Author: Kevin Youts

Breaking Down Silos to Unlock the Full Potential of Retail Media Networks

Posted on by Kevin Youts

Introduction: In today’s rapidly evolving marketing landscape, Retail Media Networks (RMNs) have emerged as a critical component of marketers’ media strategies. According to a recent article in Insider Intelligence, RMNs have grown exponentially across an exploding number of retailers, becoming a $30Billion channel in just 5 years.

With the deprecation of first-party data and the need for real-time insights into shopper behavior, RMNs offer a tantalizing antidote to both trade marketers and media buyers. However, to harness the full potential of these networks, brands must avoid the pitfalls of organizational silos, and provide the resulting teams with the optimal tools to make win/win decisions retail media decision that benefit the company overall.

Collaboration between brand and trade marketers and category managers, alongside media planners and buyers, can be crucial for maximizing return on investment (ROI) in RMNs and achieving both trade marketing and media effectiveness. Too often the two groups are functioning with separate budgetary parameters or not working off the same models to drive the most effective media buying decisions.

Let’s look at what each group brings to the challenge:

Brand and Trade Marketers:

Brand marketers, including shopper marketers, category managers and omnichannel leads play a crucial role managing sales, and coordinating merchandising and promotions with distribution partners at retailers. Their skill sets revolve around understanding consumer behavior, market trends, and effectively positioning products within both the physical and online retail environments. They typically possess the following key competencies:

• Retail Strategy: Trade marketers and brand category managers have a deep understanding of the retail landscapes in which their products are sold. They develop strategies to maximize product visibility, optimize shelf space allocation, and ensure effective pricing and promotion tactics are negotiated and well executed across an array of retail partners, increasingly in both physical and online environments.
• Merchandising Expertise: These professionals excel in creating compelling product displays and planograms that capture shoppers’ attention and drive sales. They have an eye for visual aesthetics and possess strong knowledge of consumer purchasing behavior.
• Channel Collaboration: Trade marketers and omnichannel leads and brand category managers establish strong partnerships with retailers, negotiating slotting fees, and ensuring optimal product placement. They possess exceptional communication and negotiation skills to drive mutual success.

Digital Media Planners and Buyers:

Media planners and buyers specializing in digital opportunities, play a pivotal role in targeting specific audiences and executing programmatic buying across relevant channels. Their expertise lies in leveraging digital platforms and various data analytics tactics to deliver targeted messages to the right audiences that can generate measurable outcomes. The following core competencies define their skill sets:

• Data-Driven Insights: Digital media planners and buyers are adept at analyzing consumer data and identifying targets and how best to reach them. They leverage sophisticated tools and technologies to extract actionable insights, enabling precise audience targeting and effective campaign optimization.
• Multi-Channel Mastery: These professionals have a deep understanding of various digital channels, as well as offline media, that in combination make up the optimal media mix. Their core competency lies in balancing expenditures and often dynamic investments across channels including social media, search engine advertising, display advertising, and video advertising. They are tasked with crafting integrated media strategies that reach consumers at multiple touchpoints.
• Programmatic Proficiency: Digital media planners and buyers excel in utilizing programmatic buying to automate media purchases and optimize ad placements in real-time. They possess a strong grasp of ad exchanges, demand-side platforms (DSPs), and supply-side platforms (SSPs). They lean on direction that comes from robust data analysis to retrospectively calculate ROAS, and proactively optimize their media plans to achieve optimal growth in the future.

The Pitfalls of Organizational Silos

Failure to bridge the gap between these two groups can lead to several detrimental outcomes. Budget disputes often arise when teams compete over limited resources, hindering the overall marketing effectiveness. Moreover, the absence of collaboration can result in one team allocating budgets without considering the expertise and domain knowledge of the other, leading to suboptimal decision-making and wasted opportunities.

Harnessing the Power of MMM for Optimal ROI

Retail Media Networks, especially if brands are investing in several, produce a maddeningly complex set of data to analyze (time, geo-spatial data based on physical location of stores, consumer behavior on and offline, and the staggering array of products many sales and media teams must account for).

There is renewed enthusiasm across the media analytics industry in the power of robust Market Mix Models (MMM) to make sense of increasingly complicated media planning required to optimize RMN spend as part of overall advertising strategies.

The best MMMs enable marketers to analyze behavioral, historical sales, and other causal data at the individual channel partner level across all media. With this granular understanding, organizations can identify the most effective tactics for driving sales within retail settings, analyzing overall media spend as well as the relative contribution of each media partner, including RMNs, in relation to each other.

Conclusion:

As Retail Media Networks assume greater significance in marketers’ media partner choices, organizations must dismantle the barriers of organizational silos to achieve optimal ROI.

By embracing collaboration between brand marketing and media management and providing them with effective tools like MMM, advertisers can leverage the strengths of both groups to make better-informed decisions that drive sales growth effectively.

By embracing these strategies, organizations can seize the opportunities presented by RMNs and unlock their true potential in today’s dynamic retail landscape.

