A Berlin-based digital tracking and attribution consultant who spent nearly five years at Adjust has outlined four criteria that mobile app marketing teams systematically get wrong when selecting a Mobile Measurement Partner(MMP) - and argued that the consequences tend to be severe and slow to surface.

Alessandro Deflorian, who set up as an independent consultant in March 2026 after a six-month stint as Senior Digital Attribution Manager at Delivery Hero, published the assessment on LinkedIn in late May 2026. The post, which received 13 reactions and one repost within its first week, draws on his 4 years and 8 months at Adjust and his subsequent client work navigating MMP evaluations. According to Deflorian, the patterns repeat across organisations regardless of size or vertical.

The post does not target any specific vendor. Instead it addresses the evaluation process itself - why teams land on the wrong platform not because the right one was unavailable, but because they asked the wrong questions.

What is a Mobile Measurement Partner?

Before examining what goes wrong, it helps to be precise about what an MMP actually does. A Mobile Measurement Partner is an independent, third-party platform that tracks in-app events, attributes installs and conversions to the advertising sources that drove them, and provides a unified view of campaign performance across ad networks and partners.

The major players in the space - Adjust, AppsFlyer, Branch, Kochava, Singular, Airbridge, and Tenjin - are certified by platforms including Meta, Google, and TikTok as recognised attribution providers. Each operates as a neutral layer between an app advertiser and the ad platforms buying their media, applying its own deduplication logic, attribution windows, and rules to determine which touchpoint receives credit for a conversion.

That independence is what makes MMPs valuable. But it is also what makes the selection decision consequential. Switching providers mid-campaign, or mid-growth phase, is costly in engineering time and introduces breaks in historical data comparability. The MMP a team installs effectively becomes infrastructure - and like infrastructure, it is far easier to choose well upfront than to replace later.

Why selection goes wrong: the four patterns

According to Deflorian, the same four flawed criteria drive most MMP decisions.

1. Agency recommendation as a proxy for fit

The first and most common mistake is treating an agency's preferred tool as the right tool. Deflorian writes that agencies recommend what they know - not what matches a client's specific attribution requirements, market footprint, or measurement maturity. "Agencies have preferred tools," he notes. "Make sure you understand whose interest is being served."

This is not an accusation of bad faith. Agencies operate at scale across many clients, and familiarity with a platform reduces their own onboarding and reporting overhead. But that efficiency calculus is the agency's, not the advertiser's. A recommendation shaped by which platform an agency's team already knows how to configure does not account for the particularities of the advertiser's tech stack, the markets where their app operates, or the attribution methodology their business actually requires.

The distinction matters because MMP integration is deeply technical. Each platform has its own SDK, its own requirements for iOS and Android compatibility, and its own approach to connecting with ad networks. What integrates cleanly with an agency's existing dashboards may require significant additional engineering work on the advertiser's side.

2. Competitor choice as market signal

The second mistake is selecting an MMP because a competitor is using one. Deflorian is direct about the limits of this reasoning. Job listings and posts from former employees may indicate which tool a competitor has deployed, but they reveal nothing about whether that choice has worked. "You don't know what your competitors are actually getting out of it," he writes.

This matters in part because MMP performance varies by context. The same platform may handle gaming attribution differently from fintech attribution. Its support infrastructure may be strong in North America and thin in Southeast Asia. The configuration that works for a competitor's team structure may be poorly suited to a different organisation's workflows.

Platform certification frameworks exist partly to address this heterogeneity - by establishing baseline quality standards across providers - but they do not resolve the question of which certified partner best fits a given advertiser's needs.

3. Feature count mistaken for capability

The third error is evaluating MMPs by comparing feature lists. Deflorian argues that feature lists are marketing. A platform may carry hundreds of capabilities, but what matters is whether the specific capabilities a team actually needs work reliably, integrate cleanly with the ad networks it buys from, and are supported in the markets where it operates. "A bloated feature set you'll use 20% of is not an advantage," he writes.

The relevance of this point has sharpened alongside the growing complexity of the mobile attribution landscape. Since Apple implemented its App Tracking Transparency framework in April 2021, requiring apps to obtain user permission before accessing the Identifier for Advertisers, the measurement environment on iOS has become structurally more difficult. SKAdNetwork, now rebranded by Apple as AdAttributionKit, has become the default privacy-preserving attribution framework for iOS. According to industry data, AdAttributionKit now covers 77% of all referral-based conversions to the App Store.

