2. SPECIAL REPORT
Real World MMM
Volume & Marketing Activity As a practical matter, one of the hardest as-
pects of running marketing mix models is
getting the right inputs upfront. (The devil is
Volume in the details.) To be worthwhile, the inputs
Merch Support need to be accurate, and must include the
in-market start and end dates for all events,
the specific brands and SKUs participating,
and the final version of the tactics used.
While not difficult to understand, such ba-
sic information can be hard to marshal and
requires contributions from the brand team,
Select members of Nielsen’s household panel other activity. The landscape portion is the ad agency, shopper marketing agency and
scanned all of their purchases and wore the passage of time from left to right. finance group.
meter, which attempted to record a wide To date, this statistical method of looking Furthermore, it is difficult for many manu-
range of media exposure, from radio spots, at sales revenue holds sway in the industry. facturers to tease out the costs and impact
to outdoor billboards, to in-store communica- The practice is fairly inexpensive, easy to un- of trade promotion for a specific brand. This
tions. However, the cost for a national rollout derstand, actionable and – statistically, at means that the relative effect of trade promo-
of the methodology was far too steep, even least – bridges the gap between in-store and tion versus shopper marketing can be hard
for the project’s seven prominent backers out-of-store programs. In absence of perfect to gauge.
(one of which was P&G). causal information, marketing mix analysis is Ken Madden, executive vice president of
the next best thing. However, there are issues strategy & analytics at OgilvyAction, is one
Good Enough: Marketing Mix with this valuable technique. practitioner making headway in this area.
Models Marketing mix modeling itself has an under- Madden described the recent analysis of a
Although such efforts to monitor and mea- lying fissure: It essentially is a forensic exercise, shopper marketing program and how his
sure the shopper’s exposure to marketing on with sales being the examined “evidence.”
Lump Profit Figures
team responded to the issues.
a 360-degree basis stumbled during the first Cause and effect dynamics are actually sta- “When the analytics group broke out profit
decade of the new millennium, marketers tistical correlations that attribute portions of and ROI by trade factors versus marketing
were not without tools: marketing mix mod- sales to specific advertising or promotional factors, it seemed clear that the program’s
eling had been a well-established practice for activity. But in traditional MMM, these cor- success was being driven by trade [see chart,
most CPGs for more than 20 years, and one relations are not tracked on a “per shopper” below left]. However, the brand team felt that
still used widely today. basis. Therefore, it is not certain if the same the marketing program was responsible for
Marketing mix modeling (MMM) is a sta- shopper who saw an ad on television subse- a lot of the trade success. They said that the
tistical tool that uses multivariate regression quently redeemed a coupon, for example. trade support was above CRM Profit
average because
Trade Profit
analysis to understand the relationship of var- Additionally, since statistical models use the retailer responded very positively to the
ROI: 2.14 1.61
ious marketing efforts (in- and out-of-store) sales as the measurement, other program marketing plan.
to sales. Regression analysis uses an indepen- effects aren’t captured, such as the impact “To examine this, we gathered historical
dent variable to help predict a “response vari- on brand attitudes or shopper behavior benchmarks from similar promotions at the
able” (in this case, sales). The “multivariate” (measurement of which was examined in same retailer. We modeled typical trade sup-
part refers to the presence of several “predic- the second and third articles in this series. port against typical sales lift [and] found that
tor” variables – media advertising, temporary Visit www.p2pi.org). 65% of the trade profit could be attributed to
price reductions, and so forth – that can influ-
ence sales.
The output of MMM usually looks like a
geological cross-section of a mountain land- Lump Profit Figures Profit Breakdown
scape (see chart, above). Each stratum in the
Trade
cross-section shows the amount of sales that
profit from
are attributable to a particular marketing marketing
65%
tactic. The layers are typically broken down “synergy”
into “baseline sales,” which are located at
the bottom of the mountain and presum-
ably would occur without any advertising or 35% 100%
promotion (but have been shown to erode
over time if a brand is starved of all sup- Trade Profit CRM Profit Trade Profit CRM Profit
port); and the upper-layer incremental sales ROI: 2.14 1.61 ROI: 2.14 1.61
that result from advertising, promotion, and Source: OgilvyAction
2
3. SPECIAL REPORT
Trade vs. Shopper
Rachael Norton, vice president, shopper soccer team while on the field or after the
marketing, at ConAgra Foods, discusses game. This is a program that addresses an
the relative impact of trade vs. shopper unmet need. The “solve” is for the shopper,
marketing events based on her extensive the retailer and the brand. It is a triple win.
