The Drum Awards Festival - Extended Deadline

-d -h -min -sec

Facebook Shops Ad Agency Facebook Ads

Measure your marketing like it’s 2022



Open Mic article

This content is produced by a publishing partner of Open Mic.

Open Mic is the self-publishing platform for the marketing industry, allowing members to publish news, opinion and insights on

Find out more

March 1, 2022 | 9 min read

The opportunities that come with a holistic strategy anchored in incrementality

The changing ads ecosystem

It should not be news to any marketers that new regulations and policies are implemented across the industry to meet people’s growing expectations around privacy. In some cases, these changes have required advertisers to think differently about how to run and measure the impact of digital marketing, as log-level data – all data relevant to a single ad impression – is less frequently available. This means businesses may find it harder than before to accurately measure and report on campaign results across channels.

Marketing measurement was always challenging

Even when log-level data was widely available, connecting the dots from business outcomes back to marketing efforts was no easy task and many businesses relied on proxies, such as last-click attribution. With its simple and straightforward logic and availability across tracking systems, it is easy to understand how proxy-based reporting became the norm for many, when trying to quantify the worth of marketing in the currency of business KPIs.

In general, the underlying assumption to all attribution models is that, because touchpoints – clicks or impressions – happened on the path to a conversion (correlation), they ultimately caused it (causation). Similarly, a marketing mix model assumes that when business outcomes, like sales, correlate with marketing spend, then such spend is their cause.

While models based on correlation assume (that is, take for granted) causality, the main goal for an experiment is to prove the causal relationship between marketing and business outcomes. In an experiment, statistically identical groups with the same conversion propensity are identified. While one group is exposed to advertising, the other group is not, allowing marketers to identify the incremental value driven by ads, by comparing results from the two groups.

Correcting for efficiency gaps in attribution models

Once ground truth is established through experimentation, reporting models can easily be validated to help understand how well they reflect true incremental value. A recent study from Meta found that last-click models undervalued marketing on FB/IG platforms by 47% on average. That means reported value from last-click models should have been multiplied by 1.9 in order to reflect reality. Misalignment between true business value and reported value from last-click models was even stronger higher up the funnel, stressing the challenge in using a last-click model to understand value driven from campaigns aimed to increase brand metrics such as awareness and consideration.

When taken at face value to inform strategic and tactical decisions, last-click attribution and other proxy-based models could stop advertisers from reaching optimal efficiency levels across channels. Having compared winning marketing tactics between incremental and non-incremental measurement approaches, it was found they disagreed nearly 25% of the time. And when marketers choose the wrong tactic, businesses lose out on an average cost-per-action improvement of 64% according to research.

In other words, there are efficiencies to be gained by rooting your decisions in incrementality.

Navigate the new marketing landscape with incrementality

​​Admittedly, it may sound complicated to run an experiment, but many advertising platforms have streamlined experimentation and for a good reason; advertisers with a test and learn approach tend to run more cost-efficient campaigns. In fact, a study from Meta showed that advertisers that ran at least 15 experiments on Meta in a given year, saw 30% higher ad performance compared to advertisers who did not experiment at all. In addition to this, advertisers who had also run 15 experiments or more in the previous year, saw a 45% increase in ad performance, highlighting the positive longer-term impact of a test and learn approach. Harvard Business Review agrees about the role experimentation should play: “It’s time to get more serious about experimentation and to use it strategically to guide marketing investments and build c-suite confidence”.

While successful marketing has always been in the interest of advertisers, boosting incremental revenue has become more important than ever as advertisers try to adapt to the new marketing landscape. Incremental measurement approaches, which allows for decisions to be made under the light of causal performance, is not a nice-to-have, but a crucial must-have, when trying to navigate towards success.

Misalignment across channels can be expensive

A question asked by many marketers is how much to invest and where. An optimal channel-mix should not just translate into a profitable and incremental return on ad spend (ROAS), but should aim to maximize the same. If media budgets can be allocated across channels in a more effective way, or if untapped opportunities are addressed with increased investment, this can stimulate overall growth. Consequently, a sub-optimal channel-mix can be expensive in terms of missed sales or diminishing returns.

