For a ongoing intranet consultancy project I undertook a set of stakeholder interviews as part of the research process. As we headed into the stage of assimilating the mass of gathered information - and there is a staggering amount of it - into a clear set of recommendations from observed patterns, I wondered if there was a way to visualise the interview content with the notion that it may enable further understanding.

Reading between the lines

To this end whilst wading through recordings of stakeholders, I converted them into tag clouds. Using TagCrowd, a Tag Cloud generator, I upped a transcript from each interviewee.

The resulting clouds, ignoring words like “oh”, “going”, “anyway”…,  whilst not scientific (and not the point) give an intriguing look at the content of the conversation, its themes and the ideas that are high, possibly subconsciously, on the interviewee’s agenda. It also made apparent common patterns across different interviewees.

As a supporting device I like them. Whilst independently inconclusive they give the designer a way to communicate, in an engaging visual way, to other people involved in the project the areas interviewees are gravitating towards.

The resulting images were  also well received by the client. I’m sure in part due to their simplicity. It is a diagram amidst dendrograms, tables of stats and other analysis that can be  easily digested and as such opens conversation.

Below are some of the generated tag clouds. I’ve blurred out a few client sensitive words.

^ Interviewee 1

^ Interviewee 2

^ Interviewee 3

Prospecting for gold

To compliment tried and tested formal methods I am an advocate of creative experimentation as part of this clarifying process. The willingness to peer into the problem without a clear route to the solution and knowing that it will take a bit of poking around, some dead ends but the chance of discovering something with value makes it worth pursuing. Having the time to do this however is difficult to manage which is why i think it is best done in the early stages of the process for two reasons. Firstly, if there is nothing springing from the well you can cut the activity and resolve solutions with existing methods. Secondly, the freedom from rationalised existing solutions helps to keep thinking expansive and experimental.

If you have applied existing approaches in unconventional ways or have made discoveries through serendipity post a comment.

The other day I read an interesting article about the NetFlix challenge that is very much the Zeitgeist for me so I thought I’d put it out there…net enabled recommendation engines will become critical to our everyday life activities. Your chosen engine’s result set / ‘suggestions’ will become as valuable as a good friends’ opinion when searching for that thing.

The engines

You might have heard about the NetFlix $1m recommendation engine challenge. No-one has won it yet, two years later. One of the stumbling blocks are films like Napoleon Dynamite which are proving to be very hard to predict preferences for. All good personalisation services out there have some kind of recommendation engine. I’m thinking about Apple’s Genius feature in iTunes, Amazon’s people who bought this…, Gmail ads, Last.fm’s musical neighbours..the list rolls on. In fact even Ocado just mailed me saying they have launched a service of this ilk. Awesome. I can offload a bunch of thinking.

Give and take

What these engines / algorithms are doing is taking search on a step from that of a generalist aggregator - if we both type “Palmer Eldritch” into Yahoo the same set of results come up. In fact Google, as usual, are leading the way in Search with their SearchWiki by allowing users to personalise the results using the ‘Remove’ and ‘Promote’ controls added to each result item. Over time actively helping Google learn what I think of the results will generate a set of preferences that enable it to bring me more appropriate results. Nice.

^Personalise your search engine

This type of exchange is uncomfortable for many people. It necessitates that we submit our personal choices and preferences to a privately held company, often times one that it relatively unknown to us. It raises questions such as; how will this data be used? who else will access to it? is it secure? what are the ethics of the aggregating company?

In light of this there is certainly a measure of trust that must be apportioned to the aggregator and as a consequence the residual benefit that it provides must be perceptively higher. As these services become more common the threshold for concern is likely to reduce as people become more used to deep mined personalisation in much the same way as online shopping has become ordinary.

Connected services

Facebook. The monolith. Launched Facebook connect last summer, its version of OpenID. This service lets you log into multiple disparate site with one central universal login. The idea is to wave goodbye to a plethora of passwords, which is potentially great for the user experience too.

So Facebook collect and post your activity to your wall – privacy setting permitting. This is great for the uber social sharers and also enables Facebook to start to mine your personal preferences even more so than before, supplementing their currently already abundant social graph. This gives them a great opportunity to fire content and services directly at your cerebral cortex however, given its central aim to connect people, the way they have implemented this data is still rudimentary. Speaking to friends who are daily users the adverts are often not relevant and the “People you might know” are generally people they don’t want to know.

As noted with Last.fm this kind of thing is already happening and dating sites make their living from it although Facebook (and BeBo, FriendFeed et al) have the opportunity to do it better as their data sets traverse many facets of a persons character and interests.

Interestingly this thinking has been applied to Twitter, with the recent Mr. Tweet service although i’ve not spent much time with him yet.

Speak to me like you know me

It’s out there, it’s happening but the software still feels like an algorithm. The successful service must be smart although it shouldn’t pertain to know it all. It should allow the user to amend returned suggestions and learn from these. The suggestions should be timely and have at its core appropriateness. Appropriateness of context and goal. It needs to know about the user, what mood they are in, what’s coming up on their schedule and how that could impact their preferences. It needs to be sensitive to the things the user finds offensive and talk to them as their friend might.

If anyone has used a great recommendation service or is creating one I’d be love to hear about your experience.