Here are my slides from my talk about Measuring Conversations at
Enterprise Social Media Event last week the notes below serve as a commentary:
Its just another Channel
Social Media is just the latest medium that companies will use to get their message across. If you watch Mad Men, there is a great subplot as the "New Boys" in TV grow in a tradition Ad agency
The Same Old...but Different
Like every new channel, its has some new features. Social Media has:
- a feedback loop - and it is not buffered, and is often real-time, unlike Web 1.0 systems where you ave more time to react
- you are not the "start" - its not star configured, its a small world networks, ie there are links in the network that you cannot mediate. it is an influencing, not a broadcsating, game.
- they can watch you too - Social Media is too way - users get as much of an insight into you as you do into them. Customer bulsh*t is easy to spot.
Where is the Value?
The Big Point I wanted to make is that Direct Measurement of Social Media is a second order thing, you need to deploy it where it gives bang for the buck - and to know how many bucks are banging is the First Order of business - get that right and accurate SM measurement is less critical. We are old school, and believe that ROI is a real number. There are only 3 ways of creating value in a company:
- Increase revenue
- Decrease Costs
- Utilise capital assets (including people) more effectively
Long term, Social media will only succeed in Enterprises if it does this
Increasing Revenue
Revenue is more predictably increased the closer to a sale one goes - the corollary being that a customer is worth far less the more they are just a prospect. Social Media today is mainly about prospect recruitment, so the value per customer is still very low.
Reduced Costs
So far Social media has been best at talking to existing customers and potentially reducing churn, which is a high impact area if churn is high (as it is in say mobile telecoms). Direct cost reduction/efficiency increase activity has been far harder to make work so far.
Capital Expenditure Saving
By and large the most immediate returns can be made in increasing utilisation of existing assets, then increasing efficiency, and only then decreasing future spend. in many service businesses (the ones Social media is most likely to be used in) a lot of the capital assets are the "wetware" - the people.
For both Opex and Capex reduction it depends on a company's cost structure to understand where the highest bang for buck is - its more variable than revenue. Using Social media to increase communication, teamwork, customer contact etc are all good things, but at least if you set it against a hard financial background you can calibrate what you can afford to spend on Social Media for a particular effect
Now, on to actual Measurement of Conversations
Not all Conversations are Equal
As well as understanding where value is created in a business, ite worth going therough the age old science of segmenting customers into "A" class - those who are high potential and spend, "B" class - those of average potential but higher proclivity to spend, and "C" class - the long tail - who you can waste a lot of time and money serving. Understand teh value of these different classes and resource assets accordingly. (Our experience is that treating high value clients well with Social media gives very good results fast, using it to cover Long Tail far less so in the short term)
Meme Machine
We (and doubtless others) have found that measuring snapshots and linear conversation trends ("buzz") are not particularly illuminating - what you need to do is find how different trends interlink and change over time. The way to do this is what we call memetic monitoring, trying to interlink the various topics and see how they shift over time.
We also like to monitor memetic differences between the client and the competitors to see who is impacting what differently
Social Capitalists
Understanding social capital online is at a very rudimentary pahse right now, and I think a lot of what is written is frankly new-age hope and dreams over harsh experience. If it sounds fluffy, it is probably bollocks. From a business point of view the most interesting argument si between the Gladwellian view (a small number of very infleuntial nodes) vs the Wattsian view (a much more dispersed environment) - see
this earlier post
Measuring the Metadata
Metadata (data about the data) is a key to measuring conversations - a lot is predicable about any one person by the links and activities they partake in. "Serious" data mining only takes place once good metadata analysis is put in place
The diagrams in the last 2 slides are
courtesy Mat Morrison aka @mediaczar). The "Social Capitalists" slide shows that the contacts between US Democrats and Rebublican Congressmen on Twitter is architected such that communication between the 2 groups is mediated through 2 main nodes. It would be a mistake however to deduce that they are the most "influential", as Twitter is a tiny subset of the communication transactions between all these people.
The "Metadata" slide was where mat showed you could predict which party a particular person belonged to very well by looking at their network.
Algorithms vs People?
Our experience is that if there are huge volumes, algorithms are necessary. If there is a lot of variety, people are necessary but if its low volume then its not worth heavy investment in algorithms. If you have both, you need both (cf Techmeme which started as an algorithm system but has increasingly had to use human editors)
The Software
There are a huge range of solutions/packages/etc on the market, but there is no end to end solution so one way or another you will wind up sticking them together. A typical solution will include the following operations:
Identify - Market Segmentation, Online vs Offline, User Data required etc
Search - Medium (Blogs, Twitter, Search Engines, Databases); Speed (Real time, most popular, discrete event search etc)
Aggregate - including Sources (eg Feeds, Platforms) and Subjects (eg a meme across multiple platforms)
Track - Trends over time (eg Buzz, Traffic) and making Comparisons (eg positional shifts vs competitors, meme shifts etc)
Analyze - Statistics packages (First level analysis) and Data Sifting (2nd Level – Pattern Finding)
The Wetware
Because of this range, the market is still in the phase of "stitching together" solutions for any one company that wants to measure "the conversation". What we can say is that, in increasing order of cost:
- You can go quite a long way on open source / share ware / freely available software
- Beware “freeware” from providers that then require you to hand over your data. I include Google in this as their motivations are increasingly worrisome
- Existing Software packages – most don’t integrate to give a total picture, and it is costly to knife and fork them
- System design houses – build and integrate solutions - get what you want but costs a lot of money
At any rate, MeasurementCamp (measurementcamp.wikidot.com) is an excellent resource - it was founded precisely because of the confusion in the space - offers Case Studies, a Wiki and other resources - and meets at least monthly in London
