Attribution Models Explained: Understand Where Your Sales Really Come From

The Attribution Problem

A customer sees your Facebook ad on Monday. They click and browse your store but do not buy. On Wednesday, they see your Google retargeting ad and click through again. They still do not buy. On Friday, they search your store name on Google, click an organic result, and complete a purchase.

Who gets credit for the sale? Facebook says "I introduced them." Google Display says "I reminded them." Google Search says "I closed them." Each platform claims the conversion, and if you add up their reported revenue, it exceeds your actual revenue.

This is the attribution problem. It exists because customer journeys involve multiple touchpoints, and no platform has complete visibility into the full journey.

Why Attribution Matters for Budget Decisions

Attribution determines how you allocate budget across channels. If you believe Google retargeting drove a sale, you invest more in Google retargeting. If you believe Facebook prospecting drove it, you invest more in Facebook.

Incorrect attribution leads to incorrect budget allocation, which means money flowing to channels that get credit but do not actually drive results, while channels that do the heavy lifting get underfunded.

Getting attribution approximately right is one of the highest-leverage skills in digital advertising.

Common Attribution Models

Last-Click Attribution

How it works: 100% of credit goes to the last ad the customer clicked before purchasing.

Example: Customer clicked a Facebook ad on Monday and a Google ad on Friday before buying. Google gets 100% credit.

Advantages:

• Simple and easy to understand
• Clearly identifies the final conversion trigger
• Good for measuring direct-response channels

Disadvantages:

• Ignores all touchpoints before the final click
• Dramatically undervalues prospecting and awareness channels
• Overvalues retargeting and branded search (which often get the last click)

Who it favors: Google branded search, retargeting campaigns, email marketing

First-Click Attribution

How it works: 100% of credit goes to the first ad the customer interacted with.

Example: Customer first clicked a Facebook ad on Monday before eventually buying on Friday. Facebook gets 100% credit.

Advantages:

• Values the channel that introduced the customer
• Better reflects the importance of prospecting and awareness

Disadvantages:

• Ignores everything that happened after the first interaction
• Overvalues channels that drive curiosity clicks but not purchases
• Undervalues retargeting and closing channels

Who it favors: Facebook/Instagram prospecting, display advertising, content marketing

Linear Attribution

How it works: Credit is split equally across all touchpoints in the customer journey.

Example: Customer interacted with 4 touchpoints before buying. Each gets 25% credit.

Advantages:

• Acknowledges that every touchpoint contributes
• No single channel is overvalued or undervalued
• Simple to understand

Disadvantages:

• Treats all touchpoints as equally important (they are not)
• The first ad view and the final purchase click probably do not contribute equally
• Can dilute the signal for high-performing channels

Time-Decay Attribution

How it works: Touchpoints closer to the purchase receive more credit. The final touchpoint gets the most, and each previous touchpoint gets progressively less.

Example: Four touchpoints before purchase. The most recent gets 40%, the one before 30%, then 20%, then 10%.

Advantages:

• Recognizes that recent touchpoints often have more influence
• Balances credit across the journey while weighting toward conversion

Disadvantages:

• May undervalue the initial discovery touchpoint that started the journey
• The time-decay curve is somewhat arbitrary

Data-Driven Attribution (DDA)

How it works: Machine learning analyzes all conversion paths and assigns credit based on the actual statistical contribution of each touchpoint.

Example: The algorithm determines that Facebook prospecting drives 35% of conversion value, Google retargeting drives 40%, and organic search drives 25%, based on analyzing thousands of conversion paths.

Advantages:

• Most accurate model when sufficient data exists
• Adapts to your specific business and customer journey
• Does not rely on arbitrary rules

Disadvantages:

• Requires significant conversion volume (Google requires 300+ conversions/month)
• Black box since you cannot see exactly how credit is assigned
• Platform-specific DDA only sees that platform's touchpoints (Meta's DDA does not see Google touchpoints)

Who uses it: Meta's default attribution, Google Ads (when conversion volume is sufficient)

Platform Attribution Windows

Each platform defines how long after an ad interaction it will claim credit for a conversion:

Meta (Facebook/Instagram)

Default: 7-day click, 1-day view

• If someone clicks your ad and buys within 7 days, Meta claims the conversion
• If someone views your ad (without clicking) and buys within 1 day, Meta claims the conversion

Other options: 1-day click, 7-day click only (no view), 28-day click (no longer default but available for reporting)

Google Ads

Default: 30-day click for most campaign types

• If someone clicks your ad and buys within 30 days, Google claims the conversion

TikTok

Default: 7-day click, 1-day view (same as Meta)

The Over-Counting Problem

When you run ads on multiple platforms, each one independently claims conversions. A customer who sees ads on both Meta and Google and then purchases will be counted by both platforms.

