The $1 trillion dollar Advertising world was shaken to its’ core recently, thanks to privacy. Over the past years, Advertisers have been scrambling to find solutions to their marketing measurement, sending many to try marketing measurement methodologies that account for privacy. Among the most discussed methodologies are Media Mix Modeling (MMM), and GeoLift experiments. INCRMNTAL’s methodology works differently than those of MMM and GeoLift. This article delves into the key differences, comparing the capabilities of  each methodology across various comparison parameters.

Adapting to Constantly Changing Environments

Marketers make frequent changes to their campaigns: Changing bids, changing creatives, starting / pausing / stopping campaigns, increasing or decreasing budgets, and testing new channels. Operating across several channels our own analysis shows that the average digital marketer makes over 1,400 changes to their campaigns each month!
Frequent changes pose a real challenge for traditional marketing measurement methodologies, as those measure the impact of an advertising channel, by the variation of performance before and after certain changes happened to the ad spend of the channel.

But in an environment where advertisers operate on multiple channels, and conduct dozens, or even hundreds, of frequent changes – traditional methodologies such as MMM, and GeoLift, fail to distinguish which of the changes lead to changes in marketing results. 

MMM depends on gradual and infrequent changes to measure channel performance effectively. This rigidity limits its utility in rapidly shifting environments where marketing dynamics are in constant flux.

GeoLift, meanwhile, requires controlled conditions during experiments. Changes to campaigns or external factors must be restricted, making it difficult to use in situations requiring flexibility or frequent adjustments.

Both GeoLift and MMM were originally invented for a world where Advertisers operated on mediums such as Newspapers, Radio, TV, and Billboards. Advertising was a lot more “static” in those days, and changes were not very frequent. 

MMM and GeoLift have been adapted for today’s world, but frequent changes still pose a challenge for the methodologies to operate properly. 

INCRMNTAL employs marketing activities as "micro-experiments," enabling continuous measurement even in fast-evolving marketing landscapes. Frequent changes are the fuel the INCRMTNAL proprietary models need in order to continuously measure incremental value. This adaptability ensures marketers can make data-driven decisions without requiring major disruptions – beyond those they would be doing already. 
 

Measuring Campaigns and Ad Groups

While marketing and budget plans typically work on the level of channels, or sometimes even mediums – day to day operational decisions are handled on the level of granular campaigns, ad groups, and even individual keywords for channels like Search. 

GeoLift’s ability to measure campaigns and ad groups is theoretically possible but practically constrained. Isolating experiments without introducing additional exposure often proves challenging, limiting its effectiveness.

One of INCRMNTAL’s standout features is its ability to measure campaigns and ad groups while accounting for dependent relationships between entities. This approach prevents errors from growing exponentially as the number of entities increases.

MMM, however, treats each entity independently. This results in errors scaling exponentially as the scope of analysis broadens, leading to potentially unreliable insights for campaigns with many moving parts.

Evaluating New Channels

Most Advertisers do operate with only a few channels, but scaled Advertisers often run campaigns with a dozen or more channels, constantly testing the performance of new channels and new mediums to reach new audience or increase their reach.

Testing the incremental value of a new channel is a mission critical task for any marketer operating at scale. 
When it comes to measuring the value of new marketing channels, INCRMNTAL’s incrementality approach excels. Its methodology is designed to provide insights into new channels from the outset, making it the most straightforward solution for marketers exploring uncharted territories.

When starting a new channel, within only a few days, INCRMNTAL’s platform is able to provide an insight: Is this new channel generating positive incremental results? if so – what are these? And should the advertiser continue scaling this new channel? Or should the advertiser monitor or even scale it down?

One of the strongest values in the INCRMTNAL platform is the ability to measure the true incremental value and contribution of new channels. 

MMM, on the other hand, requires substantial ad spend on new channels before meaningful insights can be derived. This upfront investment may deter marketers from experimenting with novel opportunities.

MMM needs to have “enough” ad spend and “enough” training data to estimate the results of a new channel. Typically, this will mean that an Advertiser will need to allocate around 10% of their ad spend to a new channel, and run campaigns with the channel for a period no less than 4 weeks before they can get a read of the incremental value of the channel. 

To most advertisers running at scale – this would be an absurdly expensive and unnecessary ask. 

GeoLift can evaluate new channels effectively, but its results are limited to the duration of the experiment. This means the insights may not account for long-term performance or external influences.

Operating Without Running Experiments

The gold standard of incrementality measurement is to use a randomized control test (RCT). Inspired by pharmaceutical tests, a randomized controlled trial introduces true randomization, and compares the results of exposed (to advertising) audience vs the results for those unexposed.

Today’s privacy landscape, restricting access to user level data dramatically hindered advertisers' abilities to run RCTs. Therefore, advertisers often defaulted to the 2nd best option: Planned Experiments (GeoLift Tests). 

Planned experiments and GeoLift tests are the best alternative to RCT, however, conducting a planned experiment is an extremely difficult task, as while running the test, advertisers should avoid making any drastic changes in advertising activities, or product feature release. Tests may often also be affected by external variables such as what competitors do, if there are any holiday effects, or even unknown variables that might affect the result of the tests.

Running a Geolift test will often mean that an advertiser must turn off or decrease their ad spend on a certain channel, or channels, in a region or across a whole country. This creates a disruption in operation, and whatever results may be achieved would only be relevant for the period when the test ran. 

INCRMNTAL’s ability to derive insights from day-to-day marketing activities as "micro-experiments" is a game-changer. This always-on approach eliminates the need for traditional holdouts or controlled experiments.

MMM relies heavily on holdouts and calibration through GeoLift experiments, making it less dynamic. GeoLift itself necessitates iterative experimentation, requiring significant planning and resources for implementation.

