Let’s be upfront and manage expectations right now: the answer is 42*.
Attribution is one of the favorite topic for discussions among marketing teams, and a classic argument either to request increase in spend budgets or to justify a lackluster performance in ROI.
It also epitomizes one of the paradox of digital marketing. The initial argument was that the precise tracking of every single investment and the according event on the customer size would provide an indisputable measurement of the return on investment.
Yet, we are in a bizarre situation where a number of companies are happier to accurately lose money on paid digital media, rather than investing in arguably less trackable but essential marketing activities such as awareness, content and social engagement.
The theory behind attribution is simple: before making a decision to buy a product from a supplier, a customer will carry out research and will be somehow influenced by the different interactions he has with the supplier along the way.
The attribution works consists in:
- listing the interactions in their chronological order
- appraising their contribution to the purchase. The 4 main models are first interaction (after all, without this one there might not have been others?), last interaction, and then an infinity of weighting between the first and last interaction
To make things more complex, we generally cannot track the same person across its different devices (different computers, mobile phones, tablet). More importantly, we cannot track the influences of the conversations with other human beings, articles read, programs seen, advertising on TV, radio, … and the comments on forums.
As a result, the quest for an accurate answer with some many unknown elements would easily find its place in other everlasting debates like the meaning of universe (42 is the answer).
In reality, the best way to break through this debate is pragmatic, by considering that there are two different levels of marketing activity.
1) The first one, strategic, consists in determining the right level for marketing spend in the company and to decide on the allocation of spend. These decisions are generally made for a yearly horizon.
This allocation requires indeed a degree of attribution work, but, in my experience, there are such massive gaps between the different channels (PPC generic, Price comparison, Review sites, Affiliation, SEO, CRM, Above the line) that the decision is almost indifferent to the first click or last click model. My recommendation is to appraise a cost per visit for each channel, with 2 different margin per visit, one being first interaction and the other last interaction.
For most of the channels, the ROI will be unequivocal, either positive, or negative, and so will be the hierarchy of the most to the least profitable channel**.
For a CMO, a pragmatic way to look at the allocation is to relate the share of revenue by channel to the share of investment by channel. They might realize for instance that their investment in Content and SEO (very contributive) is very small compared to an investment in paid-search (low or negative contribution).
If this is the case, the next marketing budget should aim at diminishing the gap between the revenue and investment mix, bearing in mind that, in the long run, building a strong brand requires first and foremost engaging consistently with its customers and delivering a strong experience.
To complement this approach, econometric techniques can be used: the best attribution model or more precisely ROI by channel, is the one which can best predict the return of the overall marketing budget for a given allocation. It is generally a technique relevant for longer time series, so revisiting it by quarter is about the right frequency.
2) The second level is operational, for each channel manager.
The purpose there is to improve the efficiency of the channel by optimizing the bidding process, getting accurate product feeds, selecting the right campaigns or partners, …
At a channel level, what matters most is the consistency of the measurement from one year to another. If the model is 30 days last paid click (often the case for PPC), the KPIs and targets have to be built accordingly, even if the expectations of the company are different.
One assumption there, but I have always found it robust, is that two different attribution models will track the same trends: as a result setting targets in the working environment for the channel (bidding platform for instance) will produce consistent results overall after deduplication.
As a conclusion, the responsibility for the CMO is to shape the marketing investment to deliver the long-term strategy for the company.
Getting guidelines on the expected return for each type of investment is obviously key, bearing in mind that there is no such thing as a definite attribution model. I will elaborate in a next post on some elements to help build a marketing mix.
* Reference to the Hitchhiker guide to the galaxy. 42 is the response to the question of the meaning of life …
** Of course, the profitability of brand paid-search will be very artificial and debatable, but the spend being generally very low, we can park the discussion here. There will be another post on this topic.