This technical reality means the relevant question when evaluating an MMP is not how many features it has, but whether it handles SKAN postbacks reliably, whether its modelled attribution is calibrated for the specific networks a team buys from, and whether it supports the SDK version their iOS app requires. Meta's badged MMP program, for example, includes Adjust, Airbridge, AppsFlyer, Branch, Kochava, Singular, and Tenjin - and its November 2025 AI optimisation improvements included updated reattribution windows specifically designed to align with MMP-specific definitions. That alignment matters, but it is specific to the MMP-to-platform pairing, not a property that applies universally across all MMP integrations.

A platform with a long feature list may handle SKAN poorly in a key market. A more focused tool might handle it well. The feature count does not tell the buyer which is which.

4. Brand name or price as decision anchors

The fourth failure mode is letting brand recognition or headline price drive the decision. According to Deflorian, brand name signals stability but not fit. Price signals very little. "The real cost of an MMP is never just the license fee - it's the engineering time, the data quality risk, and the opportunity cost of outgrowing it."

This framing reframes the total cost of ownership in a way that is rarely reflected in formal procurement processes. License fees are visible and comparable. Engineering costs are distributed across months of SDK maintenance, debugging, and integration work. Data quality risks are invisible until they are not - until a team discovers that its attribution model has been quietly miscounting installs, or that a mismatch in attribution windows has been inflating reported ROAS.

Kochava's MMM Data Validator tool, released in January 2026, illustrates the scale of this problem. The tool was designed to detect data quality issues before they undermine marketing mix modeling implementations. It allows app marketers to upload CSV files containing up to 2,000 rows of campaign data and receive automated reports identifying common errors including missing operating system values, incomplete network data, absent cost information, and conversion tracking gaps. The fact that Kochava needed to build such a tool suggests that silent data problems in MMP implementations are common enough to warrant dedicated tooling.

The questions teams should ask instead

Deflorian's argument is not that there are simple right answers to MMP selection, but that the industry has settled on the wrong questions. The criteria he identifies as more useful are notably less glamorous than brand rankings or feature comparisons.

How does the platform handle a specific organisation's attribution needs? That requires understanding the attribution methodology the business actually requires - last-touch, multi-touch, probabilistic, or some combination - and verifying that the MMP implements it correctly for the networks and markets in question.

How is support in the relevant region? Attribution and measurement infrastructure varies substantially by geography. The Adjust and Google Ads web-to-app handbook, published in September 2025, detailed how UTM parameter mapping between Google Ads campaigns and Adjust attribution systems must be correctly configured to preserve campaign context through web-to-app journeys - a technical requirement that depends on regional network support as much as on the tool itself.

What does migration look like if it becomes necessary? This is the question Deflorian says nobody asks until they are already stuck. MMP migration involves re-integrating SDKs, reconciling historical data under different attribution methodologies, re-establishing certifications with ad platforms, and - critically - explaining to stakeholders why performance data looks different in the new system. Teams that have not modelled this cost before selecting a vendor have no leverage in the exit scenario.

The measurement confidence backdrop

Deflorian's post arrives at a moment when industry research consistently documents dissatisfaction with measurement technology across the marketing sector. According to research from TransUnion and EMARKETER published in October 2025, 54.1% of the 196 marketing professionals surveyed reported no change in measurement confidence year-over-year, while 14.3% said it had declined. Over a quarter of respondents - 26.5% - said they were dissatisfied with their current measurement tech stack.

Those numbers span all forms of digital measurement, not mobile-specifically. But the dynamics Deflorian describes for MMP selection parallel the broader pattern. More tools have not translated into greater clarity. Teams invested in platforms that did not fit their needs, and are now either managing the consequences or planning migrations.

The measurement confidence problem has a structural dimension too. Platform-provided attribution remains the most common methodology at 65.8%, but marketers are increasingly supplementing it with incrementality testing and marketing mix modeling. Nearly half - 46.9% - plan to increase investment in MMM over the next 12 months. That shift creates new demands on MMP infrastructure, since MMM requires clean, consistent cost and conversion data at the channel level. A poorly matched MMP produces exactly the kind of data gaps that Kochava's validator was built to catch.

