Únete al AI + Data Tour para recibir formación práctica, conocer historias reales de clientes y pasar tiempo con los expertos en productos de Domo cerca de ti.
Mobile Analytics: Complete Guide to Metrics & Privacy

In 2026, mobile analytics comes down to a single variable: knowing whether your event taxonomy is sufficiently clean to withstand privacy constraints. Succeed at this point, and the metrics, attribution as well as the choice of tools will follow naturally. Get it wrong, and no dashboard will be able to save you. This guide covers the metrics, the privacy rules, and the configuration work necessary to make both usable.
Key takeaways
Here are the main points to remember:
- Mobile analytics makes it possible to track how users discover, use, and return to your application, thus helping you to eliminate friction points, deploy more useful features, and increase your revenue
- The 5 types of mobile analytics (advertising, engagement, monetization, performance, and app store) meet the needs of different teams and distinct business objectives
- Privacy protection frameworks such as Apple’s App Tracking Transparency (ATT), SKAdNetwork, and the evolution of advertising measurement standards on Android determine what you can measure in 2026
- The key metrics to monitor include acquisition and activation rates, retention cohorts, the conversion rate linked to monetization, and technical quality indicators
- A clear event taxonomy, cross-team agreement on definitions, and tools unifying data coming from different sources distinguish the teams that concretely leverage mobile analytics from those that are content merely to collect data
What is mobile analytics?
Mobile analytics consists of the collection and analysis of data describing the way users interact with your mobile application, your mobile website, and your marketing campaigns, as well as the technical performance of these experiences. This practice rests on three families of signals: the actions carried out by users (events such as app_open, view_item, add_to_cart, purchase), the regularity of their visits (retention), and the stability of the experience (crashes, display slowness,battery consumption). Associate these usage signals with campaign attribution data and you will know which channels attract the users who are worth retaining.
Mobile app analytics encompasses the tracking of behaviors integrated into the application, mobile web analytics, campaign attribution, crash and performance monitoring, as well as metrics relating to app stores. Outside this scope: desktop web analytics, backend server analytics without a link to mobile, and the company's offline operational data.
Platform privacy rules oversee the whole set. Apple's App Tracking Transparency, SKAdNetwork, and the constant evolutions of ad measurement on Android determine the level of individual detail that you are authorized to retain.
How mobile analytics works
Instrumentation begins with a software development kit (SDK) integrated into your application, which records events as the user navigates through your product. Each event relies on a standardized taxonomy, the shared vocabulary that your team has validated to define what is tracked and how it is named. A click on “Add to Cart” triggers an add_to_cart event transmitting parameters such as product_id, price, and category.
Next, identity resolution assembles these events into a unified journey. The SDK assigns an anonymous device identifier before any login. During authentication, this anonymous identifier merges with the person's user_id, thus grouping pre-login and post-login behaviors into a single profile. Your own rules dictate this merge: the consequences of a logout, the handling of reinstalls, or the elimination of duplicated events resulting from offline attempts. Teams that skip this step obtain fragmented data, and correcting the situation after deployment costs much more than upstream. Establish these rules in writing before production release.
The events are then routed from the device to a data pipeline that validates, enriches, and stores them. The result is a centralized data layer to which product, marketing, and engineering teams can all refer without having to debate the legitimacy of the numbers.
Mobile analytics vs web analytics
The objective is identical, to understand behaviors, but five technical differences often complicate the task for teams.
App stores modify visibility. Ratings, reviews, and keyword ranking within Apple's App Store and Google Play directly influence the profiles that discover you, a variable that web analytics never has to model.
Connectivity is not guaranteed. Applications continue to function offline, so events are queued locally and synchronized as soon as the connection is restored. Web analytics assumes a continuous connection.
Push notifications add an additional channel. Acceptance rates (opt-in), delivery, and clicks are all measurable touchpoints. Web push exists, but its impact is clearly lower.
Privacy rules diverge strongly. ATT and the continuous changes in ad measurement on Android impose limits specific to mobile with no strict equivalent on the web.
The device-related context is more in-depth. Device type, operating system version, battery level, and network status are all accessible signals that web analytics generally cannot perceive.
