This report explores the most promising new developments in ROI and ROAS measurement,
including MMM, singlesource, random control trials, MTA (a breed of singlesource which arose
from digital), agent based modeling (ABM – MMM projected down to simulated household level),
and other new methods.
The cross-analysis of media spend and sales data to deduce the contribution to incremental
sales produced by advertising and other marketing stimuli began with what is today known as
Marketing Mix Modeling (and as Media Mix Modeling in cases where non-advertising stimuli are
generally excluded) in 1948.
By the turn of the century, MMM had become firmly established as basic to the toolkit of major
advertisers. It was criticized for the degree of subjective analyst judgment required, weaknesses
in the treatment of the raw data, slowness of results, and the fact that management could not
understand how it worked, among other criticisms. Some advertisers tested alternative suppliers
of MMM and found a good degree of disagreement especially for the newer media types.
Singlesource (actual longitudinal measurement of ad exposure and sales in the same households,
not a feature of MMM) using small (5000 and below) panels briefly arose in the U.S. and a handful
of other countries, but were not economically sustainable, and sample sizes were too small for
most brands to see much statistically significant value.
The invention of big data singlesource (TRA, 2005) led to the use of digital ad tags, set top box
data, smart TV data and other such data sets with sample sizes in the millions of households to
be used with same-household purchase data by hundreds of companies, diminishing reliance
on MMM to a degree. With the rise in privacy laws and announcement of the deprecation of
third party cookies, advertisers refocused on improving MMM which would not be affected by
these forces.
As might be expected, artificial intelligence (AI) is already playing an important role in many of
these innovations. New evidence suggests that the most sophisticated practitioners will be using
a combination of these methods, including ABM simulated populations on which synthetic trials
may be performed, and that more subtle factors such as creative, context effects, media
saturation effects, targeting, and synergies will play a larger role in the future of outcomes
measurement and optimization of short and long term effects of marketing.
Whereas advanced forms of MMM will continue to help make better decisions about budget
allocation to broad media types vs. other types of marketing stimuli, singlesource systems will
continue to help make faster inflight decisions about phenomena not well reflected in most MMM
today, such as creative, target groups, context effects, attention/resonance, frequency, recency,
synergy effects and other leverageable variables which can make major contributions to overall
ROI/ROAS.
Random Control Trials (RCTs) and other forms of testing/experimentation in market continue
to be the gold standard and the truth standard for the modeling-based methods, and the
practical difficulties of RCTs could be assuaged by the leadership of media companies, in which
addressable commercials would be the essential enabler.