Adapting to Channels with Changing Performance

Is Google a channel? or an ad platform which includes Google Search brand campaigns, Google Search non brand campaigns, Google display network, YouTube Videos, Universal App Campaigns, Performance Max algorithmically optimized campaigns, and so on ? 

In the eyes of measurement platforms, especially in those of MMM and GeoLift, given that the granularity of measurement insights are limited, the performance insights often stay at the level of channel, or even medium (i.e. Search, Social, Display). 

But comparing the results and performance of Google Brand Search to YouTube Videos is an unfair comparison. Channels with an even bigger diversity of sub placements such as Advertising Networks, and Programmatic DSPs which bid on thousands of various apps and websites are often misrepresented in MMM or GeoLift testing platforms. 

MMM assumes that channel performance remains constant during the measurement window, which can lead to inaccuracies in dynamic scenarios. GeoLift’s frequent experimentation requirements make it impractical for tracking volatile channels over time.

For channels with fluctuating performance—such as demand-side platforms (DSPs) and ad networks—INCRMNTAL’s model thrives. By avoiding priors or static assumptions, it remains flexible and reliable in changing conditions.

Getting Started: Calibration and Readiness

Modeling and Experiments based methodologies typically require weeks to run or train on the data before they can provide useful insights. 

A planned experiment will often go for a period of two week and up to several months, during which, Advertisers will attempt to isolate each change, so that the results of the test are not influenced by external influencers. 

Media Mix Modeling requires continuous calibration for weeks or even months. The calibration process may include setting priors and posteriors, or use GeoLift tests as ways to adjust the results that are produced by the MMM model itself.

INCRMNTAL’s platform enables marketers to begin measuring results immediately upon implementation. This rapid readiness is a significant advantage in fast-paced industries.

The way that the INCRMNTAL platform does this, is by utilizing 12 months of Advertisers’ historical data (provided as part of the integration step), allowing the INCRMNTAL algorithms to scan and pre-learn patterns, scoring, causal distribution, and the causal relationship of campaigns to channels, campaigns to one another, as well as the response curves, diminishing returns, and the impact of certain key special days, holidays, weekdays, weekends, and so on. 

To learn even more about INCRMNTAL’s methodologies, we would advise you to schedule a deep dive demo with us at incrmntal.com/demo

Cost-Effectiveness

The monthly costs of marketing measurement software such as MMM, GeoLift testing platforms, and INCRMTAL may be at a similar range. 

But MMM, and GeoLift testing platforms demand frequent iterations and analysis, therefore, most MMM platforms require a high monthly maintenance costs, similar to GeoLift testing platforms which requires high service charges for test design, and analysis work. 

Cost is a major consideration for marketers, and INCRMNTAL offers an affordable subscription model that minimizes overhead. This accessibility democratizes advanced measurement capabilities for businesses of all sizes.

Revealing Organic Contribution Levels

The notion of Organic results is tricky for most marketing measurement platforms. Orgnic by definition would be the conversions happening NOT thanks to paid marketing results. 

Most marketing measurement platforms can only consider paid marketing as the data used to provide insight, leading to a strange paradox, where Organics need to be ignored. 

MMM’s top-down methodology requires specific knowledge to isolate the impact of external factors. A way to introduce Organics to an MMM model would be to stop all paid marketing activities to reveal the true organic baseline. The Organic level may later be set as a Prior within an MMM model. 

GeoLift tests, also demands a complete halt of marketing activities to assess true organic levels, a scenario that is often impractical.

Understanding organic contribution is crucial for effective marketing attribution. INCRMNTAL’s approach automatically extracts organic-related parameters, providing a clear picture without additional manual effort.

As the modeling process of INCRMNTAL is a bottom up modeling approach, INCRMTNAL starts by only measuring the causal impact of paid marketing, attributing only the results generated by a marketing campaign, in turn the channel. Once summing the results of all paid marketing activities – calculating the organic level is a simple deduction.

Selective Incorporation of External Features

No one likes to admit it – but external variables often affect marketing results more than paid marketing activities. 

We can see many examples for this: 

• The spike in sales for eCommerce and Retail during the holidays
• A spike in loan application for fintech companies during the end of the month
• Lowered engagement and attention for games during certain months of the year
• Mobile gaming seeing an increase in ap installs during weekends
• Ride hailing app seeing higher demand during stormy weather

There are dozens of more examples of how external variables may affect results. But incorporating these to improve your marketing measurement effort may be a challenge. 

Media Mix Modeling can support the inclusion of external variables, however, an MMM will always use a variable in its calculations – even when the variable does not actually impact or has any relations with the marketing results. 

INCRMNTAL selectively incorporates external variables through a feature selection process tailored to specific measurements. This precision ensures that only relevant factors influence the results.

External variables are like a thorn in the flesh for GeoLift experiments. Those are often the main reason why a GeoLift test will product inconclusive results. Attempting to create a test where external conditions do not impact advertising performance is only possible in a laboratory environment.

Conclusion: Why INCRMNTAL Stands Out

When comparing INCRMNTAL, MMM, and GeoLift, the differences are clear. INCRMNTAL’s always-on incrementality measurement methodology provides unmatched flexibility, accuracy, and cost-effectiveness in dynamic environments. Its ability to measure campaigns, adapt to changes, and deliver insights without requiring traditional experiments makes it a versatile tool for modern marketers.

While MMM and GeoLift have their merits, they fall short in key areas such as adaptability, scalability, and affordability. For businesses seeking a reliable and efficient way to optimize their marketing strategies, INCRMNTAL is the clear choice.

Bonus: Here’s a comparison table, showing the methodologies based on various parameters, as compared side by side

.