Advertising at the Last Possible Moment

Posted on by Kevin Youts

Byron Sharp makes a good case in this article here that search is more like shelf-space than it is like advertising and applies it beyond the typical discussions around paid search to the growing domain of retail media. He is right that we see a growing sense among our clients that retail media is becoming something you have to do and no longer optional. Retail media budgets being protected by the sales division; they are becoming sacrosanct. That seems very much like slotting fees – like “rent” which means all downside if you don’t have it and not a whole lot of upside if you do.

Let’s remind ourselves what it means for something to be operating like shelf-space or distribution. We expect distribution elasticities to be near 1, i.e., a 1% increase in distribution results in a 1% increase in volume. This article suggests that, in the not-too-distant future, we should expect the same from retail media. In fact, the article suggests a kind of parallel e-commerce universe where search functions like physical distribution and retail media displays function like physical in-store displays. Both operate within the fairly narrow confines of people who have either already decided to buy your product or category and just need a little nudge over the finish line (search), or those who already have your brand in their consideration set and just need a little reminder that you still exist and are relevant (display). This would argue that retail media is on its way to becoming a promotion variable like traditional trade and should be treated like trade in an MMM model.

Could online search and display ever function as something more? It seems certainly true that retail media is operating at the bottom of the funnel. But just like the world of in-store promotions, there are a lot of ways that the marketer might be influencing and changing behavior even at that last point. A key factor here is “has the decision already been made to buy in the category for the occasion or motive?” At one extreme, if we look at data from the restaurant/QSR category, the decision has already been made prior to the search, and the consumer is using search to find out where the nearest store is or to check the menu listings or prices. In this case, when we look at modeling search, we see high correlations between search and sales, but we can’t say that search caused those sales. It’s more likely that the search happened as a consequence of already making the decision to buy. In order for this variable not to have too much spurious power in the MMM model, we have to transform search to a metric like “YoY change in search” to attempt to remove the endogeneity.

In CPG, with its relatively low risk and large variety, a lot of fine-tuning behaviors are possible prior to the purchase which opens up a wider range of possibilities for the role of search. You may have decided to buy in the category, but not yet chosen the brand; you may have a strong preference for the brand but do not think of it for this occasion. Or you may be totally unaware of the product – like a new product – and encounter it for the first time on the shelf or on the site. This means that there may be a spectrum where search operates beyond simple distribution by nudging you in a direction you might not have gone. It is part of a final information-gathering or choice-confirming step prior to purchase. From a marketing point of view, it is helping you sort out the connection between what is available in the market and the fit of that to the motive in the purchase occasion. In that admittedly narrow spectrum, there is still room for some creativity and influence.

Incidentally, the belief in even a small spectrum of ambiguity in search is why we prefer to use impressions rather than clicks when we model. Many agencies want us to use clicks rather than impressions because that is “what the client pays for.” However, we prefer impressions precisely because it takes a step back from the point of purchase to the position where you might be “nudged.” The lift in a model then captures the tactic of nudging rather than treating it purely like distribution.

Search is not the only tactic that is claiming to be synonymous with distribution and that needs to be “always on.” There are a number of third-party channels and services emerging who are carving out their own specific routes to unique distribution (think Cardlytics, Connexity, DoorDash, or even Instacart); some of them don’t track impressions at all and claim to have isolated their own impact perfectly for purposes of ROAS. But there is always an awareness or behavior that comes before using an app, and a surprising amount of wiggle room to suddenly buy something different even the lowest part of the funnel.

The standard method of determining price sensitivity is flawed! But there is a better way.

Posted on by Kevin Youts

It is always tough when we discover that the gold standard for decision support in any business challenge needs an update, if not total rethinking. Such is the case with the broadly accepted “best practice” approach to determining price elasticity and the optimization of the products, features, and prices of almost all products and services.

For decades, companies have depended on choice analysis to identify the features and attributes, the tradeoffs that consumers will make, and the price parameters they will accept when considering a purchase. Choice-based conjoint (CBC) analysis ascended from its inception in the early 1980s to wide acceptance as the preferred methodology for pricing guidance, eventually with adoption of software that in theory makes the process more automated.

As with most more complex analyses that depend on higher level math, tools like CBC are best applied by the experts who understand not only how the algorithms work, but also the basic behavioral and economic science that is foundational to what the analysis is uncovering.

As one of the innovators who were part of the introduction of CBC, we are huge proponents of the elegance and utility of CBC to help marketers understand the drivers of choice. We know that psychologically, consumers assess product purchases based on complex, personal, stated and unstated decision rules, which CBC is perfectly positioned to identify and predict their impact on price.

We have been at this for decades. At some point, we have to take a hard look at the limits of CBC as it is typically implemented. And for the sake of the industry, we are forced to say that we can do better.

So what is the problem?  And how can we fix it? 

The classic approach to CBC looks at preferences, but ideally you want it to also reflect the constraints that may define the preferences.

Time can be a constraint that impacts price. If I need a product or service within a specific time frame, I might be willing to pay more to obtain it. Or if time is not as pressing, a lower price may be the key attribute that will influence a purchase. We see this principle every day in the choice of shipping options in online commerce.

But beyond preferences, the constraint of a money budget imposes a consideration that may be an even more powerful motivator/demotivator to making a purchase choice. I may want a phone with extended battery life, or a car with luxury accessories, or I may want the luxurious and expensive skin moisturizer. But my money budget constraints, relative to all my other payment obligations, may make those unacceptable choices. Traditional CBC is not structured to accommodate the money budget as an attribute, and its omission distorts the size of those factors that are impacting buyer choice behavior.