experience using custom marketing mix
models. Q. What if we got away from the analysis
and just used common sense to develop
Q. What kinds of programs have been and run programs?
your best ROI generators? Norton: We have used this as a starting
Norton: A straight trade event is usually point for some ideas. For example, we
the best ROI. The sheer impact on ROI can know that some of our shoppers’ share of
be less with a targeted program than with a stomach goes to [quick-service restaurants].
hot price point. [But the shopper program’s] McDonald’s offers a meal for one or two
impact on a specific target can be huge. And bucks. So to compete in a practical way,
that is what it’s about: long-term behavioral ConAgra created displays that featured
change and meeting shopper needs. It’s like “The $2 Meal” or “The $3 Meal.” This is a
micro-marketing on a mass scale. For exam- common-sense approach to appealing to
ple, we offer moms a meal solution for the shoppers.
the above-average trade support [see chart, generate buyer profiles. McMahon used this age of shopper moms on SNAP, the in-store
bottom right of page 2]. We call this the ‘Mar- example to explain the need for decomposing messaging might center on stretching meals,
keting Synergy Effect,’ when the retailer’s sales volume: for example.
enthusiasm for a particular program drives “Many CPG brand teams would be sur- “Additionally, these folks aren’t reading
extra trade support and thereby boosts per- prised to find out who is actually buying their newspapers, because they aren’t buying
formance.” products. If you decompose your volume – newspapers. Therefore, FSIs dropped in the
now versus four years ago, say – you might local paper won’t be redeemed by these
No Absolute Answers find that a sizable portion of your current shoppers. The list of such implications goes
Using a statistical method to look at the re- buyers are SNAP [Supplemental Nutrition As- on and on. It is imperative to know who the
turn on a marketing investment can be de- sistance Program] users [because that] now shopper is.”
ceptive if the data is interpreted incorrectly. includes families with incomes greater than
According to Mike McMahon, president of $50,000 and three kids. Once a Year, Needed or Not
research house Spire, Monroe, Conn., and “SNAP families have government-issued Usually, marketing mix models are run an-
former senior vice president of marketing at money during the first two weeks of the nually, just before the next year’s business
ConAgra Foods, “Many people look at the month. People plan their shopping trips plan. The retrospective analysis shows the
ROI as an absolute. This is a mistake. It is a around this cycle. If a brand knew its percent- impact of various investments over the year
measurement at a point in
time. This is a number to act
on, to improve upon, [but] not
an absolute.”
Marketing mix models are
a backward-looking analysis “Many people look at the ROI as
of activity. While some are
also used for scenario build- an absolute. This is a mistake.
ing, models are not accurately
predictive enough to be used It is a measurement at a point
in profit & loss statements
with assurance. Instead, the in time.”
numbers should be used di-
rectionally. Mike McMahon, president, Spire
One obvious drawback to a
typical MMM is that it doesn’t
3
4. SPECIAL REPORT
Alignment of Shopper and Brand Measures
Strategic Brand Retailer-Level
Goals Objectives Shopper Measures Brand-Level Measures
Penetration
TRIAL Switching Shopper card data National panel data
New category users
Frequency
Units per purchase
REPEAT Shopper card data National panel data
Usage rate
Loyalty percentage
Pre/post-shop interviews;
store traffic counts; Longitudinal studies; pre/post
Awareness
distribution/execution audits, advertising studies
eye-tracking
Shopalongs; mobile panel studies;
UNIVERSAL Consideration
pre/post interviews
Annual, longitudinal brand studies
In-store video;
Navigation NA
shopping cart tracking
Brand Imagery Interviews Longitudinal brand studies
on a national basis. However, a typical large information can help marketers change direc- recommends conducting a quarterly or even
national model doesn’t capture most shopper tion mid-year to increase the effectiveness of monthly analysis, so that ROI results can be
marketing efforts, such as regional programs programming. shared with teams on a regular basis and pro-
or short-term, single-retailer programs. To develop a retailer-based, custom mar- grams can be improved “on the fly.” If that’s
But, brands that have developed propri- keting mix model, a brand and its agencies a bit too ambitious, gathering the information
etary, custom models at the retailer level are sometimes pool their resources of talent and at least can help marketers make adjustments
able to measure ROI against specific programs data. At least two years of scan data are that affect the current year instead of waiting
to fill in knowledge gaps. Developing a model needed, along with the dates and details of until the next to make improvements. Brands
that uses historical brand and retailer-level advertising and retail-level promotions. Solid that use a measure, learn, change approach
data is the best way to unlock the power of quantitative skill and experience is needed in a disciplined way also can keep pace with
this technique, according to Agustin De Dios, as well. Just as often, development is out- retailers, which continually optimize their pro-
director-global analytics at Kimberly-Clark. sourced to a service provider – who still needs gramming, he suggests.
