It's all go with Firestarters right now. Tomorrow we have Cory Doctorow in the house, and last Thursday saw our inaugral spin-off event for the search marketing community. Around 200 of the great and the good of the search industry came together to debate the thorny subject of attribution – the use of data and technology to help marketers understand how their media truly performs. It's a fascinating and hugely important subject, not least because there remains so many unanswered questions.
Attribution is not easy to do well. Customer journeys are not linear and yet we often treat them as though they are, meaning that marketers often end up measuring the wrong things. Recent research that Google Analytics conducted with Econsultancy showed that last click is still the most common method for attributing value, with over half of agencies and clients citing it as the model they used. Of-course, there may well be circumstances in which last click is perfectly appropriate but often there will have been multiple inputs and influences from a diverse set of media and touchpoints (including other search terms/sessions of-course) along the way. So the promise and the challenge of attribution is in understanding the individual and relational value of different elements.
Easier said than done. As Ben Kunz has said in his post on the difference between randomness and chaos in marketing, figuring out which input causes a desired action or outcome is difficult: "Like balls on a pool table, you strike in one direction but quickly interference from other elements leads the target to another angle." So to help us unravel this complexity we had four great speakers from the search industry who each had ten minutes to give their own take, followed by a panel and debate.
Martin McNulty from Forward 3D set the theme by talking about what he called 'aspirational attribution' – how attribution is not a fix for rubbish marketing ("you can't attribute your way out of awfulnessness"), and how it's not the same as measuring but is instead about improving your odds for success. The important things he said are around correllation, causation and control – to focus on the simple fix rather than rushing straight to the big complex problem, to concentrate on where you have good quality and scale in data, and to use it to identify improvements in things you can actually control.
Helen Southgate from BSkyB said that poor attribution practice often shortens the journey unnaturally. A customer who has bought Sky has typically seen upto 50 different impressions, so a big part of what they do is targeted towards understanding the value of brand in digital marketing. She talked about the pressure to not only justify spend, but also to identify opportunities and inefficiences, and emphasised the importance of centralising data from many sources, working out what your objectives are and starting with understanding what the very top-line data is telling you before leaping in and shifting budgets. The bottom-line from the client-side is that we have to be doing it, and we have to understand it, but it's important not to try and solve all your problmes at once. The ultimate win here of-course, is the opportunity to test different messages, tools and media at different stages and eventually to respond to that data in real-time.
Similarly, David Richards (Head Of Search at ZenithOptimedia) talked about how attribution is not a silver bullet and should be used as insight, not a fix for ROI issues. If 90% of your transactions come via one click, then attribution modelling won't help you much. It should be the role of the agency to push clients to make improvements to areas that will make a difference like small changes to transactional websites that can make huge difference to business outcomes. This raises an interesting question about agency access to client proprietary data.
Tom Cull (Director at Carat Manchester) emphasised the impossibility of accessing all relevant data and building a model that answers everything. It's only worth doing attribution modelling, he said, if you can act on what the data is telling you. So go into it with your eyes open and be realistic about what is driving value.
Tom also touched on a subject which came up in the discussion afterwards – the importance of aligning incentives with objectives on the client-side in order to avoid unnecessary politics and silos. The lively debate which followed the talks seemed to arrive at a consensus that this is not a perfect world, there is no right model, no perfect tool, and no perfect data. But instead, it is important that attribution be dynamically generated based on data, and not selected or chosen. Use a standard, and build a model which is bespoke to individual scenarios. Use the data as much to tell you what the wrong model is, as the right one. And then continue to adapt and change the model to account for the myriad shifting elements involved in more effectively measuring and forecasting advertising amd marketing outcomes.
We clearly have a way to go to, but the future is an enticing one, and it's all to play for.
You can see a Storify of the event here.
Leave a Reply