MetricWorks is launching what it calls MMP 2.0 to boost performance marketing for games and other apps while preserving privacy and budget.
Ever since Apple prioritized privacy over targeted ads two years ago, mobile advertisers have been contending with the challenges of imprecise marketing. Now MetricWorks believes that by combining several different tactical solutions that help improve marketing results, it can put them all together to create better overall performance marketing.
Of course, this is a pretty complicated story and it takes a while to explain. But first we’ll note how Apple changed the market and then we’ll talk about how MetricWorks is trying to save the day. I do believe it’s worth following the tea leaves in this evolving story. And while Apple’s impact is largely behind the mobile industry, Google is likely to follow suit in implementing more privacy-facing solutions as well.
“Android will be following suit quite soon,” said Brian Krebs, CEO of MetricWorks, in an interview with GamesBeat. “There’s just been so much visibility lost. And Meta sees this as a systemic risk” and that is why it is using MetricWorks and looking for solutions in advance of more privacy regulations.
“What’s affected them so far has been absolutely devastating,” Krebs said.
And the effect on Meta’s customers, including advertisers such as mobile game makers, has been equally devastating, he said.
The background of the IDFA
MetricWorks positioned itself as a strong ally to many advertisers during this time, launching a tool for incrementality analysis, which used a series of experiments to determine the effectiveness of marketing campaigns at a time when Apple limited the use of the Identifier for Advertisers (IDFA), which once provided precise details of why a user made a purchase or downloaded a game.
Apple made it easier for users to opt out of being tracked by advertisers like Facebook (now Meta), and that hurt not only Meta but the entire mobile gaming industry, which had grown used to tracking individuals through what were known as mobile measurement providers, or MMPs. Those third-party companies received IDFA data from Apple and enabled advertisers to trace transactions and downloads back to the people who made them and the ads that spurred those actions to happen.
Such one-to-one precision in knowing just how effective an ad was became critical to companies seeking users like “Star Wars fans who liked hardcore role-playing mobile games but didn’t like fantasy yet still spent a lot of money in games” or something like that. And this precision fed the growth of adtech companies as well as the mobile game and app companies they served. But abuses in the system spurred Apple to focus on user privacy and make the data fuzzy when tracking individuals.
Apple’s SKAd Networks (SKAN) only let advertisers receive a single tracking notification a day after a transaction. So it became very difficult to measure what caused that transaction to happen, and which mobile advertiser should get credit for spurring a user to make a purchase or a download.
What were the results? For the first time since the dawn of the iPhone, the mobile gaming market shrank in 2022. While mobile game downloads rose to 89 billion, an 8% year-on-year increase. But mobile game spending dropped 5% during 2022 to $110 billion. And that dragged down the overall games market. The latest data from Ampere Analysis shows that in 2022, global spending on games content and services declined by 5.6% to $182.3 billion.
MetricWorks’ incrementality testing
When MetricWorks launched MetricWorks Polaris via an exclusive VentureBeat story in March 2021, many were in the industry were surprised, confused and unsure of the potential of incrementality to solve their measurement woes.
However, today, MMM (media mix modeling) is enjoying a resurgence, said Sanjay Raghu, vice president of marketing at MetricWorks, in an interview with GamesBeat.
It is widely accepted that Apple’s SKAdNetwork (or SKAN) is a disappointingly primitive and frustrating form of measurement riddled with holes. For instance, if multiple ad campaigns are going on at once, you can’t tell what actually triggered a transaction to happen 24 hours ago. At least not unless you’re running experiments like MetricWorks’ incrementality testing.
Krebs told us in an interview last year that this solution does not depend on bottom-up “last touch” data or device IDs (IDFA or Google’s GAID) in any way. With “last touch,” a measurement system assumes that the last marketing event before a transactions must have caused that transaction. But that’s not true.
“Last touch has a bunch of gaps in it,” Raghu said. “It’s broken, and measurement has been broken. And that’s where we come in to help solve that because [marketing] performance has suffered for our clients.”
