A cure for adtech complexity?
Going back to basics with campaign measurement in 2022
“Just because we can measure it, that doesn’t make it important.”
- Anonymous Sports pundit decrying “advanced analytics”
If adtech were to stand still today, it would be very difficult to navigate.
It isn’t doing that, however. The opposite is true. There are massive sea changes in effect and the complexity marketers have to manage is nothing short of overwhelming.
Between cookies and their crumbling state, governmental regulations and the Google-Apple-Meta merry go round, the changes in the landscape are increasingly difficult to forecast and plan for.
Whether spending thousands or millions, a tech stack today could include: Demand Side Platforms, Supply Side Platforms, Consent Management Platforms, Brand Safety Providers, Data Management Providers, Customer Data Platforms, Viewability Solutions, Malvertising Providers and a range of Publisher/Revenue/Network solutions. (You’re welcome. I spared you the acronym nightmare this typically involves.)
So where does one begin when taking stock of measurement and what good looks like in 2022?
Let’s begin with the major principles to keep in mind when it comes to advertising and measurement; attribution, incrementality and brand lift.
“What’s worse than no attribution experts on your clients team?
Two attribution experts on your client’s team.”
I voiced that aloud and stumbled backwards into an adtech Dad joke (dadtech?) when the CMA’s Adtech Committee last met. We were joined by Franni Segal, a Global Sr. Lead of measurement strategy at Quantcast, one of the leading technology companies in the measurement and insights space. She provided a refresh on the measurement landscape, and shared some of their company’s solutions. Attribution was a major part of the discussion.
What’s attribution anyway?
Attribution helps assign numeric values to touchpoints in a consumer journey – helping identify what should ‘get credit’ (and future budget), when it comes to ad campaigns.
Skimming a Consumer Journey 101 textbook will take you down the path of Awareness, Interest, Desire or Action (or other funnel frameworks). Viewed through the lens of a campaign, each stage might correspond to a different channel (TV, social media, radio, print, video ads etc.) and tactic.
Using multiple metrics, ideally a selected attribution model is used as a framework to make sense of usually noisy datasets. There are many competing (and sometimes conflicting) models like multi-touch, first-touch and, last-touch, each with pros and cons, and varying levels of cost and complexity to establish and defend. Quantifiable goals are possible, but these are usually in flux given the dynamism of all the moving pieces.
In my experience, the best attribution experts tend to be more the lone wolf type – they are data savvy, comfortable with deep work, and have just enough social decorum to sell their beliefs internally and externally, without upsetting egos involved.
(Yes, when experienced marketing leaders make sunk-budget choices based on relationships, past experience or instinct, showing them data that invalidates their decisions isn’t always met with receptive ears).
All is not hopeless, however. There are best practices: defining a “single source of truth” when it comes to the data, fixing on a consistent “lookback window”, and associating first-party data conversions for your brand/business can all establish a firmer foundation for your attribution model (and ensuing battles).
“When a measure becomes a target, it ceases to be a good measure.”
- Charles Goodhart
The economist Charles Goodhart really captured the nature of how circular the subject of measurement can be. Working backwards from your own conclusions can be the tempting path of less resistance for most marketers.
To help steer towards the unbiased, incrementality refers to the practice of comparing two or more groups of people to understand the impact of advertising (i.e. if one group is exposed to an ad and the other isn't, does the exposed group take action at a higher rate?).
Testing this scientifically can be a challenge as impact is not always conscious. The ad “seer” might not realize the ad was seen and even impacted their purchase decision.
Franni helped point out a few different methodologies, whose merits I’ll summarize.
- Post Hoc Incremental - Can be done after campaigns but has debatable control groups which could skew towards being imbalanced.
- Non-PSA Holdout Incrementality - Data can be noisy and it tends to underestimate results due to sample size and other constraints.
- PSA Incrementality - Likely the most balanced and accurate, but costly and therefore quite prohibitive for many brands.
Of all measurement models, choosing an incrementality methodology is especially case-specific.
“Every advertisement should be thought of as a contribution to the complex symbol which is the brand image.”
- David Ogilvy
Brand lift as a practice measures the impact of campaigns on consumer brand perception. Often survey-led, and usually by third-party companies (such as Kantar, Nielsen, Lucid or Quantcast), it’s an attempt to validate or invalidate insights on many aspects of the ad campaign, including timing, messaging, format, demographics and targeting.
Various providers have different takes on the practice, with recent advancements allowing for real-time survey results, ultimately making the data more actionable (i.e. with the goal of affecting the current campaign mid-flight and not just to make the next one better).
“The only form of ethical persuasion that exists is when the goals of the persuader are aligned with the goals of the persuadee.”
- Tristan Harris
When it comes to cookies, let’s cut to the chase about the (WAY) less fun and calorie-free version of the word.
The digital version. Cookies (a file stored on your device by websites, browsers and apps) have long been the way to measure user activity and interest. They’ve been manipulated to be invasive by some ‘bad actors’ in the adtech landscape over the past decade and a half and now they’re going away. Sort of.
Browsers are blocking them, platforms are limiting their usage (especially third-party cookies) and marketers as an industry are working towards the idea that targeting and measurement can be executed effectively in a more privacy-preserving or anonymized way.
This noble stance brings challenges. Targeting and measurement can become a lot less precise for advertisers (no matter what kind of actor: good, bad, or high school drama class improv).
On the positive side, there’s opportunities with the new cookieless environments.
- First-party data helps measure opted-in consumers, and alongside Consent Management Platforms, this above board methodology still offers opportunities for brands to deepen relationships with their customers, with no loss of personalization of messaging or offer.
- Contextual advertising has made huge leaps in effectiveness across platforms and providers; machine learning and AI computing now allows for much better discovery of niche targeting and ad placements compared to even 5 years ago. (Yes, finding high-intent eyeballs is possible without knowing about every single purchase, vacation, medical ailment, fast-food vice and banking habit of a consumer.)
- Multiple data signals that are built for interoperability across an ad stack can help provide guidance. Examples include cohorts, unique identifiers (via consent), geolocation, time, language and device. Measured in a high-dimension framework, these can be used to make targeting decisions that can be optimized and scaled for accuracy.
Many decisions need to be made when selecting a post-cookie identity solution. Should you use a probabilistic or deterministic model? Is your application omnichannel or only for specific channels? Is it for insights and targeting only, or also for conversion measurement? Standalone or interoperable? Free or licensed? Media specific or agnostic?
Barring a crystal ball or time machine, these questions aren’t usually answerable definitively.
Given the state of flux in the industry, it would be wise to experiment at the smallest possible scale before fully leaning into what works for your organization.
The ‘tech’ in adtech has always intended (or at least sold us) on helping humans manage complexity. That said, there has undoubtedly been a law of diminishing returns when it comes to this complexity.
“Is it even worth doing all this measurement?” is a fair question to ask, without risk of sounding like a marketing caveperson.
We’re undergoing a long-overdue maturation of the adtech landscape – from changes in government regulation to changes led by established players and an industry yearning for more unified adoption across regions.
Removing complexity altogether isn’t necessarily feasible, nor is it a cure-all, but we can reduce it while evolving in response to consumer privacy trends.
What's left to be seen is whether this evolution arrives via more cohesive adtech solutions and collaborative practitioners or by standards brought about my monopolistic big players.
For advertisers, hopefully a more manageable and stable ground is on the horizon.
It’s early in 2022, and we’re just not there yet.