Taking Control Of
Your Marketing Attribution
Attribution is the art and science of assigning credit to marketing touch‑points, be that a user, a lead, a sale, to understand which efforts drove the result. But as privacy intensifies and algorithms assume decision‑making, reliance on platform-reported metrics can mislead. To thrive in this changing environment, you must reclaim control by owning your data and building a first‑party‑driven attribution process.
What Is Attribution?
Attribution assigns credit for conversions (users, leads, orders) to marketing touchpoints (source, campaign, ad, placement). This lets you calculate ROI, make informed spend decisions, and understand where your marketing is most effective. But with rising privacy restrictions and platform automation, reliable attribution has become increasingly elusive.
Why You Cannot Trust Platform‑Reported Numbers
Greedy Platforms: One of the most pervasive issues in digital marketing today is the phenomenon of multiple platforms claiming credit for the same conversion. For example, a user might click a Meta ad in the morning, revisit the website later via a Google Search ad, and complete their purchase after receiving an email retargeting campaign. In this scenario, Meta, Google Ads, and your email platform will all report that they ‘drove’ the conversion—each using their own attribution logic, windows, and touchpoint rules. This is possible because attribution is measured independently within each ecosystem, without awareness of competing platform touchpoints. The result is heavily inflated conversion totals across your reporting dashboards, often exceeding 100% of actual sales. Without first-party data to reconcile these claims, you risk overvaluing certain channels, overspending, and missing key insights about how your customers really behave across their journey.
Heuristic Attribution: Many advertising platforms rely on proprietary, simplified attribution models that assign credit based on limited interaction metrics like last click, last view, or time spent on an ad. These heuristics are designed to make attribution appear clean and decisive, but in reality, they obscure the complexity of the user journey. A platform might allocate full credit to a final click moments before a conversion, entirely ignoring earlier brand exposures or influence from upper‑funnel channels like YouTube, display, or even offline marketing. This bias towards “closers” undervalues awareness‑driving efforts that played a foundational role in shaping consideration and intent. The consequence is that advertisers are often incentivised to double down on performance media while pulling back on the very touchpoints that first attracted or educated the user. Over time, this skews strategy, reduces brand equity, and leads to diminishing returns. Without a holistic attribution framework that spans platforms and considers all interactions, you risk optimising towards noise rather than true influence.
Opaque Algorithms: Campaign types like Google Performance Max (PMax) operate as opaque “black‑box” systems. Though asset‑level reports are now available, the attribution logic (how Google decides credit)is hidden. Google’s use of data-driven attribution (DDA) assigns fractional credit across multiple touchpoints—a theoretically sound idea, but practically dependent on internal models with hidden assumptions. Without transparency, it’s guesswork that obscures more than it reveals. Analyses of PMax show a troubling trend known as “kissed‑by attribution”: minuscule fractional credit (e.g. 0.01%) assigned to campaigns according to seemingly arbitrary heuristics, accounting for up to 64% of attributed conversions (smec). This artificially inflates performance metrics without true causal link.
Cannibalisation: Another note on PMax in particular is that it often competes with your own Search and Shopping campaigns for the same auction—on identical products, landing pages, or audiences. This can lead to self-cannibalisation and skewed attribution, with Google favouring PMax in ways that may misrepresent actual channel performance.
Own Your Attribution: A Better Path
To regain visibility and accuracy, build a deterministic-first, probabilistic‑corrected pipeline anchored in your first-party data.
Step 1: Track UTMs Rigorously
UTM parameters (Urchin Tracking Module) are short text tags appended to URLs that allow you to track the source and context of traffic visiting your website. When a user clicks on a link containing UTMs (whether from a paid advert, an email, or a social post) those parameters are captured by your analytics platform (e.g. Google Analytics 4, Snowplow, PostHog), providing crucial context around how that visitor arrived.
You can configure UTM parameters directly within most advertising platforms. For example:
In Meta (Facebook/Instagram), UTM parameters can be built during ad creation under the “Tracking” section.
In Google Ads, UTMs are added via the Final URL suffix using dynamic parameters.
For email platforms like Mailchimp or Klaviyo, UTMs can be automatically appended or customised at the campaign level.
By defining a consistent UTM structure across all platforms and campaigns, you enable clean and accurate stitching between marketing activity and website behaviour. Without this consistency, data will fragment, making attribution incomplete or misleading.
Here is an example structure:
Maintaining this structure ensures clean alignment between your marketing platforms and your analytics or data warehouse. It also provides the critical foundation for joining first-party event data (e.g. page views, purchases) with marketing metadata, supporting everything from campaign performance analysis to customer journey modelling.
Finally, be aware that some platforms auto-tag URLs using their own conventions. In these cases, enforce UTM overrides to ensure consistency across systems, this is especially important when multiple platforms interact in the same user journey.