{{custom-cta-1}}
Benefits of mobile analytics
Teams that invest in this area generally see gains on five fronts:
- Faster product decisions: usage data highlights the precise screens where engagement drops off, which allows you to correct these as a priority rather than relying on assumptions
- Better retention: certain behaviors are predictive of retention, such as completing onboarding or enabling notifications, and you can direct more users toward these steps
- Performance gains: the rate of crash-free sessions, cold-start time, and core vitals are directly correlated with store ratings and app discoverability, particularly on Android
- More effective spending: attribution makes it possible to distinguish the channels that bring active users from those that generate only installs with no follow-up
- Cross-functional visibility: executives retain strategic oversight from any device, while data engineers receive alerts on pipelines before erroneous data reaches a report
Mobile access is the advantage that most teams underestimate. A salesperson between two meetings, a marketing coordinator at the site of an event, a vice president in mid-flight: for each of them, a mobile dashboard determines whether they react to what has just changed or to the situation from the day before. Employees in the field benefit from views adapted to their role directly on their phone, without having to call on an analyst.
5 types of mobile analytics
There is no single discipline called mobile analytics. It is rather a set of related practices, each answering different questions for different teams. Knowing which type matches your situation tells you where to focus your instrumentation efforts and which mobile app analytics tools to shortlist.
Mobile advertising analytics
This is where the performance of paid campaigns is measured: cost per install, return on ad spend, and quality gap between the channels that attract real people and those that generate valueless installs.
Use it when paid acquisition is active and the budget must be reallocated. Marketing teams use it to justify spending and redirect budgets toward what converts.
In-app engagement analytics
What happens inside the application is at the heart of the matter: session frequency, feature adoption, navigation paths, time spent on key screens.
Diagnosing product-market fit, identifying value-creating features, or understanding why users disappear at the third step: everything starts here. Product managers and user experience (UX) designers devote the greater part of their time to these data.
Monetization analytics
Behavior always ends up being reflected in revenue. This type of analytics tracks subscription conversions, in-app purchases, average basket, and lifetime value calculations.
Consult it when it becomes necessary to connect behavior to financial revenue. Finance, growth, and product leaders use monetization analytics to establish forecasts and prioritize the features for which users actually pay.
Performance analytics
Technical health is the priority here: crashes, application notresponding (ANR) errors, cold-start time, application programming interface (API) latency, battery consumption.
Engineering owns these data, but the consequences ripple everywhere else. Here is how technical aspects translate into business outcomes:
A slowed startup time constrains activation, yet activation is the essential input metric. Fewer users accessing first value leads by cumulative effect to lower D+7 retention and reduced lifetime value, which explains why performance regressions are often mistakenly taken for product failures.
App store analytics
This is where the visibility of your storefront is measured: keyword ranking, page view-to-install conversion rate, ratings and sentiment analysis of reviews on the App Store and Google Play.
Use it when discoverability is at stake, or when you need to understand why internet users see your listing and keep scrolling their screen.
{{custom-cta-2}}
Privacy and attribution in 2026
Privacy on mobile is now the subject of genuine rigor of enforcement. Integrate it from the design stage rather than applying fixes to it after the fact.
- iOS ATT: Data sharing with other companies for the purposes of cross-app tracking requires authorization via the Apple's AppTrackingTransparency framework, and the user's decision is binding.
- iOS SKAdNetwork (SKAN): Apple's privacy-preserving solution for measuring installs and campaigns. Details at the individual level are limited, but aggregated and delayed signals nevertheless make it possible to evaluate campaign performance. SKAN 4 (iOS 16.1+) introduced multiple postbacks as well as coarse conversion values over longer attribution windows.
- Android Privacy Sandbox (Attribution Reporting): Google's Attribution Reporting API, now obsolete, aimed to measure ads without cross-third-party identifiers. In October 2025, Google abandoned this API as well as other Privacy Sandbox APIs, citing low adoption, thus maintaining ad measurement on Android on existing identifier-based approaches, while the industry defines their substitutes.
- Store telemetry: Android vitals exposes performance metrics that influence both the user experience and ranking. Treat them with the same seriousness as a revenue KPI.
It is useful to clarify a distinction that most guides neglect: ATT governs cross-app tracking for advertising purposes. First-party product analytics within the application remains at your disposal, provided that you are transparent and respect consent when regulation requires it. Losing ad identifiers at the device level does not mean losing the ability to see where your onboarding funnel leaks.
Compliance-by-design checklist
Integrating privacy from the instrumentation phase costs a fraction of what compliance after the fact after launch would represent. Review these points before your first production release:
- Consent first: Trigger ATT permission requests at the precise moment when the value exchange is obvious. For the General Data Protection Regulation (GDPR) and the California Consumer Privacy Act (CCPA), collect explicit consent before any tracking and document your legal basis.
- Minimize collection: Track only what you are going to use. The absence of a clear use case implies collecting no data point.
