MMM (Marketing Mix Modeling) is an analysis method that comprehensively analyzes online and offline marketing measures, including external factors other than marketing activities such as seasonal factors, and visualizes the impact of each measure on sales. It means.

In the United States, an advanced marketing country, about 8% of marketers are aware of the marketing analysis method (1), and about half of companies practice it, but as of 2020, the adoption rate of Japanese companies remained at about 1%. It was (2).

However, recently, new entrants into the MMM field have been occurring one after another in Japan. XICA, which has been providing MMM for about 2016 years since 8, feels a strong sense of crisis about this MMM trend.

Why do Japanese marketers need MMM now? And what is the warning bell that XICA is ringing against the MMM epidemic? We spoke to Yoshiaki Hirao, President and CEO of XICA, who has been involved in marketing for over 250 national clients and has been at the forefront of MMM.

(1) Announcement of the latest report from a consortium by Nielsen Media, Facebook, and Google, “How to improve advertising performance using marketing mix modeling” (Press release from Nielsen Media Japan LLC)https://prtimes.jp/main/html/rd/p/000000004.000070691.html
(2) Questionnaire survey on advertising effectiveness measurement methods asked to 212 corporate advertising personnel 2020 edition (Press release from XICA Co., Ltd.)https://www.atpress.ne.jp/news/213842

Why MMM now?

── What are the factors that made it popular in Japan?

There are two possible reasons for the popularity.

First of all,Diversification and complexity of marketing measures.

As users' lifestyles and contact points become more diverse, a variety of marketing measures are required, both online and offline, to maximize results.

This makes measuring effectiveness more complex and more difficult. This is why I think MMM is becoming increasingly important because it allows us to visualize with a high degree of precision the impact on results even when complex factors are intertwined.

── What is the other factor?

The arrival of the “cookieless era”.

In Japan, the Personal Information Protection Act was revised in April 2022, and awareness of privacy protection is increasing globally. As a result of this trend, it has become difficult for companies to obtain personal information for marketing activities.

MMM is an analysis method that uses statistics, so it is possible to understand trends from statistical data even without personal information. Therefore, I think it has become essential in the cookieless era.

── Please tell us how XICA came to provide analysis services using MMM.

XICA has been providing analysis services using MMM since 2016. We originally started as a data science company and did not specialize in advertising or marketing. After developing a product with the concept of ``anyone can easily perform statistical analysis'' and launching it for about a year and a half, we realized thatOver 9% of clients are marketersIt was.

When we talk to clients,"We're investing a huge amount of money in advertising, but we don't know how much return we'll get."The story has come up. I thought this was a problem that needed to be solved, and from there, the MMM analysis service MAGELLAN was created.

As a result of updating MAGELLAN by identifying what could not be achieved with existing MMM, it has now been used by over 250 clients.

── There was also a press release announcing that more than 200 companies had adopted it. Has it continued to grow steadily since then?

It is growing steadily. Also, our sense of direction has changed..MMM is only a starting point for creating a hypothesis.So I realized that just handing over MMM tools just like SaaS is no good.

Therefore, we are currently building a system that comprehensively evaluates marketing, from consulting such as planning client marketing strategies and formulating strategies and tactics, to execution such as creative production and media placement, based on effectiveness verification using MMM.・We support you until the operation.

Therefore, rather than expanding MMM to a wide range of companies, we are now deeply involved with each company as a marketing partner and working with them over the long term to help them grow their business.

Important things when using MMM

── What is important when using MMM?

I think there are two important things in MMM.precision and speed.

Accuracy and speed are two sides of the same coin, and unless these two are compatible, the value of MMM cannot be demonstrated.

Sales are made up of a variety of factors, including promotions, prices, and competitive conditions. By unraveling these relationships, we can finally understand the effectiveness of each marketing measure. If you do not comprehensively analyze the various factors involved, the results of your analysis may be completely opposite.

To be more technical,There is no point in analyzing marketing measures and their respective contribution values ​​unless they are understood structurally.Therefore, it is very important to improve the accuracy of MMM.

The larger the scale of marketing, the greater the influence of factors other than marketing measures, and the analysis results are likely to fluctuate, so it is better not to use MMM for decision-making if accuracy cannot be guaranteed.

── What about speed?

The second problem that arises when trying to achieve the "increased accuracy" that I mentioned earlier is speed. Pursuing accuracy makes analysis too difficult and time-consuming. In some cases, we strive for accuracy so much that we are only able to perform analysis once a year.

If analysis takes time, even if you use MMM to perform highly accurate analysis, the market environment and trends may change during the analysis period, making it impossible to utilize it for business improvement.

Accuracy and speed are two sides of the same coin, and MMM cannot be used unless these two coexist.

Sense of crisis over the MMM epidemic

── Various MMM tools have become popular these days, what do you think about them?

Since comprehensive evaluation is essential for marketing success, I think the widespread adoption of MMM is a positive sign for Japanese marketing.

However, on the other hand, there are some aspects of the current trend that feel dangerous.

I mentioned that accuracy and speed are important in MMM, but in recent years various MMM solutions have come out.The emphasis is on speed and low price, and the important accuracy is being neglected.I think there are some.

Our goal is to succeed in marketing and produce results, so as a pioneer in this field, we will continue to provide solutions that will lead to business growth.I would like to contribute to ensuring the quality of MMM throughout the industry.I believe.