Using a money budget in a choice model yields estimates of price sensitivity that are dramatically lower than the price sensitivity estimates found in choice analyses that do not employ a budget. This guarantees drastically different implications for price setting.

How does it work?

Budget Constrained CBC assumes that each person has a fixed budget for purchasing in a product category.

  • If a purchase option costs less than their budget, it will be considered and the product may or may not be bought, depending upon the benefits it provides.
  • However, if an option costs more than their budget, it will never be bought.

Let’s look at an example. These are the price sensitivity coefficients estimated in a study of purchasing a consumer durable. We studied choice behavior using CBCA with about 500 people who were “in the market” for a new model of this product.

We estimated a price sensitivity value for each person using the standard approach and also the budget constrained method. We can look at the percentiles of the distribution of price sensitivity across all people. This table shows the results.

The table shows the results of the percentiles of the distribution of price sensitivity across all people.

The price sensitivity for the budget approach is “flatter” than for the standard model and is roughly half the size (-.491 versus -1.080).

The table shows that:

  • All price sensitivities are negative, as they should be under economic theory.
  • Mean price sensitivity under the standard approach is more than twice as large as the sensitivity under the budget constraint.
  • The variability of the price sensitivity under the standard approach is about 50% more variable than under the budget constraint.

By ignoring the buyer’s budget, traditionally applied CBC will yield results that tell you that people will spend more than they actually will do and it tells you that people are much more diverse and variable in their estimated price sensitivity.

Building the Better CBC Model

The defect inherent in the gold standard of is corrected by estimating the buyer’s budget and including it explicitly in CBC.

Budget Constrained CBC overhauls this fundamental flaw and delivers more accurate estimate of willingness-to-pay, better fits to raw data, and deeper insights into buyer behavior.

Seems logical to you? We would love to prove it to you that it works. If you have an old CBC study for which you have the data and would like to see the results delivered by the constrained budget approach, contact steve@in4ins.com to discuss.

Attribution Analysis Letting You Down? Why Not Return to MMM?

Posted on by Kevin Youts

Marketing attribution analysis was great in theory. What are the things consumers see and hear that influence and prompt them to act, specifically to buy a product or service?

It would be nice if it were simple and binary: Consumers see this, they buy that.

But let’s start with the essential data and the connections brands need to make, to support decisions that impact media planning effectiveness.

Visibility into what a consumer bought: Possible in digital (if brands own their own data or can get it easily from e-commerce distribution partners.) Harder in physical stores. In both cases brands need first party data from the retailer to know what was actually purchased.

Visibility into what the consumer did to get to the point of purchase: Did they go to a retailer’s website and search for the product? Did they see a digital ad, click on it, and it took them to an e-commerce page? Did they see the product on their neighbor’s kitchen counter and make a mental note to purchase it next time they were in the grocery store?

Visibility into what advertising campaign was seen or heard before the purchase: Was it the digital ad on their favorite app? Was it the TV ad, or the OOH ad on a bus shelter? Was it an audio ad on a podcast or on the radio during their commute from the office? If it is digital media and a digital purchase with accurate UTMs embedded and an effective tracker to see what messaging converted, an advertiser may have half a shot of making a connection of the channel and the campaign to the actual transaction. But it is not a precise media planning tool, particularly if the campaign intent was brand awareness, versus activation.

Connecting all these data points to create optimal media plans

Here’s the dilemma. Attribution modeling has been described as limited, in part because it may account for one or two online channels compared to each other, but in terms of all other media and marketing exposure it can’t definitively say which is THE ONE that can be attributed to the purchase. What is more likely is a combination of factors in aggregate, in context of other behavioral conditions, drove the consumer to put the product in their virtual or physical carts.

Folks at the digital marketing agency Atomicdust back in 2021 wrote an interesting piece provocatively entitled “Marketing Attribution is Killing Common Sense”.

Still quite relevant, it contains some of the better definitions we have seen on the types of attribution (“last interaction”, “last non-direct interaction”, “linear attribution”, “multi-touch”, etc.) commonly tracked in attribution modeling. More importantly, the author basically describes how each of these models as flawed in their own way.

Why are they flawed? Two reasons, actually.

  • Attribution models are primarily limited to digital marketing, which is highly dependent on the infamous cookies that are going away due to privacy concerns.
  • These models don’t reliably accommodate the attribution of sales that are conducted in physical retail settings, in which case an advertiser may have no idea what was the last (or first) ad the shopper saw.

Frankly consumer behaviors are sufficiently complicated as to defy predictions based purely on the last or first impression of the product. Everything from seasonality and weather to time of day (particularly in the case of impulse buys at quick service restaurants or convenience stores) will be influenced by the confluence of when and how an advertising message is delivered in any given media channel.

So, if not attribution modeling, what can brands employ that is more evidenced based than “common sense”, to make better media buying decisions?

The Return to a Trusted Tool: Marketing Mix Modeling

FMCG companies have relied on Marketing Mix Modeling since the 1960s. Of course, retail options and even the media channels available to advertisers were far less complicated then. Decisions on how, where and when to plan media spend could be done at a comparatively more leisurely pace than the frenetic programmatically driven speed required in today’s digital environment.