It uses aggregated daily data and blends two techniques, econometrics and automated experimentation, to deliver truth in measurement. Together, these techniques make it easy for marketers to realize marketing effectiveness across their campaigns, regardless of IDFA retirement.
MetricWorks said its Polaris helped restore the capability to test how well a marketing campaign is performing. It makes incremental measurement real and accessible to mobile marketers with the same analytics they use today for campaign optimization. In short, with incrementality testing, MetricWorks enabled an advertiser to set up parallel tests that could determine whether a particular campaign cause a lift in sales or not.
While it helps, it’s not perfect. While alternative measurement approaches have emerged in the last two years, marketers are still battling with persistent gaps in their measurement that have hurt their marketing performance. And Raghu said the work isn’t done.
“We’ve gone on this journey. And we’re out there to continually improve performance, and deliver that through continued innovation,” Raghu said.
What are those measurement gaps?
There are three measurement gaps, said Raghu.
He pointed to poor accuracy. Due to flawed measurement techniques like last touch (including SKAN) with arbitrary attribution windows, accuracy remains a major problem for marketers.
And he noted the lack of privacy-safety: the resistance to break away from device IDs, fingerprinting and not requesting consent has led to a crackdown on privacy-violating measurement methods along with substantial fines.
Raghu also noted there is a limited measurement scope: the inability to measure completely across all media types including brand and influencer.
“We close those gaps on accuracy, privacy, and safety,” Raghu said. “We offer broader scope in terms of measuring more things than the traditional stuff can. And so once you close those three gaps, what you have is a higher form of measurement, holistic measurement, which works beautifully. And we’ve got success stories. We are calling this MMP 2.0.”
Available measurement options and their uses
Given those gaps, MetricWorks surveyed the available measurement methods and why each one by itself does not address the gaps above to provide a complete and effective measurement solution. As you can see in the table in Fig.1, there are four available methods: Deterministic Last Touch, SKAN Last Touch, Geo Lift Testing and MMM (media mix modeling).
The table describes the strengths, weaknesses, and common uses of each method. Here are the seven key takeaways from this table below.
First, no one measurement method is ideal on its own. They all have weaknesses. So, if you’re relying on only one of these, you are bound to have gaps and hence performance issues.
Therefore, to maximize accuracy, multiple measurement methods must be unified into a single source of truth in a way that exploits each method’s strengths (in gold) and minimizes each method’s weaknesses.
Mobile marketers already have existing infrastructure built around MMP (mobile measurement partner) data, which is a combination of rows 1 and 2, namely, deterministic last touch and SKAN last touch data. This is MMP 1.0 data which suffers from the same weaknesses described above in rows 1 and 2, namely, that it doesn’t measure incrementality, and cannot measure non-addressable media (offline, TV, influencer, etc.).
Despite those weaknesses, MMP 1.0 is still used to feed the following business processes: Tactical day-to-day user-acquisition processes (e.g., campaign budget optimization, bid optimization); strategic marketing processes (e.g., budgeting, media planning); prediction models (e.g., LTV); business intelligence and marketing reports.
MetricWorks noted that the MMP data is tactical. Its common use is tactical day-to-day decision-making (refer to the common uses for the first two rows in Fig.1). This means that it is daily, grouped into cohorts of data, and supports a variety of performance metrics.
Separating the use of multiple measurement methods into unique processes will inevitably cause conflicts, the company said. For instance, using media mix modeling (MMM) for budgeting while using last touch for day-to-day campaign optimization could result in some channels (that MMM evaluates favorably and last touch does not) getting big budgets that never get spent since UA teams would be setting low campaign spend caps and bids. Alternatively, it could result in extremely complex, interconnected processes which hamper the scaling of a business.
To be truly useful and easily adoptable, all chosen measurement methods must be adapted to be tactical with the single source of truth maintaining the exact same structure as MMP data. This is what MetricWorks refers to as the “MMP look and feel”.
The Birth of MMP 2.0
MetricWorks reflected on the above takeaways and wondered how to truly make progress to deliver the next generation of measurement.