- Exclude personally identifiable information (PII) from events: Email addresses, phone numbers, precise geographic coordinates, and official identifiers have no place in event parameters. Hashed or anonymized identifiers fulfill this role perfectly.
- Define retention policies: Determine the lifespan of the data and automate its deletion. The right to erasure provided by the GDPR requires that purging on request be an operational feature, and not a manual procedure carried out in urgency.
- Secure transmissions: Event data must transit via HTTPS with encryption. Audit the configuration of your SDKs to identify any channels that are not airtight.
- Document data flows: Know where the data goes, which third parties receive it, and how you will respond to data subject access requests (DSARs).
Measurement survives all these evolutions. It simply requires more aggregation, more modeling, and more rigorous event design than what the era of identifier abundance required.
How different teams use mobile analytics
Different profiles ask different questions from the same data. Defining the target audience from the outset avoids having to rework the dashboards afterward.
- Product managers: rely on engagement and retention data to prioritize features, spot friction points, and verify that the changes deployed have shifted the indicators
- Marketing teams: leverage advertising analytics and attribution to adjust campaign spending, assess channel quality, and connect acquisition cost to the value generated downstream
- UX designers: study navigation paths, session recordings, and drop-off points to identify confusing flows and support their design decisions with concrete facts
- Engineering teams: monitor crash rates, ANR events, and performance metrics in order to intercept problems before they ripple through to ratings and retention
- Customer success teams: analyze adoption and engagement trends to spot at-risk accounts as early as possible, including within embedded analytics experiences where a faulty mobile display quietly feeds churn
- Field staff and line-of-business (LOB) managers: consult pre-filtered dashboards adapted to their role, and ask questions in natural language from their phone between meetings, without requiring the slightest technical skill
- Executives, IT teams, and data leaders: keep within reach, from any device, a cross-functional view of KPIs, governance, and pipeline health alerts
Core metrics that matter
Mobile analytics metrics are worthwhile only if the whole team agrees on their calculation formula. Before creating the slightest report, agree on the KPI definitions and on the special cases that they must handle.
Acquisition and activation
First question: do new users go beyond the first contact? Two checkpoints reveal to you the essentials of what you need to know. How many installs turn into first opens, and how many first opens generate a key action such as account creation, tutorial completion, or the addition of a first item.
Here are the formulas you need:
- Install-to-open rate = first opens / total installs
- CPI (cost per install) = total advertising spend / number of installs
- CPA (cost per action) = total advertising spend / number of desired actions
- CAC (customer acquisition cost) = all sales and marketing costs / new paying customers
Spend $10,000 on ads for 5,000 installs and your CPI is $2. If 3,000 of these people complete their sign-up, the CPA comes to $3.33.
A large number of installs accompanied by few opens calls into question your store page or your post-installation experience. Opens without key action call onboarding into question.
Stickiness is measured quickly using the daily active users to monthly active users ratio (DAU/MAU) or weekly active users to monthly active users (WAU/MAU), which shows what share of your monthly audience shows up each day or each week.
Engagement and retention
After activation, two questions remain: do they come back, and how quickly do they perceive value each time? Track the time to first value, the moment when the application proves itself to a new user, then observe the cohorts for day 1, day 7, and day 30 retention.
Here is how to calculate retention:
- Day 1 retention = people who return on day 1 / people who started on day 0
- Day 7 retention = people who return on day 7 / people who started on day 0
- Day 30 retention = people who return on day 30 / people who started on day 0
- DAU/MAU stickiness = active people per day / active people per month
Utility applications generally stand around 20 percent of DAU/MAU. Forty percent signals solid engagement, and 60 percent or more suggests a product anchored in daily habits. Benchmarks exist to calibrate expectations, so that aiming for 60 percent stickiness for a utility application prepares for disappointment rather than for a stimulating challenge.
Healthy retention curves bend at the beginning, then stabilize as a loyal core settles in. Curves that continue to slide toward zero generally conceal friction points (slow screens, confusing navigation) or the absence of a valid reason to come back.
Cohort definitions include special cases that quietly distort the numbers. Time zones matter, because an open at 11 pm can count as day 0 or day 1 according to your rule. Reinstalls inflate retention when deduplication is poorly handled. And the move from anonymous status to authenticated status fragments journeys as soon as identity stitching is poorly configured.
Monetization
If your application sells products or services, measure whether engaged people convert and continue to spend at a reasonable pace. Track the move from the trial period to the paid subscription, or from the first purchase to repurchase, then integrate frequency and average basket.