Step 2: Unique Landing Pages
One of the most effective ways to enhance attribution is by creating unique landing pages for each campaign, channel, or target audience. These pages serve not only as conversion points but also as structured containers for tracking the provenance of traffic—helping you build a clean, end-to-end funnel from ad to action. When multiple campaigns or channels point to the same generic landing page, it becomes much harder to confidently attribute success. UTM parameters may still help, but inconsistencies in URL handling, form prefill, or user behaviour (e.g. returning via direct traffic) can lead to attribution loss.
In contrast, distinct landing pages:
Preserve campaign-level clarity: Each URL acts as a self-contained trace of source, medium, and intent.
Improve message match: Tailoring the content to match the ad copy, creative, and user expectation increases conversion rate, reducing bounce and reinforcing relevance.
Enable controlled testing: You can easily test variations in messaging, layout, or call-to-action by isolating traffic flows to specific pages.
Reduce cross-contamination: With dedicated URLs, you avoid overlapping audience traffic across campaigns, which can lead to mixed or double-counted attribution.
Every landing page URL should carry UTM parameters that reflect its campaign context.
How To Tie Together
On that page, embed hidden form fields or backend logging scripts that capture the UTM parameters at the moment of conversion (e.g. form submission, sign-up, or purchase). This ensures the marketing data is preserved and passed into your CRM or backend system alongside the user’s details. These identifiers (such as campaign ID or ad ID)can then be joined to transaction data, enabling deterministic attribution.
For lead generation or subscription models, assign a unique form ID or name per landing page, allowing you to attribute submissions even if UTM data is missing (e.g. in direct revisit scenarios).
Step 3: First‑Party Event Collection
The core of reliable attribution is the ability to connect a marketing touchpoint (such as an advert click) to a downstream outcome (such as a transaction, lead, or app install). This requires comprehensive and consistent event tracking, on both web and mobile. For websites, tools such as Google Analytics 4 (GA4), Snowplow, or Segment allow you to capture detailed user interactions including pageviews, form submissions, button clicks, and e-commerce transactions. At a minimum, you should configure your tracker to:
Capture UTM parameters from inbound URLs.
Persist these values across the user session or store them in cookies or session storage.
Log key events (e.g. purchase, lead, registration) with the associated UTM data attached.
- Assign first-party identifiers, such as user_id, transaction_id, or lead_id, to these events.
This identifier is critical, it provides the bridge between frontend user behaviour and backend system data. With it, you can join an ad click to a transaction in your database, creating a fully traceable and verifiable marketing funnel.
Mobile Tracking: Why an MMP Still Matters
On mobile apps, tracking attribution becomes more complex due to platform restrictions, especially on iOS. Apple’s App Tracking Transparency (ATT) framework requires user opt-in for tracking, and many users decline. In this environment, deploying a Mobile Measurement Partner (MMP) such as AppsFlyer, Adjust, or Branch remains best practice. These tools:
Capture app installs, in-app events, and re-engagements.
Provide probabilistic attribution methods when deterministic identifiers (like IDFA) are unavailable.
Help de-duplicate conversions across channels.
Integrate with ad networks to provide more detailed conversion data than what native SDKs often allow.
While some may question whether GA4’s mobile SDK is enough, in practice it lacks the depth of attribution capability that MMPs offer, particularly when working with mobile-first businesses, app-only funnels, or multi-touch campaigns across mobile and web.
Cross-Platform and Cross-Device Linkage
In order to track journeys across web and apps, you need to unify event data. This requires:
Consistent identifiers (e.g. hashed email address or CRM ID).
Robust event naming conventions.
Cross-domain or cross-device tracking logic (especially when a journey begins on mobile and completes on desktop, or vice versa).
This unified view ensures your attribution model can account for complex user journeys, rather than treating each platform as a silo.
Expert help is only a call away. We are always happy to give advice, offer an impartial opinion and put you on the right track. Book a call with a member of our friendly team today.
Step 4: Backend Join & Clean Data
Once you have reliable frontend event tracking in place, capturing UTM parameters, timestamps, page paths, and user identifiers, the next step is to integrate this with your backend data to establish a complete picture of attribution. Ad platforms and analytics tools can only see what happens within their own domain. Your business systems (such as your order management system, subscription platform, or CRM) hold the definitive record of whether a conversion took place. Attribution accuracy depends on linking these two worlds.
By extracting event logs (such as a pageview or form submission that includes UTM data) and joining them to actual conversions in your database, you form a deterministic attribution layer: a clear, timestamped chain of events showing how a user arrived, what campaign they came from, and when they converted.
This process removes the guesswork. It ensures that:
You only attribute real, completed transactions or qualified leads—not just clicks or form interactions.
You control the attribution logic, not the ad platforms.
Your reports align with actual business outcomes, not inflated metrics from multiple sources claiming credit.
How It Works in Practice
Extract event data from your tracking tool. Whether you are using GA4, Snowplow, or Segment, set up an export of raw event data into a warehouse such as BigQuery, Redshift, or Snowflake.
Extract conversion data from your backend systems. Pull data from your e-commerce platform, CRM, booking system, or any other platform where conversions are recorded. Ensure that identifiers such as transaction_id or user_email are available.
Perform the join Using SQL (or tools like dbt), write logic to link the marketing events to their corresponding backend outcomes.