But the beauty of MMM is its inherent data analytical sophistication, capable of incredibly complicated models with far more degrees of granularity, is that it can tackle the combination of many new kinds of data feeds at different levels of specificity (target audience, time granularity, geographic unit, type of metric, etc.) And this level of complexity requires data analytics robust enough to parse the results at a much more sophisticated level to much better media planning support.

As retail has also evolved to include far more distribution outlets (including e-commerce) and media followed suit with the introduction of digital advertising, MMM provides the perfect solution to making highly complex decisions about media buys. But you need accomplished data scientists running the models, equipped with the expertise to interrogate the algorithms with a brand’s specific objectives in mind.

  • MMM when done right, can provide far more detailed road maps for achieving ROAS (return on advertising spend) encompassing virtually endless variables that can impact spending.
  • MMM done right can provide guidelines for spending just enough in any one media channel so as not to overspend in terms of overall return on investment.
  • And we could (and will) dedicate and entire post to the unique benefit of MMM to facilitate targeting not only the many more channels in the new world of media used by people, but the situation and the journeys that got them there. Attribution never really fulfilled its promise to get a whole lot closer to this level of targeting visibility.

Brands are moving away from strictly asking themselves: “Should we spend less on TV and more in digital?”, to the much more relevant question: “What’s the role of the big-screen in my campaign and which new channels are going to replace the ubiquity of TV?” To guide the resulting media planning choices, the best MMM provider/partner will actually surface the more discrete evidence as decision support. The ideal MMM partner will reveal which of all the digital and other channels available will yield the best contributions to ROAS compared to each other, so media buying decisions can be more fine-tuned and effective.

Attribution may be letting you down. But have no fear. Our data analyst teams at in4mation insights are prepared to help you tackle media planning, delivering optimized scenarios to achieve return on your investments and ongoing growth for your brands.

Marketing Mix Modeling Retakes its Role in Media Optimization

Posted on by Kevin Youts

Mark Garratt, Partner and Co-Founder at in4mation insights

We are at an interesting inflection point in the world of advertising media, for sure. Since the days of media being dominated by broadcast advertising (remember those voices shouting that “TV is KING!” in the early days of digital? How quaint that sounds today!) through to the moment in time when the act of buying media was disrupted by programmatic and all manner of tech enabled deployment, the most common descriptor we heard of the media landscape was that it was becoming “fragmented”.

From where we sit, we have seen factors for building the optimal media plan as increasingly complicated and opaque. This has always made it challenging for brands to make informed decisions on how and to whom to allocate your media budget to get the best return.

It is also important to note that for a while the gold standard for managing media spend decisions, marketing mix modeling (MMM), seemed to take a back seat to the elusive but tantalizing promises of multi-touch attribution alone as the holy grail for measuring media effectiveness. Despite perhaps not being in vogue, we stuck to MMM because of what we knew in our gut was the better solution overall IF the challenge was complexity and multi-dimensional influence.

We knew there were no short cuts. So we have spent over 20 years building, refining and applying our core competency for our clients in a technical approach to MMM that leverages the power of robust computational modeling tools (for exactness it is called Hierarchical Bayesian analytics, but that is a subject for another post). It has been proven to deliver better, more granular and flexible results to measure and optimize media spend for optimal return.

It was therefore gratifying to us when the ARF dedicated an entire track in their upcoming Attribution Accelerator conference to the topic: The Renaissance in Marketing Mix Modeling: Meeting New Expectations.

We will be presenting there with our client from OLLY, specifically on how the right kind of MMM fueled by a robust Bayesian approach can solve for the most complex combination of factors, and fragmented data.

Why a renaissance, you might ask, and why are there new expectations?

If anything, the fragmentation in the media industry has accelerated exponentially since the days when the sword fights over media effectiveness were confined to the battle over TV versus the nascent category loosely defined as Digital. Today’s media plan is distributed across hundreds if not thousands of potential touchpoints that involve choices across channels that didn’t exist 5 years ago: TikTok, Instagram, Podcasts, DOOH, and now the exploding sector that is referred to as Retail Media.

MTA as a standalone as originally envisioned has yet to really deliver on its promise, and its future reality is made all the more challenging with the pending demise of cookies and increasingly technical and regulatory focus on privacy. With the right integration into MMM, (delivered as quickly and cost effectively as we have proved is possible with our own solutions) has the potential in the future to be game changing, but we need both to be functioning optimally in a world of much more complex expectations. We are developing just such an integration.

The “perfect storm” that is also brewing, of course, is the added specter of brands reducing marketing spend in the face of inflationary factors leading to an economic downturn.

Increasingly the C-Suite is challenging all business teams to do more with less, and marketing is not immune to this directive.

What we at in4mation insights are pursuing with our clients is the objective of “doing the best with less”, specifically within their media budgets that were being stretched thin even before the drumbeat of reduced marketing spend! How do we do this? By providing robust evidence-based MMM analytics that reveal the optimal contributions of all media channels no matter how fragmented and complex their interactions might be.

