Today’s CMO occupies a challenging and dynamic place in his or her organization’s ecosystem. He is accountable to the CEO for producing results – whether these results are leads, sales, brand equity or a combination – that help achieve the company’s business goals. He partners with the CRO (chief revenue officer) to implement programs that fuel the sales organization’s pipeline and empower it to close business. He is accountable to the CFO for producing those results at an acceptable cost, and must justify future budget spending based on past and predicted performance. And he manages a team that’s responsible for implementing a veritable Rubik’s cube of marketing channels, campaigns and tactics designed to produce the results that his C-suite colleagues are expecting.
For these reasons, marketing attribution should be the CMO's best friend.
Back Up Just a Half-Step
In today’s multichannel, multi-touch marketing environment, marketers almost universally acknowledge that measuring success – and optimizing future spend – based on giving full conversion credit to the last touch or last act experienced by a prospect before converting is a highly flawed, inaccurate methodology.
And most advertisers are coming to realize that subjective, rules-based methodologies for assigning credit – where the marketer simply picks an arbitrary percentage to assign to each touch based solely on its order within the conversion path – are also highly flawed. A rules-based methodology can actually be even less accurate than last touch, as it only accounts for position within the conversion chronology and doesn’t even consider the channel, creative, size, publisher, or any other attribute of those touches that can have significant impact on producing the conversion.
In contrast, algorithmic attribution gathers and analyzes every attribute of each touchpoint experienced by all prospects across all channels to produce a customized model. These models examine every attributes and touchpoints combination to identify the lift in conversion produced by every touch experienced by every prospect. It then recalculates the organization’s success metrics based on the ACTUAL impact of every touchpoint. Although marketers tend to reject the first two methodologies, they intuitively tend buy into the validity of the algorithmic attribution methodology.
That’s Where the Friendship Starts
Implementing and utilizing algorithmic attribution within a marketing organization provides the CMO with an ecosystem whose performance is much more accurately measured and predicted. As a result, it’s much easier for him to meet the needs of his C-suite colleagues, as well as to chart a strategic and tactical course for his marketing colleagues. Here are a few key deliverables of attribution on which this friendship is built:
It’s a Matter of Trust
As with any friendship, this one is built on trust -- trust that’s built upon seeing the actual result more closely match the predicted results -- and seeing a lift in performance using true, attributed metrics compared to last touch or rules-based metrics. As trust is built, confidence is similarly built, both within the mind of the CMO and her organization, and within the C-suite colleagues with whom she collaborates.
Anto, an excellent informative article for Mediapost's readers.
I just want to add on about rules-based metric. Its true that rules-based system severely under-performs in dynamic environment such as marketing since behaviors of customers change over time, but this shortfall can be remedied via hybrid-computing (or soft-computing as berkeley people called it). One popular hybrids today is neuro-fuzzy which is still dominant in the engineering domain as in electronics and control system design but it has now found application in general data analytics.
Neuro-fuzzy, which is a combination of neural network and fuzzy logic inference, can take an initial sets of (fuzzy) rules (those rules could have been designed by a human expert, lets say by a marketing expert or simply the rules were discovered autonomously via fuzzy-clustering) and then optimize or adapt the parameters of the rules. The final sets of rules are now more robust than the initial rules (because the rule parameters have been automatically tuned for optimal performance). Rules that are not important from the initial sets are simply pruned or deleted as they contribute nothing to the prediction/classification (feature selections).
Hybrid computing such as Neuro-fuzzy has been shown to have outperformed other pure crisp segmentation/clustering & classifications algorithms in most cases. IMO, I think that rules-based system still have a role to play here in marketing analytics but the rules must be tuned first by a neural network before deployment. The other good point about this is that the Neuro-fuzzy learner is adaptive, ie, when the data changes, then it adapts itself by continual tuning of its parameters. Manual rules (those designed by marketing experts) are still valuable, because, human experts can sometimes spot something that makes sense where a machine can't (especially the black-box system like neural network).
The following publication may be of interest other software designer readers. Google it, the PDF is freely available for download.
[e-CRM – Data Modeling Using Adaptive Neuro Fuzzy Model]
I totally agree with your approach. I have to comment though, that in order to really create an understanding about the market place and marketing effectiveness, you need to look at to big picture along with the in-depth analytics. I would call this challenge, many companies have "corporate autism". More about how to cure corporate autism here :) http://futurecmo.org/2012/05/30/cure-for-corporate-autism/
I have been testing lateral data dashboarding a lot lately. Just visualizing variety of data like: marketing spending, sales, email visitors, total visitors, un-paid search, weekly research results with top-of-mind and spontaneous adrecall or brand recall, social media sharing and brand hits in articles, Google search index for the brand and their competitors, weather conditions, competitors spending, etc.. It is just amazing what kind of insights you can come up with by just making graphs from data that is rarely connected. The corporate autism article was about recognizing the 1% you need to conciously pay attention to and understand the big picture. Lateral data analysis and visualisation tells a story that would be impossible to come up by looking at any in-depth datasource. The beauty of this is that it can be done in excel and it only costs time. Once you have done it, it's much clearer what you need to pay attention to and also the in-depth data and analytics start giving you a lot more than before. Try it! I'd love to hear about your experiences too!!!