Quick answer: An attribution model is a rule for assigning credit for a conversion across the marketing touchpoints that led to it. Last-click gives all credit to the final interaction before the conversion; first-click credits the introduction; linear, time-decay and position-based split credit in different patterns. Use model comparison in Google Analytics to see how channels shift under each rule, and do not make budget cuts from last-click alone when your path to purchase is long.
Someone searches your brand, clicks a blog post from organic search, returns a week later via a retargeting ad, then converts after an email. Which channel “did it”? Attribution models are formal answers to that question. None is perfect. Each is a lens.
This guide explains the standard models available to most marketers in 2018 (especially in Google Analytics), their biases, and how to use them in decisions.
Why last-click dominates, and why that is a problem
Many ad platforms and default analytics reports emphasise the last non-direct click before a conversion. It is easy to compute and easy to optimise toward.
The bias is obvious: channels that introduce new demand (content, video, display prospecting, some organic discovery) look weaker than they are, while brand search and retargeting look stronger because they sit at the end of paths they did not always create.
If you cut the “weak” top-of-funnel channel based only on last-click ROAS, you may be cutting the reason the late-click channels had anything to convert.
Common models
Last click (last non-direct click) 100% credit to the final touchpoint before conversion (often ignoring direct). Good for understanding closer performance. Bad as the only budget brain.
First click 100% credit to the first known touchpoint. Highlights discovery. Ignores nurturing and closing work.
Linear Equal credit to every touchpoint in the path. Fair-minded, but treats a tiny bounce visit the same as a serious demo request session.
Time decay More credit to touchpoints closer in time to the conversion. Useful when recent interactions matter more, such as short sales cycles.
Position-based (U-shaped) Often 40% first, 40% last, 20% split among the middle. A compromise when you care about introduction and close.
Data-driven attribution (DDA) Available in some Google Analytics 360 and Google Ads setups when you have enough conversion volume. Uses your data to estimate contribution rather than a fixed rule. Not every account qualifies in 2018.
Google Analytics also offers a Model Comparison Tool in Multi-Channel Funnels so you can put two models side by side. Use it.
Paths, lookbacks and Direct
Attribution only works on the paths you can see. Limitations:
- Cookie deletion and cross-device paths break paths
- Safari and privacy changes increasingly constrain identifiers
- Offline closes (phone, retail) need import strategies or they stay invisible
- Direct traffic often means a dark prior touch (typed URL, bookmark, unclear email)
Set lookback windows consciously. A 7-day window on a B2B deal that takes 90 days will under-credit early content. A 90-day window on impulse ecommerce may over-credit ancient visits.
For GDPR-era consent and analytics configuration, see GDPR and Google Analytics and IP anonymization.
How to choose a working approach
- Map your real sales cycle. Days or months? High consideration or impulse?
- Use last-click for tactical optimisation of bottom-funnel campaigns, not for proving that content is worthless.
- Compare models monthly for channel mix discussions with leadership.
- Align definitions with ad platforms. Google Ads has its own attribution settings; they may not match Analytics. Know which one people are quoting in meetings.
- Supplement with experiments. Geo tests, incrementality tests and simple on/off holds beat model theology when budgets are large enough.
Also verify that conversions are tracked correctly before arguing about models. Broken AdWords conversion tracking makes every model wrong.
Cross-domain and multi-property issues
If checkout lives on a different domain, paths break unless cross-domain tracking is set up. Attribution meetings often invent drama that is really a measurement gap.
Similarly, if paid landing pages sit on vanity domains without consistent tracking, first-click discovery gets mislabelled.
Reporting habits that help
- Report assisted conversions alongside last-click conversions for upper channels
- Separate brand vs non-brand search before declaring paid search “efficient”
- Watch time lag and path length reports in Multi-Channel Funnels
- Annotate big campaigns so model shifts have context
What attribution cannot do
- Prove absolute causality by itself
- Fix poor offers or slow sites
- Replace qualitative customer research
- Survive bad UTM hygiene (everything dumped into Direct or (other))
Clean campaign tagging is boring and essential.
Offline and phone sales
If a meaningful share of revenue closes by phone, digital attribution will always be incomplete until you import those conversions or use call tracking that ties back to sessions. Even a simple “how did you hear about us” field, used consistently, beats pretending last-click ads closed deals that sales closed on a call.
CRM reverse ETL and offline conversion import into Google Ads exist for teams ready to operationalise them. Start with honest reporting of the gap if you cannot import yet.
Agency and in-house politics
Attribution debates are often budget debates in costume. Agree up front which model you will use for quarterly planning, and which reports are informational only. Put the agreement in the analytics brief so a new channel owner cannot redefine success mid-flight.
FAQ
Which model is correct?
None universally. Prefer a primary model for day-to-day ops (often last-click or data-driven where available) plus periodic comparison views for planning.
Should we switch Google Analytics to first-click globally?
Usually no. Keep defaults understandable for the team, and use comparison tools for alternative views.
Is last-click fine for a small shop?
It can be good enough when paths are short and budgets are simple. Revisit when you add content programmes, affiliates or long remarketing cycles.
A simple monthly ritual
Once a month, export or screenshot model comparison for your top five channels, note path length and time lag, and write three bullets for leadership: what last-click says, what an assisted or position-based view says, and what you will do next. Keep the ritual short so it actually happens. Attribution without a decision habit is trivia.
Credit is a model, not a moral verdict
Attribution models explain how you choose to divide credit, not who “deserves” a thank-you note. Learn what each model emphasises, compare them, fix tracking gaps, and spend based on a mix of model evidence and real-world tests.
If you want help cleaning up conversion tracking and reporting so channel debates are based on better data, our Know what’s working service covers analytics setup, attribution reviews and the dashboards teams actually use.
