Approaching Attribution Modeling in Marketing and Advertising
When done properly, attribution modeling will expose both the good, the bad and the ugly of your marketing operations. Some people call it, measuring ROI, some people call it performance measurement, or sales behavioural measurement.
Either way, once complete, it is extremely powerful at allowing organizations to redirect company resources toward the areas that matter.
Early on, a decision needs to be made as to how far into the sales funnel you wish to go (see funnel diagram below). This is the point you’ll need to draw a line for measurement of Marketing’s perceived role of responsibility with regards to departmental goal setting.
For example is the goal to generate short term impact with lower commitment outcomes such as brochure downloads and store visits? Or do should marketing delve deeper into the generation of marketing qualified leads (MQLs), sales qualified leads (SQLs) or even focus on attaining high CLV customers above all else?
How you choose these target metrics will radically change the perception of performance of decisions made upstream. It will also radically affect the attribution modeling process.
There are essentially two disparate approaches when it comes to this very polarizing discussion. Note: these decisions you’ll make as a marketer will be fraught with significant political risk.
Group 1 – The Traditional Approach
These departments are typically led by an investment-fund-style manager who will outsource the majority of the departmental function to a selection of agencies all attempting to work together toward a common goal.
Defining exactly what this specific goal is, will vary considerably between businesses. Some teams will focus on ‘conversions’, others will prefer a focus on easily achievable vanity metrics or the perennial ‘brand [something]’.
Media choices will heavily favour paid media advertising, perhaps accompanied by a few topical buzzword initiatives for good measure. Each 12 month cycle will be precipitated by a creative ‘big idea’. Agency firings and ‘re-brands’ will shortly follow announcements of poor sales performance as will the default rebuttal statement that, “there was increased brand awareness”.
When probed, the relationship between the main goal and sales revenue is rarely proven nor tested to any great degree.
The team culture will be either gut-led or opinion-driven where personal veto powers will rein supreme during a highly political decision-making process.
Phrases uttered during internal discussions typically include: reach, brand awareness, engagement, customer experience, customer journey, touch-points, downloads, impressions, clicks, share of voice and many other iterations of the word ‘brand’.
This group, I’d estimate to cover the vast majority of marketing departments around the world.
Group 2 – The Datarati Approach
In the second group lies a much smaller, but growing collection of professionals who are taking advantage of new technology and more-often than not, more consistent in their growth trajectory.
They will prefer to control critical and resource-intensive tasks in-house, not just to conserve resources but also to improve agility. Select functions will be outsourced to agencies only when necessary – not by default. Some organizations will in fact, control all tasks in-house and senior staff will share a collective disdain for agencies of any type.
It’s not unusual for creative production, digital media buying and measurement to be completely the domain of internal teams. The general work approach will favour scientific data-led testing over any one person’s opinion, and a continuous improvement cycle culture will dominate with a tolerance for the risk associated with experimental failure.
The media approach will be agnostic, mixing both organic and paid media, partnerships, affiliate/referral programs, sales promotions and more, into a holistic growth strategy that will change quickly if necessary. When probed, the team will be scientifically able to prove contributions to sales revenue and the correlation between different performance metrics.
Phrases commonly uttered during internal discussions will include: northstar, customer database, audiences, conversions, sales-funnel, lead-scoring, split testing variables, SQLs, MQLs, CAC, LTV, CLTV etc.
From experience, the main forces preventing Group 1 from becoming Group 2 are fourfold:
- Ingrained cultures
- Shortage of skilled professionals
Measuring the contribution of marketing to sales revenue is not always easy – especially when there is a mix of offline and online influence and your offering has a service component.
On one hand it’s too easy for The Datarati to be dismissive of traditional media and frequently exhibit a strong prejudice toward digital mediums. While digital is inherently easier to measure, the value of offline touch-points should not be ignored. Trust and brand credibility are traditional media’s strength. Too often I find the Datarati view rife in the SaaS tech and Direct to Consumer (DTC) industries, but their views are less valid in other sectors.
Offline media, including ‘archaic’ mediums such as the radio, can have a powerful effect on the overall purchase decision – especially with higher consideration offerings.
Recent personal experience selling million dollar paintings at auction has only further exposed the limited nature of digital’s effectiveness in this specific product case.
Direct mail brochures can be cheaper and more effective than Google Ads when targeting younger demographic segments as this audience rarely receive physical mail! It still astounds me how ‘channel agnostic’ marketers will still exhibit prejudice toward some channels, especially traditional media.
It is important that marketers realize that the yield of different channels is fluid and not fixed when making longitudinal comparisons.
What’s old, often becomes new again, making it shortsighted to treat leanings from books such as ‘hacking growth’, ‘the lean startup’ or ‘traction’ as marketing gospel without first performing your own tests in a modern context. A lot can change in the digital space within the space of months, let alone years.
On the other hand, it’s all too convenient for The Traditionalists to rest on their laurels and continue with:
- what is familiar
- what used to work,
- what they believe to still work, and..
- trusting a distorted commission-led traditional media-buying market
Just because something is difficult to measure, doesn’t mean it shouldn’t be measured. After all, “what can’t be measured can’t be managed”.
Outsourcing work to external vendors is also a very convenient way to limit any blame, but the negative effects can also severely hamper team morale and skill transfer. Who wants to be a glorified ‘agency coordinator’ when their peers are uncovering valuable marketing insights which will future-proof their careers.
Which one are you?
This burgeoning field of ‘marketing attribution’ or ‘attribution modeling’ is the holy grail of any marketing department and a field which I receive frequent requests for my expertise.
However, the answer to your attribution woes is complex and varies significantly between organizations.
Custom attribution models are frequently the preference of the larger organizations. They are validated over time as the model constantly tested and validated by live feedback data. However, this approach is expensive to initiate and maintain over time.
Smaller firms will preference a mix-mash of digital dashboards, base much of their decisions via analysis courtesy of various click attribution models and they will often struggle to combine their online-online contributions. Attribution software tools are often poorly implemented, misinterpreted or myopic in their scope.
Choose your metrics and data approach wisely
From recent experience there still seems to be much variation between the metrics used between businesses, even those within the same sector.
IMO – attribution experts must be highly skilled in three areas…
- A deep knowledge of the performance metrics related to each medium which extends beyond vanity metrics.
- A discerning ability to critique the accuracy of performance data, and…
- An understanding in customer purchase behaviour psychology
Without these three critical areas, conclusions can be very easily misguided and the any subsequent attribution models flawed.
A common misstep by both groups is an over-reliance on descriptive data vs predictive or prescriptive data.
Which of the two groups above best describes your department and how do you attack the attribution questions from your CEO or CFO?
This guide was written to get you thinking about how to approach the inevitable questions sounding return on marketing spend investments.
Tell me your thoughts below or reach out to me for a discussion specific to your situation.