Attribution modelling mistakes distort your understanding of which channels actually drive conversions, leading to misallocated budgets and flawed campaign decisions. This guide walks through the most damaging errors practitioners make when building or interpreting attribution models, from single-touch oversimplification to ignoring cross-device journeys and misattributing assisted conversions.
Last-click attribution assigns 100 percent of conversion credit to the final touchpoint before a sale or lead, which systematically undervalues every earlier interaction. This model inflates the apparent value of direct traffic, branded search, and retargeting while starving awareness channels like display, video, social prospecting, and top-of-funnel content of budget justification. The mechanics are simple: a user discovers your brand through a YouTube pre-roll, researches via organic search a week later, clicks a Facebook retargeting ad the following day, then converts by typing your URL directly. Last-click gives all credit to direct, ignoring the three earlier assists. The danger is compounded when optimization algorithms inside Google Ads or Meta are fed last-click conversion data—they learn to chase bottom-funnel intent exclusively, creating a self-reinforcing loop that chokes off new customer acquisition. Switching to a position-based or time-decay model immediately surfaces the hidden contribution of upper-funnel tactics, but many practitioners stick with last-click out of inertia or because it matches the simplicity of legacy CRM tracking.
Modern buyers research on mobile during commutes, compare options on tablets at home, and convert on desktop at work, yet cookie-based attribution models treat each device as a separate user. When you ignore cross-device behaviour, mobile traffic appears to have terrible conversion rates because the final transaction happens elsewhere, leading to underinvestment in mobile campaigns and creative. Google Analytics 4 uses probabilistic and deterministic signals to stitch user journeys across devices, but the data is only as good as your login prompts, newsletter signups, and account-creation flows that create authenticated touch points. In Canada, where bilingual content and regional targeting often fragment traffic across language-specific landing pages, cross-browser fragmentation compounds the problem—users switching from Chrome to Safari lose cookie continuity unless you implement server-side tracking or leverage Google's User-ID feature. The fix requires enabling cross-device reporting in GA4, auditing your conversion funnels for login opportunities, and layering in probabilistic matching where deterministic links are sparse. Without this, mobile and tablet campaigns absorb the cost of early research but never receive credit for the desktop conversions they enable.
Data-driven attribution uses machine learning to assign fractional credit based on the incremental contribution of each touchpoint, but the algorithm requires a minimum threshold of conversions and path diversity to produce stable weights. Google Ads data-driven attribution needs at least 3,000 clicks and 300 conversions in the attribution window within 30 days to activate; below that, the model either falls back to last-click or produces erratic coefficients that swing wildly month to month. Forcing a data-driven model onto a seasonal B2B account with 40 conversions per quarter introduces more noise than insight—the model may overweight a single high-performing blog post one month, then ignore it the next as sample variance shifts. The alternative is a rule-based model like linear, time-decay, or position-based, which applies consistent logic regardless of volume. Linear splits credit evenly across all touchpoints, time-decay gives exponentially more weight to recent interactions, and position-based allocates 40 percent to first touch, 40 percent to last, and 20 percent divided among middle assists. These deterministic models lack the precision of machine learning but produce interpretable, stable results when conversion volume is thin or highly cyclical.
Attribution models that only track online form fills and e-commerce checkouts ignore a substantial share of revenue for local businesses, professional services, healthcare, and B2B companies where phone calls, in-store visits, and offline invoicing close the loop. A personal injury lawyer running Google Local Services Ads and organic SEO might see 60 percent of leads arrive via phone, yet if those calls are not imported into Google Analytics or the CRM, the attribution model credits zero value to the campaigns that drove them. Google Ads call tracking extensions and third-party platforms like CallRail or Invoca capture the referring keyword, campaign, and landing page for inbound calls, then push that conversion data back into Ads via offline conversion imports or the Measurement Protocol. For brick-and-mortar retailers, Google's store visits metric uses aggregated, anonymized location data to estimate in-store traffic driven by digital campaigns, though the feature requires thousands of impressions and a verified Google Business Profile. Canadian businesses serving both anglophone and francophone markets need to ensure bilingual call-centre tags and CRM fields are normalized before import, or attribution breaks at the language boundary. Skipping offline conversion tracking creates a blind spot that systematically undervalues channels with high call or visit intent, leading to budget cuts for the very tactics that drive foot traffic and consultations.