When exploring advertisers who ran a Cross-Publisher Conversion Lift study on Meta to measure incremental value driven by their usual strategies on Meta and paid search, 75% had a strategy in place where Meta was significantly more cost-efficient than paid search. On average, an incremental conversion driven by paid search ads, was 5.5X more expensive than incremental conversions driven by ads on Meta. This illustrates how businesses may find it difficult to optimise their strategies across their full channel-mix and highlights an opportunity faced by many: to anchor marketing strategies in incrementality and see stronger business results as a result of more effective marketing decisions.


Let’s explore how this can be done.

Experiments bring harmony to measurement

An effective measurement framework considers all points of sales and provides marketers with the right tools to inform both tactical and strategic decisions. A framework adopted by many, is one where marketers rely on attribution for tactical day-to-day decisions and turn to marketing mix models (MMM) when making decisions of a more strategic nature. Few however question how different models add up, and since different solutions often contradict each other when not calibrated, it can create confusion in the decision making process. This is where experimentation comes into play.

By running regular experiments across all marketing channels, ground truth is established and models such as attribution and MMM can be validated and calibrated if needed. Experiments bring harmony to measurement and make the overall framework incremental, allowing marketers to make efficient decisions on the back of data that reflects reality well.

What type of experiment to run, will depend on channel measured and what solutions are available. At Meta, the golden standard for measurement has for a long time been lift studies for conversion and brand metrics. Lift offerings have evolved with the changing ads ecosystem and Conversions-API (CAPI) Lift and Private Lift are two options that allow advertisers to run accurate and actionable low funnel experiments on their Meta campaigns, in a private and secure way.

Another approach to consider, which is particularly useful when resilient channel-specific solutions are not available, is geo-experimentation. Due to its reliance on first-party data only - often aggregated sales by geographical location - it’s a relevant approach to consider in the privacy-first era. Geo-experimentation can be tricky to get right, but luckily there are multiple open-source solutions on the market to consider. GeoLift, for example, is a new addition to Meta’s Open-Source Techniques and specifically developed to make it easier for analytical teams to run robust geo experiments across marketing channels.

A holistic approach can improve ad effectiveness further

While channel specific experiments allow advertisers to identify incremental value driven by each media platform, how channels interact with one another is rarely explored. The journey from discovery to purchase often stretches over various touchpoints and over a longer period of time, so the question is: does a holistic approach to marketing translate into more effective campaigns?

British online Fashion retailer J.D Williams wanted to explore this and ran a Cross-Publisher Conversion Lift for their marketing campaigns on Meta and paid search to explore potential synergies. They found that the two channels worked harder together when driving new customers. While Meta proved to be 3.5x more cost-efficient than paid search in driving new customers, Meta ad effectiveness increased by 26 percentage points when allowing paid search to run at the same time. Prospecting customers clearly engaged with both channels and ad exposure across multiple touchpoints increased the overall likelihood to purchase when being new to the brand.

When knowing how channels correlate, marketing activities can be aligned for optimal return. This indicates how channel strategies should not just be anchored in incrementality, but should be defined holistically across the full channel mix to increase effectiveness further.

Bringing it all together

The marketing landscape will without doubt continue to evolve as policies and consumer preferences continue to shift. A measurement framework that relies on inaccurate and outdated proxies can create more confusion than clarity and research suggests business growth could be slowed down as a result.

It’s time to view experimentation as a driver for growth and a unifying source of truth. Three key steps to help you measure your marketing like it’s 2022:

• Run regular experiments across each media channel and calibrate your existing reporting tools. Utilize resilient channel specific solutions when available like CAPI Conversion Lift for Meta and consider Open Source Techniques like GeoLift for cross-media experiments.

• Compare different on-channel strategies by running multicell experiments. Don’t forget to act upon results and implement the winning tactics

• Strive towards a holistic mindset by aligning strategies across channels and exploring channel interactions.

Important resources

- Conversion API for a resilient data integration with Meta.

- GeoLift for geo experimentation.

- Robyn for marketing mix models.

Get in touch with your Meta representative to set up a Conversion Lift

Facebook Shops Ad Agency Facebook Ads


Industry insights

View all
Add your own content +