Assign attribution rules. Once joined, you can apply your own attribution logic—whether that’s first click, last click, position-based, or a custom model that fits your business cycle.
Build a central attribution table. This becomes your single source of truth for performance reporting, feeding dashboards, cohort analysis, ROAS calculations, and lifetime value modelling.
Step 5: Blend Probabilistic Modelling
Even with robust tracking in place, full visibility is no longer guaranteed. A growing share of users arrive on your site or app with partial or missing attribution data, due to privacy settings, browser restrictions, or platform limitations. For example:
On iOS devices, since the introduction of App Tracking Transparency (ATT) in version 14.5, users must explicitly consent to being tracked across apps and websites. The majority choose not to.
On web, third-party cookies are being phased out, limiting retargeting and cross-site attribution.
Many browsers, such as Safari and Firefox, block tracking scripts or delete cookies within hours.
This creates a significant attribution gap: you know a conversion occurred, but you cannot link it back to a specific campaign, channel, or journey.
Using Probabilistic Methods to Bridge the Gap
Rather than letting this data disappear into a black hole, you can leverage your known, first-party data (the users you can track) to build probabilistic models that estimate the performance of your campaigns across the entire user base. These models use statistical inference to predict the likely source of a conversion, based on observable patterns. Some commonly used techniques include:
Uplift Modelling: Measures the incremental impact of a campaign by comparing conversion rates between exposed and unexposed audiences. This can be used to isolate the true value of paid activity, even when some users cannot be directly tracked.
Regression Analysis: Establishes relationships between marketing inputs (such as spend, impressions, or click volume) and outputs (such as revenue or leads), controlling for confounding factors like seasonality or product launches.
Bayesian Inference: Allows you to incorporate prior knowledge (e.g. past campaign performance) with current, incomplete data to produce probability-based attribution estimates that adjust dynamically over time.
Markov Chains and Path Modelling: Simulates the likelihood of a user converting based on their sequence of touchpoints, estimating the relative contribution of each channel or step—even when individual paths are partially obscured.
Using Probabilistic Methods to Bridge the Gap
Rather than letting this data disappear into a black hole, you can leverage your known, first-party data (the users you can track) to build probabilistic models that estimate the performance of your campaigns across the entire user base. These models use statistical inference to predict the likely source of a conversion, based on observable patterns.
Practical Applications
Attribution Correction: Blend deterministic data (users with full UTM or ID coverage) with probabilistic models to estimate how many “invisible” conversions each channel is likely to have generated.
Spend Allocation: Use modelled outcomes to guide budget decisions when real tracking data is incomplete or skewed—particularly useful for upper-funnel, mobile-heavy, or awareness campaigns.
Performance Forecasting: Predict expected outcomes under different spend scenarios, even when attribution gaps exist.
The best strategy is a hybrid attribution model: deterministic where data is available, probabilistic where it is not. This approach acknowledges the reality of today’s privacy-first environment, while still enabling robust measurement, campaign optimisation, and strategic decision-making.
Step 6: Incrementality & Experimental Design
Incrementality refers to the portion of conversions that are truly caused by your marketing activity. In other words, it measures how many additional conversions you gained as a direct result of running a particular campaign or channel, compared to how many would have occurred without it. For example, if a campaign generates 1,000 conversions but 800 of those users would likely have converted anyway via organic search or direct traffic, then the campaign’s incremental value is only 200 conversions. This is a critical distinction: traditional attribution may assign credit to all 1,000, but only testing can reveal the true causal impact. Without measuring incrementality, you risk mistaking correlation for causation. This leads to:
Overinvestment in low-impact channels that appear to perform well but actually intercept users who were already intent on converting.
Underinvestment in upper-funnel or awareness activity that supports the journey but rarely gets credit in click-based models.
Inaccurate ROI calculations, bloated attribution reports, and misguided strategy decisions.
How to Test Incrementality
There are several ways to structure incrementality tests depending on the channel, platform, and scale:
A/B Holdout Testing Create two comparable audience groups: one exposed to your ads (test group), and one completely withheld from them (control group). Measure conversion rates across both. The difference is your incremental lift, the added value driven by the campaign.
Geo-Testing Run a campaign in select geographic regions while holding out others. Useful when platform-level control groups are not possible or when testing at scale.
Budget Suppression Temporarily pause or reduce spend on a specific channel to observe whether conversions decrease. If performance remains unchanged, the channel may not be adding true incremental value.
Platform Lift Studies Some platforms, such as Meta or Google, offer built-in lift tests that automate randomisation and control group creation. While these can be useful, treat them cautiously as they still operate within the vendor’s data ecosystem
Conclusion
If you are spending money on marketing, you can simply not accept that attribution is impossible…that is not the case! The real path to clarity is building a controlled, deterministic attribution foundation anchored in first-party data, augmented with probabilistic models for gaps. This pipeline becomes your true ROI engine, resilient to black-box systems and privacy disruption.
If you need help building out your marketing analytics, why not talk to the experts at 173tech?