(To hear more of De Dios’ views on adapting data inputs from the client. In whatever way What’s more, this approach can help jus-
MMM for shopper marketing, visit www. the smaller model is built, the granular ap- tify spending on shopper marketing fairly
p2pi.org.) proach is the best way to ensure effective use quickly. According to Rachael Norton, vice
As importantly, the usefulness of the model of MMM for shopper marketing programs. president of shopper marketing at ConAgra
can dramatically improve if the analysis is run McMahon, who developed custom MMMs Foods, the company shifted marketing dol-
more often than annually and at a retailer- for ConAgra, believes that best practices take lars from above-the-line activities to shopper
specific level. Obtaining quick and accurate a “measure, learn, change” approach. He programs because the transparency provided
4
5. SPECIAL REPORT
by frequent ROI reporting brought credibility
to the discipline across the organization.
the impact on the larger marketing objective.
One simple example is to report the number
About the Author
of coupons redeemed without understanding
Aligning Brand and Shopper Goals whether the effort rewarded current users,
As has been stressed throughout this series, induced trial among new users, or converted
sales aren’t the only measure of success for switchers. Another is to track the number of
shopper marketing programs. Achievement “shares” online without knowing whether
of other objectives is often at least as impor- they drove conversions or awareness. Savvy
tant. In shopper marketing, most objectives marketers keep critical brand objectives front
can be expressed in terms of shifts in shopper and center when evaluating program perfor-
behavior. This can include brand switching, mance.
alterations in purchase or occasion frequency, Aligning the objectives of above- and be-
or changes in list making. low-the-line efforts doesn’t need to be overly
But marketers are challenged when they complicated. The chart on page 40 is a simpli-
attempt to measure shopper performance fied model identifying how they might line up Liz Crawford has more than 20 years
against overall brand objectives. Perhaps the and where the data could be sourced. of brand management and consulting
biggest stumbling block is the temptation to The shopper card data acts like a kind of experience with a concentration in
use tactical measurements and thereby miss mini-panel of shoppers for a particular retail- strategic innovation. Over the last
er. This smaller panel might generate data us- few years, Crawford has focused
ing swiped frequent-shopper cards or mobile on developing integrated shopper
devices – handheld devices or smartphones marketing strategies for Fortune 500
Series Schedule with apps, or even retailer apps. clients. Currently, Crawford is an
While this sort of alignment isn’t perfect, it analyst and contributing writer for the
Part 1: Rationalizing the can at least help keep the brand on track to Path to Purchase Institute. McGraw-
Investment attain success benchmarks (assuming they’ve Hill released her book, “The Shopper
been clearly articulated at the outset). Economy,” in March.
Part 2: Measurement of
Shopper Behavior Next: Single Source Input
The next step should be a big boost to mar- JWT/OgilvyAction Inc., conducting
Part 3: Measurement of Brand keting mix modeling: single source data and business under the OgilvyAction and
Impact cross-media tracking. As discussed in the sec- JWT Action brands, is a fully integrated,
ond article of this series, single source data end-to-end shopper marketing and ex-
Part 4: Effective Integration refers to tagging a specific shopper’s activities periential marketing agency with main
Practices across media platforms and retailers. Adding offices in New York, Chicago and Akron,
this data to MMM delivers more real-world Ohio. It is part of the WPP Group.
Part 5: Retail Collaboration tracking and fewer forensics. Marketers will
begin to see relationships between shoppers
Part 6: Directions for the and programs like never before. This concept
Future will be discussed in detail in the final article of
the series.
5