The attribution window defines how far back in time a touchpoint can receive credit for a conversion, and both extremes introduce bias. A seven-day click window for a high-consideration B2B SaaS product misses the prospect who clicked a LinkedIn ad on day eight, researched competitors for two weeks, then converted via branded search on day twenty—the model ignores the LinkedIn assist entirely. Conversely, a 90-day window for an impulse e-commerce product dilutes signal by crediting a display impression from three months ago that the buyer has long forgotten, inflating the apparent value of stale upper-funnel tactics. Google Analytics 4 defaults to a 90-day click and 1-day view-through window for conversions, while Google Ads uses 30 days for search and 90 for YouTube and display. The correct window length depends on your actual sales cycle: analyze the distribution of days-to-conversion in your CRM or GA4's path analysis reports, then set the window to capture 80-90 percent of conversions without extending so far that you credit noise. For Canadian seasonal businesses—ski resorts, tax preparedness services, summer tourism—consider dynamic windows that expand during peak research phases and contract during low intent periods, though implementation requires custom conversion import scripts.
A view-through conversion occurs when a user sees but does not click a display, video, or social ad, then later converts via another channel. Excluding view-throughs from attribution systematically undervalues brand awareness campaigns, especially YouTube TrueView, programmatic display, and Facebook/Instagram reach campaigns that rely on passive exposure rather than immediate clicks. The standard view-through window is one day, meaning the impression must occur within 24 hours of the conversion to earn credit, though you can extend this to 30 days in Google Ads for upper-funnel video campaigns. The risk of over-crediting is real—a user who saw your pre-roll ad three weeks ago and converted after a branded search probably does not owe much to that impression—so view-through attribution works best when paired with a short window and frequency caps to avoid crediting accidental exposures. Assisted conversions, tracked in GA4's conversion paths report and Google Ads attribution reports, reveal touchpoints that did not close the sale but influenced the journey. A blog post that ranks for an informational query, earns a click, but sees the user convert two days later via email deserves partial credit as an assist. Ignoring assists penalizes content marketing, organic social, and educational channels that prime the buyer but rarely trigger immediate conversion.
Attribution reports show which channels earn credit, but without layering in campaign structure, audience segmentation, and creative variation, the numbers mislead. A time-decay model might show that organic search receives 45 percent of conversion credit, but if you do not segment by branded versus non-branded queries, you will not realize that branded search—users already aware of your company—accounts for 40 of those 45 points, leaving non-branded SEO with minimal incremental contribution. Similarly, attributing value to Facebook Ads without separating cold prospecting, warm retargeting, and lookalike audiences lumps together campaigns with wildly different roles in the funnel. The fix is to build attribution segments that mirror your campaign taxonomy: separate brand from non-brand, new visitors from returning, top-of-funnel from bottom, and geo-targeted from national. In GA4, use custom channel groupings and UTM parameters to enforce this structure; in Google Ads, apply labels and filters before exporting attribution data. Canadian businesses running parallel English and French campaigns must also verify that language-specific landing pages and UTM tags are correctly mapped, or the attribution report will misattribute cross-language journeys as separate users rather than a single bilingual buyer path.
Over-reliance on last-click attribution is the most common and damaging error. It gives all credit to the final touchpoint, systematically undervaluing awareness and consideration channels like display, video, content, and prospecting social ads. This distorts budget allocation by inflating the apparent value of branded search, direct traffic, and retargeting while starving upper-funnel tactics that actually initiate the buyer journey.
Cookie-based attribution treats each device as a separate user, so when someone researches on mobile and converts on desktop, the mobile touchpoint often receives zero credit. This makes mobile traffic appear to have poor conversion rates, leading to underinvestment in mobile campaigns. Google Analytics 4 cross-device reporting and User-ID features help stitch these journeys together, but require login flows or authenticated events to work reliably.
Data-driven attribution requires sufficient conversion volume and path diversity to produce stable weights—typically at least 3,000 clicks and 300 conversions within 30 days in Google Ads. Accounts with seasonal traffic, long sales cycles, or fewer than 100 conversions per month should use rule-based models like time-decay or position-based instead, which apply consistent logic regardless of sample size and avoid the erratic coefficients that plague low-volume machine learning.
Many businesses close deals via phone calls, in-store visits, or offline invoicing that never touch an online checkout. If these conversions are not imported into your analytics or ad platforms, attribution models only see a fraction of revenue and systematically undervalue the channels that drive calls and foot traffic. Call tracking platforms and Google's store visits metric help close this gap, but require integration between your CRM, call centre, and digital tracking stack.
Analyze the days-to-conversion distribution in your CRM or GA4 path reports to see how long buyers actually take to convert. Set the window long enough to capture 80-90 percent of conversions—often 30 days for e-commerce, 60-90 days for B2B—but not so long that you credit touchpoints from months ago that no longer influence the decision. Seasonal businesses may need dynamic windows that expand during peak research periods.
A view-through conversion happens when someone sees but does not click a display, video, or social ad, then later converts through another channel. Excluding them undervalues brand awareness and upper-funnel campaigns, especially YouTube and programmatic display. Use a short view-through window—one to seven days—to avoid over-crediting stale impressions, and pair it with frequency caps to ensure you are measuring genuine exposure, not accidental background views.