First, MetricWorks asked what if, instead of using each of the above methods desperately, the company merged them into one? The team maximized the strengths and minimized the weaknesses to deliver far superior accuracy than any of the individual measurement methods, Krebs said. You may have noticed that the MMP 2.0 Uses column in Fig. 1 aligns with the Strengths column.
Secondly, the team asked, “What if we created this new measurement with the same look and feel as existing MMPs (mobile measurement partners)?” The company said MMPs offer simple and easy-to-use dashboards that are intuitive to use. This became a critical design requirement.
Going beyond the above product requirements, MetricWorks also wanted the solution to be recognized as a boon to UA teams. The MMP 2.0 accomplishes this in two ways:
It removes the guesswork (and potential conflict) associated with using multiple measurement signals by outputting a single credible source of truth (blended metrics) that can be shared across departments, including key stakeholders (aka execs).
And it fits into existing UA process without forcing its clients to make any changes or to forsake their hard-earned UA best practices.
Krebs said the combination of all of these practices works, effectively yielding, “Same Look. Superior Measurement. MMP 2.0.” MetricWorks calls the solution Polaris and it has been well received by a number of their clients and partners such as Blizzard, FunPlus, Kabam, Nexon, Meta and TikTok.
Making MMP 2.0 accessible to all
Today, MetricWorks is also launching a free tier for Polaris to make MMP 2.0 accessible to all marketers at no risk and no charge. Developers who advertise games can start using the Polaris Free Tier by signing up here to take advantage of its capabilities.
“We are now introducing a free tier so people advertisers can get to try the power Polaris for free,” Raghu said.
Those capabilities include Polaris’ modeled incrementality metrics for one title on one either iOS or Android. And it has one cohorted incrementality metric (eg. D7 Revenue) across all countries, channels, campaigns, sub-campaigns or creatives.
To start, MetricWorks offers visibility into an advertiser’s historical daily incrementality going back as far as 12 months. It also gives access to Blended Metrics down to the campaign, sub-campaign/creative levels.
MetricWorks compared how MMP 1.0 feeds typical business processes today versus how MMP 2.0 drives those same business processes with superior accuracy that leads to higher performance.
As you can see MMP 2.0 utilizes all four measurement methods rather than just the two that MMP 1.0 uses. This improvement results in a significant jump in measurement accuracy. Additionally, because MMP 2.0 fits into an advertiser’s existing UA process and has the same look and feel as MMP 1.0, it seamlessly replaces MMP 1.0 in a measurement stack.
“Ultimately, our mission is about demanding more performance from one’s data,” Krebs said. “These days, mobile marketers are under a great deal of pressure to deliver higher performance while preserving budget. The good news is that they can succeed under these tough constraints…with MMP 2.0. It brings a smile to my face when we get to offer people a fresh look at measurement and true marketing effectiveness.”
Media mix modeling (MMM) focuses on the broader activities that a user pursues during a day, from reading books to creating their own content. For decades, companies like Nielsen measured the result of brand advertising at maybe a couple of times a year and that helped determine where the ad budgets went. That doesn’t fly in the mobile advertising world, where you have to know what happened to cause a transaction on a daily basis. This kind of work is too difficult for small shops to do. And the data is not very granular.
“In terms of broader scope, marketers are not limited to just the traditional things that they can measure now,” Raghu said. “We can do brand and influencer marketing as well. And the pathway that we’re we’re going down yet is going to lead to more effective creative optimization.”
Meta is starting to get behind the concept of media mix modeling as the future of measurement. While this data is long-term in terms of collection and insights, MMP is more immediate and measured on a daily basis. Last touch might be good at measuring which of two campaigns is working, but it is weak on low volumes. MMM can be weak on new campaigns starting on a new platform, where there isn’t much data available yet. But it turns out to be more accurate than the horrible way of measuring via last touch.
“We have coalesced a variety of features into what we’re terming MMP 2.0. Because it MMP 1.0 is driven by last touch, and that has always sucked,” Krebs said. “Last touch is crappy and it’s getting even crappier. We need to take the MMM power and put it in an MMP package.”
VentureBeat’s mission is to be a digital town square for technical decision-makers to gain knowledge about transformative enterprise technology and transact. Discover our Briefings.