Key monetization formulas:
- ARPU (average revenue per user) = total revenue / total active people
- ARPPU (average revenue per paying user) = total revenue / paying people
- LTV (customer lifetime value) = ARPU × average customer lifetime (in months)
Gather these elements into a simple view of customer lifetime value over time in order to make cohorts comparable and to verify whether efforts on onboarding or performance translated into revenue. A clean KPI report far surpasses an unnecessarily complex report in this specific case.
Quality and performance
Technical quality conditions each of the metrics presented above. Sessions and people without crashes, ANR events on Android, and cold-start times must appear on the same page as your funnel. When these metrics drop, funnels and retention drop with them, and the cause is far too often wrongly attributed to design changes.
Performance also produces a second-order effect. Android core vitals influence visibility on the Play Store, and on iOS, users reward fluid applications with better ratings. Interpret them as leading indicators situated upstream of activation and retention.
How to use this process
Go through the loop from left to right: acquisition, activation, engagement, monetization, with quality underlying these four stages. A drop observed at any level requires going back to the previous stage to find its origin. Fix the weakest link furthest upstream on which you can act quickly, then verify whether the cohorts stabilize and whether revenue follows.
Setting up mobile analytics the right way
Three prior agreements are required before integrating the slightest SDK: what you measure, the way in which you name it, and what constitutes a sufficient level of quality. Set these points, and the pipeline design becomes crystal clear. Receive the raw events exactly as the SDKs emit them so that anomalies remain verifiable, transform them into a clean layer with consistent names and types, then publish an organized layer grouping together the tables that teams actually consult, such as daily active users, cohort retention, and funnel stage summaries.
Event taxonomy and naming conventions
Twelve to 20 events associated with key tasks are enough to get started. As soon as your platform supports recommended events and parameters, use them in order to benefit, free of charge, from clearer reports.
Here is an example of an event schema for a signup funnel:
Naming in object-action format respecting consistent casing guarantees the readability of the whole. user_signed_up, cart_item_added, and checkout_started are suitable. ButtonClick1, signUp, and CHECKOUT are not. Inconsistent naming turns into a real tax on analysis much more quickly than most teams imagine, and by the time someone notices, it is months of disorganized data that must be cleaned up.
Parameters must answer the questions “which one” and “why.” An event view_item associated with item_id, category, and price can shed light on a dozen questions. Without these details, it makes it possible to address almost none of them.
Global properties require the same rigor. Associate app_version, platform, locale, and acquisition_source with each event in order to be able to segment your versions and your acquisition channels without having to rework your schema later.
Identity resolution
Assign an anonymous device identifier before login, then merge it with the person's user_id at the time of authentication. Special cases must be the subject of documented rules:
- At logout, decide whether a new anonymous identifier takes over or whether continuity is maintained
- At reinstallation, determine how to treat a user who returns with a new device identifier
- For offline events, design a retry logic that eliminates duplicates as soon as the connection is restored
Governance and quality
Define properties sparingly. Lasting attributes like plan_tier deserve their place; personally identifiable information (PII) must never have one.
Version your schema. Renaming events after the fact is a costly approach; therefore add new events without hesitation and deprecate the old ones by setting an obsolescence date and a documented migration plan.
Validate before any production release. Debug modes and test devices are there for that, and malformed events must interrupt the build rather than reach production. Continuous integration (CI) constitutes the appropriate checkpoint for this control.
Designate an owner. One person responsible for validating new events, ensuring compliance with naming conventions, and keeping the tracking plan alive in the form of an operational document rather than a simple relic.
Reading the data: funnels, paths, and cohorts
Once the events are transmitted, three views do most of the work, each answering a very distinct objective. Funnels make it possible to locate the place where the momentum runs out of steam. Go through them from left to right and identify the step combining a sharp drop and high traffic, because that is your first target.
Journeys show how users actually navigate, as opposed to what your design assumed. Compare users who converted to those who did not convert, while giving increased attention to the two steps preceding success, where small adjustments often pay off disproportionately.
Cohorts tell you whether a change lasted over time. Group users according to the week when they opened the application for the first time, then compare day-7 and day-30 retention between the groups. Post-change cohorts that retain more people constitute a victory that you can truly defend during a review.
Decision-driven segmentation
Segmenting data without an associated decision is a waste of time. Follow this sequence first:
- Formulate your hypothesis: "People acquired via paid social have lower D7 retention than those coming from organic"
- Define the segment precisely: acquisition_channel = paid_social vs acquisition_channel = organic
- Make sure of a minimum sample size: at least 1,000 people per segment to guarantee statistical confidence
- Define guardrail metrics: confirm that you are not buying retention at the expense of revenue or engagement
- Define the action: if the hypothesis is confirmed, what changes?