When a particular technology or method becomes popular, not just MMM, it is common for solutions based on surface-level understanding to emerge and the essential value to be lost.

── What is not achieved by MMM with low accuracy?

The premise is that the elements that make up sales have a "structure."

For example, a structure such as ``TV commercials contribute to improving the performance of online measures.'' More specifically, there is an indirect and fluid state of mind, such as, ``As a result of the airing of TV commercials, the brand accumulation effect (brand equity) has increased, resulting in higher click rates, increased conversions, and increased sales.'' It is a structure created by human actions.

Correctly understanding this structure is the key to correct effect measurement. will be important.

Light MMMs are also coming out, but theseIf the structure, brand accumulation effect (brand equity), and time series are not analyzed correctly, there is a possibility that the decision-making will be in the opposite direction.So I think it's very scary.

── What is the structure?

Interrelationships and correlations, including indirect effects, that occur during the series of impressions, clicks, and conversions..

MMM analysis methods include a multi-stage analytical model and a single-stage analytical model, but in order to clarify the indirect and fluid state of mind and the structure created by human actions, a multi-stage analytical model is required. must be used.

In addition to being able to understand the structure described above, a multi-stage analytical model can alsoFairly evaluate the effects of recognition-related measures (measures aimed at obtaining profits through medium- to long-term brand building) and reaping-related measures (measures aimed at obtaining short-term profits)Available

── What is the brand accumulation effect?

What is the brand accumulation effect?The influence that advertising has on users' purchasing behavior over the long term.is. For example, TV commercials are a typical example of a situation where decisions to reduce fees are easily made due to incorrect analysis, but in the first place, TV commercials are about long-distance running, not short-distance running. This is a measure that is often implemented with the aim of establishing a brand's position over a 5-10 year period.

In order to accurately understand the effectiveness of TV commercials, it is important toEffects that persist even after the advertisement ends, such as brand accumulation effects, ripple effects, and residual effects, must also be included in the analysis.

By multiplying and analyzing these effects with the indirect effects that boost KPIs, the way we view the effectiveness of TV commercials may change dramatically.

Joint research between Fuji Television and XICAMore quotation

*Baseline: Sales that would occur without advertising as a result of brand building that has been accumulated in the past.
(*3) This is a score (249-2017) created from statistically processed analytical data (2021 cases) owned by XICA, so it is not common to all cases.
(*4) In this study, “long-term” is defined as a period of 3 to 5 years.

── What kind of analysis is based on time series?

Time-series data is data that is arranged according to time order, such as daily, monthly, quarterly, or yearly. Sales and impressions are also time-series data.

MMM uses statistical methods to express relationships between multiple time-series data that may affect sales using model formulas. By doing so, the structure of the factors becomes clear and the interrelationships and influence of each factor can be evaluated with high precision.

The analysis will not be accurate unless we understand the structure of investment and results based on the indirect effects, brand accumulation effects, and time series that we have talked about so far.

Intrinsic value of MMM

── What do you think is the essential value of MMM?

Being able to run PDCAI think it 's.

── What do you mean?

What MMM provides is "hints for forming a hypothesis", not answers.I mentioned earlier that "accuracy is important",In order to improve accuracy, we must have a hypothesis.Hm.

Also, the results of the analysis are not constant over time. It changes rapidly depending on the market environment, such as competition and trends. Therefore, it is necessary to continue the analysis on a regular basis, preferably once a month, or at the latest once a quarter.

The companies that continue to produce results for our clients require a lot of trial and error. We are able to achieve sustained results precisely because we continue the cycle of strategy formulation → execution → reflection using MMM → decisions on course corrections and enhancements based on hypotheses.

The essential value of MMM is to be able to perform the PDCA cycle with high precision and speed.I think

── What are XICA's thoughts on MMM as various tools are released and foreign-affiliated MMMs are increasingly becoming open source?

As I mentioned earlier, MMM is a common practice for many overseas companies, so I think Japanese companies should do it as well. It is therefore welcome that it becomes popular in Japan.

Although,MMM is a method that can easily produce unbelievable results if you do not pursue analytical accuracy, so we must properly educate people about the value of MMM.That's what I think.

The last thing I want is for people to think that ``MMM can't be used'' when they see the results of using an MMM tool that doesn't have high analysis accuracy, so I'd like to properly educate them on the importance of understanding the structure and understanding the time series. , I want to maintain order in MMM.

── What do you think will happen to MMM in the future?

There was a time when just saying, "You can see the effects of TV commercials with MMM!" was considered revolutionary. However, now that awareness has spread,An era where people are judged by how they use MMMI think it 's.

── Are there any changes in client requests?

yes. Until five or six years ago, the mainstream usage was to utilize MMM for optimal media budget allocation.

However, in recent years, questions such as ``What kind of creative is good?'' ``What is the balance between branding and product promotion advertising?'' ``(When looking at each user funnel) This cannot be done without analyzing the effects of long-term brand accumulation, such as how the measures are impacting and what kind of marketing mix is ​​optimal for improving overall sales.I want to use MMM for more upstream strategy design.The number of clients is increasing. MMM can be used to identify issues in order to develop marketing communication strategies.

── What is your future strategy as XICA?

At XICA, we are strengthening our consulting services. Based on MMM, I would like to put more effort into converting the analysis results into strategies and tactics and putting them into action.