Real example: Your store made $5,000 in revenue last week. Meta reports $4,000 attributed revenue. Google reports $2,500 attributed revenue. Total platform-reported revenue: $6,500. Actual revenue: $5,000. That is $1,500 in double-counted conversions.

This is normal and unavoidable with multi-platform advertising. It is why blended metrics are essential.

Practical Attribution Strategy

Step 1: Use Blended Metrics as Your North Star

Calculate your blended CPA and blended ROAS across all advertising channels using your actual revenue (from Stripe or your payment processor), not platform-reported revenue.

Blended ROAS = Actual total revenue / Total ad spend across all platforms

This number cannot be inflated by double-counting. It is the truest measure of your advertising efficiency.

Step 2: Use Platform Metrics for Relative Comparison

While individual platform metrics may overcount, they are useful for comparing performance within a platform. If Ad Set A has a $12 CPA and Ad Set B has a $20 CPA on Meta, Ad Set A is almost certainly more efficient, even if the absolute numbers are inflated.

Use platform metrics for:

• Comparing ad sets, creatives, and audiences within a platform
• Identifying trends (is CPA rising or falling?)
• Day-to-day campaign management

Step 3: Run Incrementality Tests

The gold standard for attribution is incrementality testing: measuring what happens when you turn a channel on or off.

Simple incrementality test:

1. Record your baseline revenue for a normal week
2. Pause Google Ads for one week
3. Measure the change in total revenue (not just Meta-attributed revenue)
4. The difference between baseline and test-week revenue is Google's true incremental contribution

Example result: Total revenue drops from $5,000 to $4,200 when Google is paused. Google's true incremental contribution is $800, even though Google was reporting $2,500 in attributed revenue.

Warning: Incrementality tests are disruptive (you lose revenue during the test) and should be done during low-stakes periods, not during Black Friday.

Step 4: Apply the 80/20 Rule

Perfect attribution is impossible. Do not let the pursuit of perfect data paralyze your decision-making.

Follow the 80/20 rule: blended metrics plus basic incrementality understanding gets you 80% of the way to optimal budget allocation. The remaining 20% of precision requires sophisticated multi-touch attribution tools that cost more to implement than the additional insight is worth for most stores.

Attribution for Different Business Stages

Early Stage (Under $1,000/month ad spend)

• Use platform-default attribution settings
• Focus on blended ROAS as your primary metric
• Do not worry about multi-touch attribution complexity
• One or two channels means minimal double-counting

Growth Stage ($1,000-$10,000/month)

• Track blended metrics weekly
• Compare platform-reported ROAS to blended ROAS to understand inflation
• Run simple incrementality tests quarterly
• Use platform metrics for within-platform optimization

Scale Stage ($10,000+/month)

• Consider third-party attribution tools (Triple Whale, Northbeam, Rockerbox)
• Regular incrementality testing across channels
• Marketing mix modeling for budget allocation
• Blended metrics remain your source of truth

Common Attribution Mistakes

Trusting Platform Numbers at Face Value

Meta says your ROAS is 5x. Google says 4x. You feel great. But your actual blended ROAS is 2.5x. The gap is double-counting and view-through attribution inflation.

Killing Top-of-Funnel Based on Last-Click

Facebook prospecting campaigns often look terrible on last-click attribution because they are the first touch, not the last. Pausing them reduces the pipeline of new visitors, and retargeting performance eventually collapses because there is nobody new to retarget.

Ignoring View-Through Conversions

View-through conversions (someone saw your ad but did not click, then purchased later) are real but inflated. Ignoring them entirely undervalues video and brand advertising. Overcounting them overvalues those same channels. Use a 1-day view window as a reasonable compromise.

Over-Engineering Attribution

Spending more time analyzing attribution data than optimizing your ads and landing pages. Attribution should inform budget allocation decisions at the channel level. Within a channel, focus on creative quality, targeting, and conversion rate optimization.

Key Takeaways

• Attribution determines how you allocate budget so getting it approximately right matters for profitability
• No attribution model is perfect since each has tradeoffs between simplicity and accuracy
• Blended ROAS is your north star because it cannot be inflated by platform double-counting
• Use platform metrics for relative comparison within a platform, not for absolute truth
• Run incrementality tests quarterly by pausing channels briefly to measure their true contribution
• Do not kill top-of-funnel channels based on last-click data because they feed the entire funnel
• Apply the 80/20 rule since blended metrics plus basic testing gets you 80% to optimal allocation without expensive tools