Is our MMM approach new? We would be more likely to say that it is a product of years of investment, innovation, and validation to make media return on investment more predictable and attainable no matter the complexity of the factors. If that constitutes a renaissance, we are happy to guide our clients out of the Dark Ages!

Contact us at info@in4ins.com to talk about how our solutions can help you apply the best of MMM techniques to the increasingly complicated world of media spend management.

What flavor of Bayes is your MMM provider using?

Posted on by Kevin Youts

What flavor of Bayes is your MMM provider using?
MARK GARRATT

Synopsis: In the last 20 years, Bayesian Analysis methods have become the preferred tool for predictive analytical models. With a few exceptions, almost every Marketing Mix Modeling (MMM) vendor says: “Yes, we use Bayes.” Under the hood, however, there are a lot of different versions of Bayes, and they produce different results. These differences are big enough to warrant changes to decisions made by media managers and trade specialists especially as media and distribution channels are becoming more fragmented. Choosing an MMM partner who uses the best form of Bayes will guarantee you more accurate analysis and help you make better decisions.

Fully Bayesian methods outperform alternatives

Many MMM vendors say they use Bayesian methods to calculate your optimal media mix. But not all Bayesian methods are created equal. When evaluating potential partners for MMM it is good to understand what really matters in Bayes to achieve actionable, reliable media optimization solutions.

Some vendors who say they are doing “Bayes” are really using a method developed in the 1960s called Empirical Bayes. This is not a fully Bayesian model. Both Empirical Bayes and a fully Bayesian model (often referred to as Hierarchical Bayes) deliver estimates of sales drivers at a detailed level. For instance, both are capable of estimating separate advertising and promotion coefficients, which get translated into sales lifts for individual DMAs, separate UPCs, and so on.

The difference is that fully Bayesian methods also estimate, understand, and gainfully use the uncertainty in those coefficients.

What it means to understand uncertainty

The next bit is technical, but important because it illustrates why understanding uncertainty matters.

Let’s say an advertiser needs to determine the impact of simultaneously investing in advertising and in-store promotion – essentially a period of feature & display. For illustration, consider our analysis of 2000 UPCs sold in supermarkets from 2019 to 2021.  The charts show the impact of a feature and display (F&D) for 516 of the 2000 products that executed this tactic in that period.

Empirical Bayes and Hierarchical Bayes

In the top chart, using Empirical Bayes, we show that we estimated the average F&D coefficient at .55 and then translated it into an average sales lift of 73%.  You will also note a tight distribution of coefficients around the average of .55, with the vast majority of F&D coefficients ranging from .40 to .75.

In contrast, in the bottom chart, Hierarchical Bayes gives a lift of about 28%, which corresponds to the average F&D coefficient of .26.  You will also note that the Hierarchical Bayesian model shows that this effect is more dispersed over the 516 products.

How do we know which estimate is correct?

Two facts support the second analysis. First, the fit of the Hierarchical Bayesian approach is better, as measured by a smaller mean absolute % error (MAPE). But second, if we take the 516 UPCs that executed this promotion and run a regression on these products one at a time, the median value of the lift coefficients is 0.26, which is the same as the Hierarchical Bayesian result.

Why does this matter?

There is an enormous difference between the advice for action that would be delivered from the two approaches.  The key recommendations that each would produce would lead marketers to decidedly different courses of action.

What advice would you get from Empirical Bayes?

“The lift from combining feature and display is 73% over base, and we are sure of that. About 80% of all the promotions fall between 57% lift and 82% lift. You should keep doing these, with great confidence in future results!”

The superior advice you would get from a Hierarchical Bayesian approach.

“The lift from combining feature and display is 28% over base. But there is a lot of variation across UPCs. About a third of UPCs have very low lifts.  While we investigate further, you should consider stopping F&D with those. About half of the UPCs perform near the average of 28% lift. If it is profitable to continue these, then do so. And, notably, about twenty percent of the F&Ds have lifts averaging about 150%. Let us look into those top performers and figure out why they are doing so well, so that we can duplicate their success with other products.”

We cover the reasons for this in expansive detail in a [white paper] demonstrating the science behind Hierarchical Bayesian methods.  We show how they consistently outperform other flavors of Bayes, both in the overall size of the effects that drive marketing KPIs and the richer detail they produce on the true variation of those effects across products and geographies.

What is the in4mation insights advantage?

As the marketing world becomes increasingly complex and fragmented, the need for reliable, actionable guidance for optimizing media and marketing spend will only grow.

The behavior of analytic models with intermittent and intricately mapped media and promotion variables, both online and offline, demands a very different approach than MMM models developed in the 1980s and 1990s.  You should not expect these models, originally developed in a world of continuous, national-level streams, like adstocked national TV, to continue to perform well when the number of media inputs is exploding (all digital and social channels, and retail media, as examples) and their nature is drastically changing. Only a fully realized and sophisticated analytic tool like Hierarchical Bayes in the hands of a skilled MMM partner can help you understand the new world of marketing.

If you would like to discuss how Hierarchical Bayes can make a difference in optimizing your unique media and marketing mix, reach out to us for a conversation at info@in4ins.com.

