commercial use.  Of the four tools I’ve looked at, it’s the only one that has a built-in layout animation capture option, if you want to demonstrate changes to your graph.

Weaknesses: Like GraphViz, I wasn’t able to create nodes or connections manually.  The limited formatting options means the graphs looked flat and boxy; I couldn’t find any 3D, shading, or shadowing options.  As far as I can tell, AiSee only imports its GDL data format.

Summary

I think for some applications I’m going to have to stick to yEd, but I’m going to implement some export options that will allow me to work with OmniGraffle for a additional variety and flexibility.  I’m sure I missed some details since this post was based only on a couple hours’ mucking about with the various packages I could find; I invite corrections from those who know better, but perhaps even my errors will be an indication of the learning curves you might expect with each package.

In the future, I may review network animation visualization tools like SkyRails…. those are always splashy!

Here’s a quick quote from Ryan Olson who was also impressed by the leaderboard:

“What makes Zen Pinball stand apart from all other pinball games is the fact that your scores are constantly fed into a leaderboard where you can filter results. You are able to battle the entire world or a small circle of friends for a top score.”.

The purpose of the leaderboard is to keep people “hooked” and active in the game’s network; to instill a bit of friendly competition, recognition, and self-pride into what would normally be a somewhat “solo” past-time.  In short, it’s what the Forrester Institute in the Groundswell book refers to as a reputation reward system.  What stands out in my mind is how well the Zen Pinball leaderboard mixes different levels and types of performance with a bit of banter… national level top scores, local networks, and then just to make you think you are doing well and encourage you to try a little harder, throw your own scores into the mix to give you the impression that everybody nationally is seeing your name pop up once in a while.  By including many different types of performance stats - not just top total scores, but “highest 60 second scores”, “best one-week scores”, and rankings among friends and regions as well as national, it creates more opportunities to make the individual feel they are getting recognized for their efforts.

Here’s a little video clip of a couple of Zen Pinball’s leaderboard text; my own login ID is “sky-spirit”, which you will see pops up right alongside a national high-scorer.

How is the Zen Pinball leaderboard useful for other social media applications? In general, it’s a smart and adaptive, network-based reputation reward system, designed to fuel and reinforce a desired activity or behavior.  Imagine a charity that leverages volunteers to raise donations via “fun runs”, walks, or endurance runs, such as the American Cancer Society’s Relay for Life, the American Heart Association’s Heart Walk, or the Susan Komen Breast Cancer Foundation’s Race for the Cure.  These foundations rely on their volunteers to leverage their personal networks; and while volunteers certainly contribute their time and energy for benevolent reasons, there’s plenty of opportunity to provide encouragement to reinforce and reward their efforts.  Imagine a Facebook application, iPhone/Droid app, or personalized email bulletin that regularly shares messages like the following:

• Person Y (your friend) is only $$$$ short of displacing Person Z as the top fundraiser in the Miami-Dade Region!  Help them reach their goal!
• Person W is gaining widespread attention, having raised over $$$ in donation commitments in one week! (an amount slightly more than you have recently made yourself)
• Person X is the #3 top fundraiser for the State of Georgia, having raised over $$$ in one month!
• Person Q (you!) is a tough competitor, for having raised almost $$$ in donations in one week!

In today’s “down economy”, charities are looking to leverage their core fundraising techniques better, and this would certainly be a creative way to bring new dimensions to old techniques.  Reputation/reward applications can be extended by creating tools for letting friends, family, and others send kudos and thanks as well, similar to Facebooks “so-and-so likes this” feature.

Remember, it’s not just about network connections, friends and followers, but reinforcing and rewarding network behaviors!

In the next few days, I’ll make sure the two sites handle these errors better, and present users with a “sorry, try back next hour” message.  Darnit!

“It is said that sometimes, the motion of a single butterfly’s wings can lead to a chain of events that build into a mighty hurricane.  However, expertise in raising butterflies does not translate into expertise in raising hurricanes.  Similarly, expertise in creating videos, even with a success story to point to, does not translate into expertise in creating viral social media!”

In short, VidSquid is a crowdsourced, free, audience response system for YouTube.  Take any YouTube video, and VidSquid lets users rate the video second by second, positive or negative.  Some of the applications for VidSquid include:

• peer evaluation
• improving speeches and presentations
• commercials and advertising
• discussing home movies
• message campaign improvement
• speech/language coaching
• user interface evaluation
• etc….!

