Attribution is the process of crediting channels for conversions. Last-click attribution credits the last channel before conversion. First-click attribution credits the first channel. But reality is more complex: multiple channels contributed to the decision. Cross-channel attribution models aim to fairly distribute credit across all contributing channels. This chapter covers different attribution models and when to use each.
Attribution Models Explained
Last-Click Attribution
The channel of the last touch before conversion gets 100% credit. Example: organic search → email → paid ads → conversion = paid ads gets 100% credit.
Advantage: simple, easy to implement, favors bottom-funnel channels. Disadvantage: ignores supporting roles, undervalues awareness channels.
First-Click Attribution
The channel of the first touch gets 100% credit. Example: organic search → email → paid ads → conversion = organic search gets 100% credit.
Advantage: favors awareness channels, easy to implement. Disadvantage: ignores nurture and conversion roles, inflates credit for early touches.
Linear Attribution
Credit is split equally among all touches. Example: organic search → email → paid ads → conversion = each gets 33% credit.
Advantage: fair to all channels, accounts for multiple touches. Disadvantage: doesn't account for which touches matter most.
Time-Decay Attribution
Recent touches get more credit than old touches. Example: organic search (30 days ago) gets 20%, email (10 days ago) gets 30%, paid ads (1 day ago) gets 50%. Total = 100%.
Advantage: realistic (recent interactions are usually most influential), balances awareness and conversion. Disadvantage: more complex to implement.
Custom/Algorithmic Attribution
Credit is allocated based on historical data analysis. Machine learning identifies which touches are most predictive of conversion. Example: data shows that when email is in the sequence, conversion rate is 15% higher. Email gets extra credit.
Advantage: most accurate, accounts for interaction effects. Disadvantage: requires data and technical sophistication.
Choosing an Attribution Model
Early stage (just starting): use last-click. Simple, implementable immediately. Undervalues awareness, but good starting point.
Growing company: use time-decay. More accurate than last-click, accounts for multiple touches, relatively simple.
Mature company: use custom/algorithmic. You have enough data and team sophistication to implement advanced models.
Best practice: track multiple models simultaneously. Report last-click to executives (simplicity), time-decay for strategy (balance), first-click for awareness metrics. This gives complete perspective.
Implementing Attribution Models
Step 1: Choose your model. Start with time-decay unless you have a reason to use something else.
Step 2: Define your attribution window. How many days before conversion do you track touches? 7, 14, 30, 90? Use 30 as default.
Step 3: Configure in your analytics platform. Most platforms support multiple attribution models. Google Analytics and most product analytics platforms have native support.
Step 4: Compare models. Run all models on your data. Compare results. Where do they differ? Does it change your strategy? If different models suggest different channel priorities, dig deeper to understand why.
Challenges in Cross-Channel Attribution
Channel cannibalization: if you run both organic and paid search for the same keyword, which channel gets credit? Both may claim the conversion. Solution: set clear rules (first click = organic, last click = paid) and apply consistently.
Missing data: offline interactions (phone calls, in-person meetings) aren't tracked in analytics. Solution: import offline conversion data into analytics when possible. Acknowledge that some conversions have missing attribution data.
Privacy and tracking limitations: iOS privacy changes, ad blockers, and privacy tools reduce tracking capability. Some user journeys are partially invisible. Solution: use first-party data whenever possible. Acknowledge tracking limitations in your reporting.
Which attribution model should I use for my business?
Time-decay is the best default for most businesses. If your sales cycle is short (1-2 days), use last-click. If it's long (30+ days), use first-click or linear. Test your top 3 models for 1 month and see if recommendations change. If they don't, stick with the simplest (last-click). If they do, use the most realistic one for your business.
How do I explain attribution models to non-technical stakeholders?
Use an example: 'Three people helped close a $10k deal: person A introduced us (awareness), person B answered questions (consideration), person C signed the contract (decision). Last-click gives all $10k to person C (unfair). First-click gives all to person A (ignores B and C). Time-decay gives 20% to A, 30% to B, 50% to C (realistic).' Use this framework to explain why time-decay is more fair.
Should I use different attribution models for different channels?
No, stick with one model across all channels so they're comparable. Different models for different channels creates inconsistency. Use one model consistently, then analyze within that model: organic search's role vs. email's role, etc. If you need different models for different purposes, run them separately (reporting and strategy), not mixed.
How do I update my attribution model if my sales cycle changes?
Re-analyze your data: how long is the average sales cycle? If it changed from 15 days to 45 days, your 30-day attribution window may be cutting off important touches. Update to 45-60 days. Re-run analysis and see if channel priorities change. Update your model and document the change.
Can I use attribution data to predict which campaigns will be successful?
Partially. Historical attribution shows which channel combinations work. If successful customers came through organic → email → paid, new campaigns following that pattern are likely successful. Use this as a hypothesis: invest in those sequences. But each campaign is different; test and measure rather than rely solely on historical patterns.
How do I handle offline conversions in my attribution model?
Import offline data into analytics: phone calls, in-person meetings, contracts signed. Match by customer ID (email, phone, or name) to web sessions. Online touches appear in attribution; offline conversions are the outcome. This creates a more complete picture. Note: offline attribution is harder; focus on the biggest offline deals first.