About Mark Garratt 

Mark Garratt is a partner and co-founder of in4mation insights. He is an accomplished analytics professional with a distinguished career in both business and academia. Mark has been a trusted advisor to some of the world’s biggest brands and an analytics leader at CPG companies including P&G, SABMiller Brewing, and The Gillette Company.

Retail Media can be profitable for brands – here’s how!

Posted on by Kevin Youts

Retail Media can be profitable for brands – here’s how!
Retail Media networks are growing at a rapid pace, but brands are wondering about how to allocate their spend. Learn how i4i’s Robust Bayes™ approach to MMM can help.

MARK GARRATT – Retail Media is a rapidly growing channel. However, the top concerns of CPG marketers are whether RM is profitable and incremental to their business. Ultimate incrementality can only be determined by Marketing Mix Models (MMM). But unless you take account of the unique retailer footprint, your MMM value will likely be overstated, leading to money left on the table.

The 2022 Skai™ and BWG survey estimate that retail media networks will become the third-largest digital channel this year with projections of $100B in ad spend by 2024. Since 2020, this media channel has grown in both complexities and in the range of services offered to CPG manufacturers. The number of big retailers establishing their own networks grows at the rate of about one a month, and there are now multiple middle-layer agencies like Criteo, Citrus Ad, and Flywheel that are adding liquidity to the channel by allowing CPG companies to deploy spending across multiple retailers. As marketers rush to try out new capabilities, questions persist about profitability. According to the 2022 Skai™ survey, the top concern of senior managers is driving positive ROI (44% of those surveyed) and the second most important issue is proving incrementality (33%).

The High Peak and Sharp Tail of Retail Media:

in4mation insights includes Retail Media as a standard media input in their CPG models. And that has made us aware of a very important phenomenon. The initial impact of retail media networks is impressive. The RoAS has a high peak. But ultimately, the impact at any one retailer faces a limit or a “ceiling” based on the number of people who can be reached through that retailer network, and hitting up against this ceiling causes a rapid drop in RoAS that we call the sharp tail. How does this happen? And does it happen the same way to all retailers?

For the digital founders of retail media networks (Amazon, Apple) this ceiling was high enough to not be a factor. The audience reached by Amazon per month is 65% of the population 18-65. Among the retailers, Walmart has the biggest footprint but below Walmart, the drop-off is considerable.

When we consider retail media, we first think of online shoppers, but the world is more integrated than that. People who see sponsored ads on Target.com may be looking for one thing and see another advertised. They may think about it for a while and pick it up the next time they are in the physical store. It is not just the online rankings that matter but also the percent of the target population who visit the brick-and-mortar stores, and the overlap between online and offline shopping. Retail media networks sustain profitability in proportion to the percent of people reached, and the percent of people reached is determined by both the online and offline customer base.

The following chart demonstrates the high peak and the sharp tail: This chart is from our Optimetry™ software. It shows spend per week on the x-axis and return on ad spend (RoAS) on the y-axis (labeled ROI here). The green line is the marginal RoAS (revenue return on the next dollar spent on Retail Media) and the yellow line is the absolute profit. The data is for a large CPG manufacturer investing in the Target (Roundel) retail media network. The charts show that the marginal RoAS (green line) peaks out very early at around $100k per week spend. The RoAS level of over $10.00 attained is very high and that is good. However, the sharp decline in RoAS after that is due to reaching the hard ceiling on reach determined by the retailer’s customer base. This pattern will have an effect on optimizations. As you simulate spending more on retail media, there will reach a point where there is a media channel that improves overall profitability faster than retail media.

Other important factors to consider:

        • CPMs for retail media are rising fast. It will be important to monitor the rise in prices versus reach ceilings and this will vary by retailer.
        • A new, indirect pathway is being created between social media sites that redirect users to retail media sites. Credit will need to be split between these channels.
        • As with all digital media, visibility will need to be monitored as ads become compacted in the screen space around the search box.

What is the in4mation insights advantage?

Retail media is yet another digital variable competing for space in an already crowded digital field. The hierarchical Bayes methods we use have the following advantages:

        • Retail media can introduce a lot of sparsity into models as CPG media managers target both specific retailers and specific products. Advanced methods are needed to solve this (Click here to download our latest white paper on Robust Bayes™)
        • Media simulations/optimizations must account for hard ceilings on reach for different retailer audiences.

Retail Media Networks can be a highly profitable channel and will attract a lot of spend. But watch out for the sharp tail; it can easily become a channel where you overspend. Contact us at in4mation insights (info@in4ins.com) to help you assess the macro profitability for retail media across online and offline customers.

About Mark Garratt 

Mark Garratt is a partner and co-founder of in4mation insights. He is an accomplished analytics professional with a distinguished career in both business and academia. Mark has been a trusted advisor to some of the world’s biggest brands and an analytics leader at CPG companies including P&G, SABMiller Brewing, and The Gillette Company.

Choice Modeling – Menu-Based Conjoint

Posted on by Kevin Youts

Menu-Based Conjoint Analysis
Menu-Based conjoint analysis allows researchers to correctly predict how customers’ buying behavior changes when they’re presented with a menu of products and services.

STEVE COHEN – Choice modeling is a type of conjoint analysis that allows researchers to determine which product features are most important to consumers. Over time, it has developed into a powerful tool for analyzing consumer demand for any product and its combinations, substitutes, or complements.