The BarCamp team is producing a video of my presentation, but since the venue had technical difficulties that prevented me from showing my Keynote presentation on the projector, I’m not holding high hopes of how useful it will be.  Therefore, I’m attaching my presentation below as a Quicktime movie.  It’s a huge file - 91MB - and in movie format, since it includes a number of animations/movie clips (hey, VidSquid is all about movies!).

Although there are features I’ll be adding over the next few weeks, VidSquid.net is live and open to the public.  Give it a try!

Drs. Duncan Watts and Albert-Laszlo Barabasi literally wrote the book on the subject of the dynamics of scalability (Dr. Barabasi was kind enough to send me an advance copy, actually). The title “Linked: How Everything Is Connected to Everything Else and What It Means” is a wonderful introduction for laypeople and experts alike to the dynamics of network-based systems like the internet.

All things being equal, it demonstrates why “first to market” is so important in these environments.  As “scale free networks” form over time, new connections (read as “users” or “customers”) are more likely to connect to providers (nodes) that already have lots of connections… a sort of “snowball effect”.  The image below (from Nature magazine) shows how early nodes have an advantage for attracting connections, thus giving structure to the overall network.  In scale free networks, connections don’t follow a random bell-curve (normal) distribution, but rather what analysts call a “power law” distribution.  All things being equal, once a scale-free network is established it’s unlikely that network nodes with lots of connections will ever lose significantly to newcomers.

However, all things are rarely equal.  Just because you are first to market doesn’t mean you offer better or more satisfying value.  Thus we enter “competition” into the equation, and “first to market” doesn’t guarantee that you will never lose your position on the top of the heap.  After the forerunners demonstrate the value of the market, we often see “wanna-bes” follow.  Most of the time, newcomers don’t displace the incumbent, and in the marketplace either fade away, get consumed, or find a smaller niche to cling to.    However, Dr. Barabasi and others have introduced a second type of scale-free network: the “fitness model”.  The fitness model presumes that each market participant offers a different level of value, satisfaction to its subscribers; the provider that offers the theoretical perfect service, that meets all of its subscribers needs and provides unflawed user experience would be considered to have 100% “fitness”.

What does the fitness model research tell us?  First, that incumbents can be displaced by newcomers that offer sufficiently greater value (greater fitness) to overcome inertia, brand loyalty, networks-of-friends, learning curves etc. can eventually displace firstcomers for market share.  Newcomers that provide equal, or slightly better value will stay behind… there is a sort of network inertia that resists newcomers whose only real offering is “option B”.

But here is the shocker - the research also shows that sometimes, if the newcomer’s fitness is high enough, a sort of “crystallization” can occur where they steal all the connections in a sort of massive upheaval.

What does this mean for Microsoft Bing? Microsoft released its new search engine with massive marketing efforts to displace Google.  But Google is very much the fat spider in the middle of the web right now for search engines with a 65% market share.  Given that Comcast’s July 2009 figures don’t show any big movement, I think it’s apparent that for most users, Bing isn’t sufficiently better than Google to displace it.  However, remember that Google wasn’t the first search engine either; despite being the latecomer in 1998 or so, it beat out Lycos, Excite, and AltaVista because it provided more relevant results and a simple GUI.

But the next big question: what about Twitter and its competition?  Twitter has a very strong microblogging niche, and has grown to over 30 million users by some sources.  It has successfully resisted newer services like Pownce, Jaiku, Plurk, and Friendfeed, and attempts by Facebook, LinkedIn and Plaxo to offer microblogging services.  But now Yahoo is stealthily rolling out a new microblogging service, Meme.  Despite Yahoo’s size and experience as an internet giant, that puts it in the “newcomer” seat, and we must ask the question - how much more fit is Meme than Twitter?  Is it better enough to make a difference?

If you are an invited beta user, let me know your thoughts.  But factor into the mix that Twitter has demonstrated many problems that put it far below “perfectly fit”, some of which are only now beginning to make an appearance:

• Constant technical / growth problems (you can join the Fail Whale club!  A recent attack on one user takes down all of Twitter!)
• The development of spam tweets - and no tools for intelligently filtering/rulesetting them (no, I don’t want to join your mafia).
• Surprise!  Results provided by the Twitter API may not match reality.
• Unrealistic exponential follower bases (I mean really, 1 in 10 Twitterers listen to everything twought by Ashton Kutcher and Ellen Degeneres?  Seriously?  Does anybody really know what having even 20,000 followers means?)

So, this means the existing twinfluence site will suffer “as is” as we get out the wrenches and blow torches… but in the end twinfluence should be more useful than it ever was.

Hang in there!