11 June 2017
Everybody knows the hackneyed “Half of my marketing dollars are wasted – if only I knew which half!” As the marketing profession has evolved, the concept of marketing accountability established an ever-stronger foothold. Whereas marketing used to be considered more of an art than science, the data-driven revolution has made a lasting change to the lives of marketers. Gone are the days of unfettered spending of marketing dollars. Nowadays, these budgets have come under scrutiny of traceability and accountability.
The rise of digital marketing has opened up a new avenue of near perfect sales-to-marketing attribution. Or so it seems. Search engine optimization (SEO) and search engine marketing (SEM) campaigns make it possible to establish quantitative models that allow budget allocation with surgical precision. Digital campaigns are like an experimental playground where you can run endless exact tests. Outcomes of these tests then drive subsequent campaign funding on the basis of prior (conversion) results.
Behind this radical change in marketing accountability rests a customer lifecycle model (!) that we assume (!) to calculate observed efficacy of marketing efforts. Some people refer to the “customer journey”, others talk about “customer lifecycle management.” It begins with a latent consumer desire or need, through information gathering and selection, to purchase. However, customers in the digital multi-channel world can follow an almost infinite number of paths. And here’s the rub: not all of the steps along that journey can be measured, at least not with comparable accuracy and reliability. Yet everything that contributes to making your product or service more available, either mentally or physically, can help to drive sales.
Conversion models are great for tracking on-line activity, but that is also where their shortcoming: they reduce reality to deliberately (!) simplify allocation of costs and revenue. A model is a purposeful reduction of reality. Customer lifecycle models are used so we can allocate cost and revenue. Like any model, this does not imply that the reduced version of reality is a perfect and valid representation of what actually happens “out there.”
We have entered an inherently multi-channel world. Internet and smartphones are ubiquitous. But although mass media spending may have declined in this mix, it has all but gone away! Conversion attribution is the attempt to quantify, as exact and direct as possible, how discretionary marketing spend relates to sales. This is the area where the lion’s share of progress in marketing accountability has been coming from. By measuring all successive steps for the on-line conversion funnel, we tie back sales to expenses in the digital world.
What does it mean when we say: “We are going digital”? For most companies, this “merely” implies they are adding (more) digital channels to their existing (fundamentally) analog business. The reality is that if your company and business model had a “pre-internet” footprint, the odds are that internet has augmented, and not replaced your existing business model. This is not to say that digital isn’t important, it is. But unless you are a digital only company, digital (apps) and internet are channels that have been added to the mix.
The “customer journey” is the (entire!) path a consumer has traversed, from initial need and awareness, all the way through to purchase. Note that some of these steps may be digital, and some of them might have had their roots in the analog world. And even for the digital part of the journey, not all components will be equally transparent and measurable. Yet we apply a model that assumes we can tie back digital sales to digital expenses (only). George Box is credited with “All models are wrong, but some are useful” metaphor, from his 1976 paper. The digital sales funnel is another example of a slightly over simplified representation of reality. Let’s have a look why.
Paid search like Google Adwords is a great example of a relatively transparent and accurately measurable digital channel. But it still has some caveats. You can turn digital campaigns on and off at very short notice, and then the rise in sales over and above the existing trend line is typically attributed to the Google marketing spend. Easy and straightforward, isn’t it? Note, however, that we assume we can reliably determine the trend line before and after this campaign. And also note we assume that competitor activity during this time period plays no role. There is yet another assumption: activity prior to this Adwords campaign is assumed to play no role. Factors like brand awareness are assumed not to interact with purchase decisions, at least not in the model… What should be clear from this array of assumptions is that the evidence may sometimes hint at violations of some of these assumptions, and in that case we are mostly going to ignore that evidence. The model may still be very useful, just don’t ever assume it is correct.
With the rise and spectacular success of these conversion models to track effectiveness of the digital marketing funnel, you get the impression that the limitations of this “customer journey” model gets pushed to the background. This makes you wonder, if companies are still employing a mix of media channels, then how come this mix doesn’t get reflected in the models they use? That appears inconsistent. Either you have “a” place for mass media and other channels in your model (typically rather difficult to measure with precision), or you stop spending money on it. But to calculate digital effectiveness with three significant digits, and not have a component in there for such a large chunk of your spending like mass media doesn’t make sense to me. The fact that these other channels are so much more difficult and expensive to measure can hardly justify ignoring these influences altogether.
Mass media has clear expenditures, and may be considered for inclusion in your models. But what about social media, and free publicity? Modern marketers can surely value the worth those bring to the brand. But most of this “free publicity” doesn’t come free at all. Someone, somewhere has to focus time and effort on it, and those are “expenses” just like buying space on a billboard, or running a TV or radio ad. Wouldn’t it make sense to attempt to factor these influences into the mix as well? If so, that implies they need to have an explicit component in your marketing mix model. In a time when sustainability matters, we need to consider how community engagement plays a role, too. And like this, there are several less tangible components that we invest time and money in, but often without quantifying how we expect these activities to pay off. They are simply not represented in classical conversion models.
Reality is that many of the “old school” marketing mix components are (much) harder to quantify than digital campaigns. But if we want to rationalize budget allocation, you have to make an attempt to measure. There is no reason A/B tests ought to be confined to the internet, and they shouldn’t. In this realm, important innovations like the Bandit algorithm have been applied, recently. Why not test a large-scale regional roll-out of billboards in such a way, for instance. Go “all in” in one part of the country, and stop using billboards for a while in another – and then measure the impact, etc. Marketing accountability has evolved, and our approach needs to reflect this. Needless to say, for these less pronounced effects, you need a solid grasp of Statistical Power Analysis to be sure you don’t abandon effective channels prematurely (see an editorial on Statistical Power Analysis I recently wrote about this subject). If you aim to optimize your marketing dollars, doesn’t it make more sense to optimize holistically, rather than merely within your digital channel? For multi-channel companies, the relative spend across “old school” and modern marketing channels justifies a more integrated approach to measurement and modeling.
In my opinion, instead of a fragmented view where you attempt to optimize for each of the customer contact points individually, a more holistic view is called for. Analyze your customer journey as a whole, and determine which components in the mix could do with more funding, and which components in the mix that budget needs to come from (where you choose to spend less). Immediately, an organizational challenge surfaces: how many companies can “move budget around” without interfering with long-term departmental plans, and without triggering World War III with regards to budget allocations?!? The typical annual budget cycles form the basis of so much planning and staffing decisions that that an agile responsive stance in light of research findings is a tall order at best. Needless to say, you would also need an independent and “strong” research arm to weather all of the storms an inconvenient research finding might trigger…