This article looks at the development of Menu Models in conjoint analysis.

Menu Models

Choice from a menu involves picking one unit of more than one product, brand, or item from a list of all displayed items. Go to a fast-food restaurant and buy a burger, fries, and a beverage from the menu. The example below shows the choice from a menu of services that could be added to a phone plan. This “menu” is taken from a study that I conducted on behalf of a telecom company during the early 2000s. As services and features were chosen, the survey participant displayed the total.

Any solution to estimating consumers’ interests from a menu should, at the least, deliver these results:

  1. How many people choose each item?
  2. How price-sensitive are people to each item?
  3. Which items are substitutes?
  4. Which are complements?

Old Solutions for Menu Models

In 2000, I co-authored an academic paper in the Journal of Marketing Research (JMR) that first outlined a solution to estimating the menu model. Our approach handled item choice and price sensitivity in a straightforward manner, using an advanced choice (probit) model. Substitutes and complements were estimated through the correlations of the items with one another. The paper was one of four that were nominated for Best Paper that year in JMR. Unfortunately, we did not win the award, but the paper generated lots of interest and comments.

A few years later, Sawtooth Software introduced Menu-based Conjoint Analysis (MBCA) as a solution to the menu problem, building upon their existing discrete choice technology (their early technical paper referenced the JMR work). However, their solution to complements and substitutes requires the additional estimation of effects across individual item equations. If all of those cross-effects were important in a menu of 30 items, the model would need 900 additional terms. Since it’s unlikely that all 900 are needed, the analyst must search for the best of those, which could be very time-consuming.

The New Solution

The new method demonstrates that neither my solution nor Sawtooth’s can capture the true depth of complementarity and substitution. This is mainly because both solutions are really estimating the demand for one item on the menu at a time, whereas the new method is estimating the demand for all items, combinations, substitutes, and complements simultaneously.

How good is the new approach when compared to the other two existing methods?

Approach in JMR Paper Sawtooth Software New Menu Model
Overall Hit Rate 56% 53% 58%
Predict Choice of Single Items 52% 54% 51%
Predict Choice of Combinations of Items 36% 35% 47%

If we compare the Overall Hit Rate and Single Item Choice, the performance is relatively the same. Compared to the two older approaches, the new approach shines when looking at correct predictions of combinations: 47% correct versus 35%-36%, an increase in predictive power by 12 percentage points, or an overall 32% increase.

Applying the New Menu Model to Your Business

The use of conjoint analysis over the years has been proven to be a valuable tool for estimating consumers’ preferences from a menu of choices. It can be used to help you gain a better understanding of product demand, combinations, substitutes, and complements. The new menu model can help you revolutionize your research and allow you to understand your customers like never before. If this sounds like an area that could help your business, please contact me today – Schedule a call or Send an email.

About Steve Cohen 

Steve Cohen is an award-winning entrepreneur, specializing in the design of research, analysis of marketing data, and using marketing science tools to solve business problems. In 1984, Steve was the first commercial researcher to do choice-based conjoint, and subsequently, he authored the first conference presentation and paper on MaxDiff. He is a co-author of the first academic paper on Latent Class Choice-based Conjoint and Menu-based Conjoint. Steve was won numerous awards for his research including: AMA “Parlin” Award, INFORMS Society for Marketing Science “Buck Weaver” Award, NextGen Marketing Research Disruptive Innovator Award, NY Marketing Research Council’s Hall of Fame, ESOMAR Goodyear Award for the Best International Research Paper, AMA Hardin Award for the Best Paper in Marketing Research Magazine, and Best Presentation at Sawtooth Software Conference. Read More

Why Marketing Mix Models (MMM) are Ambitious

Posted on by Kevin Youts

Why Marketing Mix Models (MMM) are Ambitious
Effective media planning needs to accommodate the complex variables that impact your target audience and drive sales.

MARK GARRATT – Marketing mix modeling is a very ambitious type of analysis – it seeks to provide a single, holistic view of all the drivers of sales. The purchase behavior that we are trying to model may have many causes. Some of those causes are due to advertising and promotion and some are due to uncontrollable factors like the weather or the economy. Within the advertising-driven behaviors, there is another major dimension of difference: where does the advertising act in the consumer journey? Does advertising act at the top of the funnel generating broad awareness and reach, or does it act at the bottom of the funnel, capturing demand through timely and clever interventions? Linear TV is a typical top-of-funnel medium; retail media (such as the search box on Target.com) is one of the newer lower-funnel or demand-capture media.

Yet another dimension that a marketing mix model (MMM) must address is whether an ad is acting directly or indirectly. For instance, a display ad that is targeted to your phone is direct. But a TV spot that advertises a web domain is acting indirectly. MMM is ambitious because it seeks to capture all these impacts, from the most general to the most particular, in one model framework.