Reddit's dual attribution feature, launched in May 2026, offers a practical illustration of why MMP configuration matters at a platform level. The feature surfaces both Reddit's first-party attribution data and MMP data side by side, making the gap between them visible. That gap - which exists because MMPs apply deduplication logic and their own attribution windows that differ from those of the platform reporting the conversion - is a permanent feature of the multi-source attribution environment. How large that gap is, and how reliable either side of it is, depends directly on how well the MMP is configured for the specific platform integration.

What the Adjust background adds

Deflorian mentions his time at Adjust - nearly five years across what his LinkedIn profile lists as multiple roles - not to endorse the platform but to establish that the patterns he describes are structural. Between his Adjust tenure and the launch of his own practice in March 2026, he served as Senior Digital Attribution Manager at Delivery Hero in Berlin for six months, from September 2024 to February 2025, giving him direct exposure to how attribution decisions play out at a large consumer app company. Adjust is one of the certified badged MMPs across Meta, Google, TikTok, Reddit, and other major platforms, and is the same platform whose web-to-app handbook with Google Ads was published in September 2025. Having worked on the vendor side and then in-house, Deflorian is positioned to observe how clients arrive at the decision - and how frequently the reasoning is disconnected from what actually drives measurement quality.

The post does not position any single vendor as the correct answer. It is instead an argument that the procurement frameworks teams use for MMP selection are systematically misaligned with the technical and operational realities of running app measurement infrastructure. The right MMP for a given organisation depends on factors that agency recommendations, competitor benchmarks, feature catalogues, and price lists cannot surface.

Industry context: certification, SKAN, and the stakes of switching

The complexity of the current mobile measurement environment makes vendor fit more consequential, not less. Google's enhanced measurement tools for iOS app campaigns, launched in August 2025, include Target ROAS bidding and expanded on-device measurement capabilities - capabilities that depend on App Attribution Partners, including certified MMPs, maintaining specific SDK versions and integration standards. Similarly, Amazon DSP's Events Manager expansion in February 2024 connected in-app conversion data from MMPs including Kochava, Adjust, and Singular directly to Amazon DSP campaigns - but only within the specific certification framework Amazon established.

These platform-specific integrations mean that switching MMPs is not simply a matter of installing a different SDK. It requires re-establishing certification relationships with each platform integration, reconfiguring attribution windows and postback rules, and accepting a period during which historical data comparability is broken. The engineering time and data quality risk that Deflorian identifies as hidden costs are, in the context of platform-specific integrations, very real and quantifiable.

TikTok's real-time iOS conversion tracking, enabled through a Kochava partnership in October 2025, is an example of how certification relationships between specific MMPs and specific platforms create value that is not transferable. A team that selected its MMP without investigating which platforms had the deepest integrations with which partners may find itself without access to capabilities that competitors are already using.

Deflorian closes his post with a question rather than a prescription: "What drove your MMP choice?" The response this generates - or does not generate - is itself a form of data. Teams that cannot articulate what drove their MMP selection, beyond agency recommendation or competitive benchmarking, are the ones most likely to find themselves facing a painful migration 18 months later, or living with data that nobody questions.

Timeline

Summary

Who: Alessandro Deflorian, a Berlin-based independent consultant in digital tracking and attribution, who spent 4 years and 8 months at Adjust and subsequently served as Senior Digital Attribution Manager at Delivery Hero before launching his own practice in March 2026. He is writing for an industry audience of app marketing and mobile measurement professionals.

What: A LinkedIn post identifying four criteria that teams routinely use when selecting a Mobile Measurement Partner- agency recommendation, competitor benchmarking, feature count, and brand name or price - and arguing that all four are structurally misaligned with what drives measurement quality. According to Deflorian, the right criteria are attribution fit for specific needs, regional support quality, and migration cost.

When: Published in late May 2026, approximately one week before May 31, 2026.

Where: LinkedIn, in a post from Deflorian's professional profile, directed at the mobile app marketing and ad tech community.

Why: Deflorian draws on repeated observation of MMP evaluation processes - both as a vendor at Adjust and as an independent consultant - to argue that the patterns leading to poor MMP selection are systematic and predictable. The consequences he identifies are concrete: a painful migration 18 months after the initial decision, or quietly bad data that goes unquestioned. The post is aimed at teams currently in the middle of an MMP decision, with the aim of surfacing the questions they are not yet asking.