This last step is the one that teams tend to ignore, and it is by skipping it that segmentation becomes a pastime rather than a lever.
10 common mobile analytics mistakes to avoid
Most failures in mobile analytics are recurring mistakes. Here are those that waste the most time for teams:
- Sprawling event taxonomies make analysis slow and fragile. Keep a concise and versioned schema.
- Chasing installs while ignoring activation produces only illusory victories. Measure the journey leading to the first value.
- Excluding quality from the dashboard leads to interpreting performance drops as product problems. Place crash and latency indicators next to your conversion funnels.
- One-off exploits do not scale. If transformations and tests cannot be replayed, growth will break your reporting.
- The absence of an identity reconciliation strategy fragments user journeys. Define merge rules before deployment.
- Unvalidated event schemas allow malformed or missing events to corrupt downstream reports. Integrate validation into the CI pipeline.
- The absence of a governance process for modifying events encourages the proliferation of unnecessary events and metric drift. Assign responsibilities and require approval for any schema change.
- Diverging metric definitions between teams generate debates around dashboards that erode trust. Agree on formulas and edge cases before any development.
- Tracking everything “just in case” drowns the signal under the noise. Collect data with a clear intention.
- Postponing compliance and privacy to launch week accumulates regulatory debt that is costly to resolve. Integrate privacy protection from day one.
Best mobile analytics tools for 2026
Choosing a mobile analytics tool begins by defining your main question. Questions related to behaviors point toward product analytics platforms, questions linked to campaigns point toward attribution providers, and stability questions point toward crash and performance monitoring. Most teams end up using two or three of them, because no category answers all three perfectly at the same time.
Four criteria make it possible to distinguish the selected solutions from the others. The event model and identity management determine whether your user journeys from anonymous to identified remain intact. Privacy and consent controls determine whether you can comply with ATT and the GDPR without custom development. Data export and synchronization with the data warehouse determine whether your data remains portable. Governance features determine whether your schema survives into its second year.
Here is how the main options compare:
The free versions are suitable when the event volume remains below the announced caps, when identity reconciliation across applications is not required, and when standard retention windows cover your analysis horizon. The figures above reflect the plans published as of the publication date and are subject to change; therefore check them before committing yourself. Paid plans become necessary as soon as you need longer retention, export to a data warehouse, governance controls, or enterprise access management.
A few constraints deserve to be known before committing yourself. Firebase/GA4 samples data at high volumes and limits custom event parameters. Amplitude and Mixpanel both cap monthly events and reserve their advanced features for paid plans.Marketing analytics tools designed for attribution, notably AppsFlyer and Adjust, reserve detailed reporting for paid offerings.
The scenario changes the game. A start-up phase application having only one engineer should opt for the solution requiring the least effort and offering free crash tracking. A product-led scale-up that runs weekly experiments needs in-depth behavioral analytics and cohort tools. A consumer application strongly focused on advertising first needs attribution, then behavior analysis. Regulated companies should give more importance to governance, consent controls, and audit trails than to the number of features.
Whatever the combination chosen, the reports that matter will extend across several tools. Marketing analytics data, product events, and crash telemetry are rarely found in the same place; therefore plan for a layer making it possible to unify them. AI reporting tools can accelerate this task, and privacy-respecting attribution on iOS and Android allows you to remain compliant, while allowing marketing to obtain a fair measurement of performance.
Bringing mobile analytics together in Domo
You can run the entirety of this loop in Domo without cumbersome setup. Connect the data from your application as well as a few key marketing sources, then use Magic ETL and DataFlows to standardize your event schema and publish clean and reusable tables.
The platform is unified by design and modular by adoption, thus allowing a single product to solve a first problem before you extend your scope. As adoption progresses, the data, the logic, and the governance are defined a single time and reused everywhere, including in the workflow of ML model management. Domo operates directly on top of your existing cloud data platform rather than replacing it.
Each role accesses exactly what it needs. Field employees open pre-filtered data dashboards and ask questions in natural language from their phone, with AI agents operating according to the objectives and constraints defined by the teams. Executives track KPI dashboards without going through a computer, and the operational and executive views remain governed by the same definitions. Data engineers receive mobile alerts during pipeline failures, thus preventing your CFO from discovering the problem via a failing dashboard. The results of AI business analytics integrate directly into the tools that teams already use, and Domo Everywhere makes available to customers secure, white-label, and mobile-optimized embedded analytics.
Start small: a dozen key events, a funnel, a cohort, and a change targeting your highest attrition.