If we think of the major ways that data inputs can vary in a Marketing Mix Model (MMM), then we can identify several factors:

  1. Controllable versus non-controllable
  2. Upper, middle, or lower funnel
  3. Direct or indirect
  4. Whether costs can be defined or not
  5. The geography in which the advertising is planned or acts

These factors can apply equally to digital or linear media driving sales in brick & mortar or online classes of trade. The typical models used for MMM are based on regression. Regression is flexible. Nothing prevents us from trying to sum up the data to a fixed geography (store, chain, household, individual) in a fixed time (day, week, month) and put all these different types of variables in one model.  However, the model does not “know” things that you might take for granted. It does not know how the factors described above may distort the impact of advertising. Here are two examples of how complexity can find its way into models:

      • When it’s raining, I don’t want to go outdoors, especially if it’s cold or stormy. If I am trying to get dinner for the family, I may be more inclined to use a delivery service like GrubHub or UberEats. If I use one of those services, I will see ads for restaurants that may change where I end up placing my order.
      • As an advertiser, I spend part of my budget on social media. Part of that social media is national, but another part is market-level. Over time, I have shifted money from national social to market-level social. I have also used that market-level social specifically to introduce a loyalty app.

In the first case, we are dealing with an uncontrollable but important factor (the weather) that makes me shift from one distribution channel to another (a change in geography/place). While in that distribution channel, the consumer gets exposed to retail media (which may be originated by the retailer, not by the advertiser) which results in a purchase different from their regular brand.

In the second case, the advertiser is changing the weight of the impression on social media but that really is a shift to market-level (geography) which also comes at a higher cost. That depresses their short-term ROI. The local social media is operating at mid-funnel level, but it is being used to launch an app that is at lower-funnel or demand-capture level. The app is designed to bring the customer back to the store to get more “share of wallet” (an indirect effect) and part of the cost of market-level social media needs to be traded off against building customer loyalty with the app.

Range of Media Drivers, Why Marketing Mix Models (MMM) are Ambitious

The purpose of these examples is to show just how common it is for the impact of intended marketing effects to get amplified, diluted, or changed, sometimes by things out of our control (like the weather) and sometimes by the unintended consequences of our own plans. Also, it may begin to show the way that just putting more and more variables into regression models and expecting to get true reads on contribution to sales might be naïve.  There are specific interactions. There are effects at different levels of action. Costs keep changing. And there are a lot of potential effects.

How Hierarchical Bayes Helps Us Handle Ambitious Models

Bayesian analytics was named after the Rev. Thomas Bayes 1701-1761, who Brittanica describes as an “English nonconformist theologian and mathematician who was the first to use probability inductively and who established a mathematical basis for probability inference (a means of calculating, from the frequency with which an event has occurred in prior trials, the probability that it will occur in future trials.”

Such theories couldn’t be practically applied until the 1990s when major advances in computation made practical versions of it possible. Hierarchical Bayes describes a model that has layers or hierarchies; most often those layers are defined by different geographies or by-product characteristics. As described above, Bayesian models also have “priors”. Priors (which simply put are preliminary assumptions that may come from experience or data) are a useful way to build domain knowledge such as sign constraints into the results.

Hierarchical Bayes Regression allows us to construct models where we can handle many of the problems that make MMM tricky:

      • They allow us to break up the data in hierarchies of granularity. So that store-level effects (like local weather) can act at the store-level, DMA-level effects (like terrestrial radio) can act at the DMA level and national media (like national social media) can act at a national level. This ability to break up the data into levels of action solves many problems of misattribution.
      • They allow us to apply priors to variables so that domain knowledge derived from external norms, known constraints (e.g., advertising effect is positive), MTA findings, A/B tests, etc. is incorporated into results. This ability gives answers that make sense, are consistent over time, and lead to KPIs that drive action through sequential measurement.
      • But even more important, they are self-norming, so that if you have multiple products, packages, outlets, etc. pooled together, they can share the measure of advertising impact across similar products or geographies and reinforce the granular estimates. And if you have outliers, their values are guided toward the central tendency, preventing mistakes and phony anecdotes.
      • They are robust when media (or any effect) applies to only a subset of products or distribution channels (e.g., this retail promotion only applies to Target sales). This addresses the complexity in media that we see especially in the evolving online space.
      • Because they operate flexibly at granularity, Bayesian models allow us to measure many more variables than simpler forms of regression executed piecemeal.
      • They are efficient. Because when they are done right, Bayesian models offer high levels of control against “random acts of data.” So, analyst time can shift from babysitting models to interpretation of results and implications for action.

As we said at the start, MMM analytics are ambitious and complex. To get them right you need the kind of analytics partner who is an expert at wielding the sharpest tools (we would argue Hierarchical Bayes) to get you actionable recommendations for efficiency in media spend. Your analytics partner also needs to be knowledgeable of the evolving media landscape to demystify and decipher the real contributions of any media channel to your growth objectives.

In future posts, we will give you our POV on how to account for retail media when looking at your overall mix as well as topics like the challenges and opportunities modeling the impact of PR, affiliate, and influencer marketing against all your other marketing investments.

If you have a critical topic that is impacting your media mix decisions and would like to know more about how we can apply our special expertise to the advantage of your brand, let’s have a conversation.  We would love to help – contact us at info@in4ins.com.

About Mark Garratt 

Mark Garratt is a partner and co-founder of in4mation insights. He is an accomplished analytics professional with a distinguished career in both business and academia. Mark has been a trusted advisor to some of the world’s biggest brands and an analytics leader at CPG companies including P&G, SABMiller Brewing, and The Gillette Company.