RSStling with RSS

A few months ago I took stock of what I had to offer versus what I thought the market was going to demand. I had a strong feeling that monitoring was going to become really important, and so I’ve got to grips with Yahoo Pipes and Netvibes and made these two systems work well for me.

Measurement is also important – always has been – and so I’ve been looking into smart, repeatable, scalable, actionable, objective (objectionable?) ways in which to measure online performance. Again, I’ve had success with this and generally when I show people how to do it, they ‘get it’.

But one thing still eludes me, and it’s really really really important. And that is: how to produce reports.

It’s fairly easy to produce an outreach report, say, or a report that talks about what a small group of bloggers has been up to. But I need to scale, so that what works for ten sites will work for 100. Put it this way: I need to produce a report detailing the online coverage for the ITPOES report which has hit literally hundreds if not thousands of sites globally.

I’ve spent most of today wrestling with this problem. I’ve tried subscribing to feeds in Google Reader and pulling reports out of that. I’ve tried importing the xhtml into Word and getting that to play ball. I’ve even tried downloading so-called XML reporting apps that don’t really work at all, it seems to me.

So I’m still left with one more nut to crack. Reporting. Cloud computing has come to the rescue with monitoring – I know, I know, it’s my own in-house solution but it works for me – and I’d dearly love it to help with reporting. Just something that lets me format coverage, group it, allow me to feed it into other systems that can get a view on influence for each source. If anyone out there has a take on this, I’d really like to know.

Posted via email from Brendan Cooper – your friendly social media-savvy freelance copywriter and social media consultant.

6 thoughts on “RSStling with RSS

  1. Brendan, I have exactly the same issue, although I use Radian6 and Live Buzz from Market Sentinel to do the donkey work for me.

    Ultimately, where we add value is in sifting through ALL the “junk” and adding insight to the findings.

    Sure, you will always want to know sheer volumes, but the impact and voice of influencers on the campaign will be much smaller and therefore worthy of more attention.

    We use technology as above which, infuriatingly scales like mad, yet the insiught which delivers value (i.e. me and the team!) which doesn’t.

    We work to something called “Actionable Insights” – which is effectively a small set of rules which help us refine what we have seen on the monitoring tools. What impact has our outreach activity had and what conclusions can we draw from it?

    What was perception, how did the perceptions align themselves with our other marketing activities, what do we need to change etc etc.

    It will always need un-scaleable people involved but therein lies the value.

  2. Hi Paul,

    Very good point about the scalability – and suddenly I realise that I’ve touched on this issue myself before – see

    I wonder how long it will be before we can have accurate sentimenting and issues analysis? That would help – chuck in a load of content and see what comes out. But as for insights – that is, intuition based on experience – well I don’t see how that can scale, because I don’t think computers will ever be able to offer this. The best they can do is offer us data, information, knowledge to an extent.

  3. FYI, there are already tools out there that DO measure sentiment. Both the location of the mention (i.e. the words immediately surrounding the term being monitored), but also the context of the article.

    This enables us to determine whether or not a good mention is really good (in the context of a generally negative article) or vice versa. The sentiment score is also affected by the site credibility of the mention to ensure that a positive score from, say the guardian is recognised as being more influential or important than the same comment from a numpty’s blog.

    All very clever stuff!

  4. I’ve tried using them and personally I avoid them.

    I tried using the Twendz tool soon after it came out. It was on the day Jade Goody died. The very first tweet that came up said “So sad Jade died” – and was classified as negative.

    I posted about it at the time and a developer responded saying it was ‘correct’ behaviour. Maybe in the context of ‘death’ – as in, death is sad – but most certainly it was not negative about the subject matter ie Jade Goody.

    That is a simple example of how keyword proximity can get it plain wrong. Then add up all the nuances of English – not least irony – and you can see that it’s really riven with difficulty. I’m not sure I would be confident in front of a client in this scenario.

    I still think human beings are the only true measure of sentiment but they don’t scale. Perhaps there’s room for a compromise here: take the automated results and compare them with, say, human sentimenting for the top five most influential sites, or a random sample for confirmation. Or outsource a larger, more representative sample to an agency that does this sort of thing – even Amazon’s Mechanical Turk, for example – in which case you’re still exposing yourself by relying on someone else’s judgement.

    It’s not easy. Nothing ever is!

  5. I totally agree on the human front – a machine can only “learn” so much – which takes us back to the whole issue of scalability. Humans are needed to do it, but don’t scale.

    Interesting you use the Amazon Turk reference. I’d done some digging into that a while back and was gutted in a way, that they had pretty much turned humans into processing machines in the same way that they did with online storage (S3)!

  6. Ha! I guess we’ll all be Mechanical Turks before long…! Btw, thanks for the comments. I’d almost forgotten what it was like to have a good comments thread at the bottom of a post…

Leave a Reply to Brendan Cancel reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google photo

You are commenting using your Google account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s