Attribution Metrics & KPIs: What to Measure
The core attribution metrics worth tracking are attributed revenue and conversions by channel, cost per acquisition by channel, assisted conversions, conversion path length, and time-to-conversion — each answers a different question, and none of them alone tells the full story.
- Attributed Revenue and Conversions by Channel
- Cost Per Acquisition (CPA) and Return on Ad Spend (ROAS) by Channel
- Assisted Conversions and Conversion Assist Ratio
- Conversion Path Length and Time-to-Conversion
- Marketing-Sourced vs. Marketing-Influenced Pipeline
- Customer Lifetime Value (LTV) by Acquisition Channel
- Model Comparison Delta
- New Customer vs. Returning Customer Attribution
- Data Quality Metrics: The Ones Nobody Tracks but Should
The mistake we see most often isn’t tracking the wrong metrics — it’s tracking the right metrics but reading them as if they’re absolute facts rather than outputs of whatever attribution model happens to be running underneath them. Every metric in this list changes depending on the model. Know the model before you trust the number.
Attributed Revenue and Conversions by Channel
This is the headline metric in almost every attribution report: how much revenue (or how many conversions) does each channel get credit for. It’s useful for budget conversations, but it’s entirely model-dependent — the same underlying data produces different attributed revenue numbers under last-click versus data-driven attribution.
Always report the model alongside the number. “Organic search drove $40,000 in attributed revenue (data-driven model, 30-day window)” is a defensible statement. “Organic search drove $40,000” without qualification invites a challenge you can’t answer.
Cost Per Acquisition (CPA) and Return on Ad Spend (ROAS) by Channel
CPA and ROAS divide spend by attributed conversions or revenue, which means they inherit every distortion already baked into your attribution model. A channel that looks like it has a terrible CPA under last-click attribution might look completely different under a model that credits its assist role earlier in the path.
- Compare CPA under at least two models (e.g., last-click and data-driven) before cutting a channel’s budget based on this metric alone.
- Watch for channels with high CPA but strong assisted-conversion numbers — they may be doing real work that a single-model view hides.
Assisted Conversions and Conversion Assist Ratio
Assisted conversions count how many times a channel appeared in a conversion path without being the final touch. The assist ratio (assisted conversions divided by last-click conversions) tells you whether a channel plays more of a supporting or closing role in your funnel.
A high assist ratio with low last-click credit is the classic signature of top-of-funnel channels — organic content, social, display — being undervalued by simple last-click reporting. This metric is often the clearest evidence a marketing team can show leadership to justify continued investment in awareness-stage channels.
Conversion Path Length and Time-to-Conversion
Path length (average number of touchpoints before conversion) and time-to-conversion (average days from first touch to purchase) describe the shape of your funnel rather than crediting any specific channel. These are diagnostic metrics — they tell you whether your attribution model choice fits your actual sales cycle.
If path length is trending up over time, it can signal rising consideration friction (more research needed before purchase) or simply better tracking catching touchpoints it used to miss — worth distinguishing before drawing conclusions.
Marketing-Sourced vs. Marketing-Influenced Pipeline
Common in B2B reporting: “sourced” pipeline is revenue from deals where marketing generated the first touch or the lead itself; “influenced” pipeline is any deal where marketing touched the journey at any point, even if sales sourced the original contact. Both are legitimate metrics, but conflating them inflates marketing’s perceived impact and erodes trust with sales and finance.
- Report sourced and influenced pipeline as separate line items, never blended into one number.
- Define both terms in writing and get sales leadership to agree on the definitions before the first report goes out — disputes over definitions are far more common than disputes over the actual numbers.
Customer Lifetime Value (LTV) by Acquisition Channel
Acquisition-focused metrics (CPA, ROAS) only tell half the story if channels differ in the quality of customer they bring in. LTV by acquisition channel closes that gap — a channel with a higher CPA but a customer base that spends 2x more over a year may be the better investment even though first-touch ROAS looks worse.
This requires connecting attribution data to downstream customer data (repeat purchase, subscription retention, churn) — a heavier lift than most other metrics on this list, but often the single most strategically important one for subscription and repeat-purchase businesses.
Model Comparison Delta
This isn’t a standard off-the-shelf metric, but it’s one we build into every attribution dashboard: the percentage difference in attributed revenue for each channel between your primary model and a secondary model (commonly last-click vs. data-driven, or last-click vs. first-touch).
A large delta on a channel is a flag — it means the model choice materially changes the story for that specific channel, and any budget decision about it deserves extra scrutiny rather than a single-model rubber stamp.
In practice, we track this delta as a simple percentage column next to the primary attributed revenue figure in client dashboards. A channel sitting at a 5% delta between models is stable and safe to act on; one sitting at a 40%+ delta needs a conversation before any spend change, because the “right” answer depends heavily on which model’s assumptions actually fit that channel’s role in the funnel.
New Customer vs. Returning Customer Attribution
Blending new and returning customer conversions into one channel metric hides a common distortion: retargeting, branded search, and email tend to skew heavily toward returning customers, while organic content, top-of-funnel social, and cold prospecting ads skew toward new customer acquisition. A channel with a great blended ROAS can be almost entirely re-selling to people who already know the brand, which is a very different business result than genuine growth.
Split attribution reporting by new-versus-returning customer status wherever your platform allows it (most major ad platforms and GA4 support this segmentation). This single split often explains why a channel that looks efficient on paper isn’t actually contributing to the growth goals leadership cares about most.
Data Quality Metrics: The Ones Nobody Tracks but Should
Before trusting any of the metrics above, track the health of the data feeding them: percentage of sessions with “(not set)” or missing source/medium, percentage of conversions with no recorded UTM, and the gap between platform-reported (self-attributed) conversions and your independently tracked total.
- Untagged traffic above roughly 10-15% of sessions should be treated as a data-quality emergency, not background noise.
- If summed self-reported conversions across ad platforms exceed your actual total order count by a wide margin, that’s expected — it’s a reminder to always attribute from a single source of truth (GA4 or your CRM), not from adding up platform dashboards.
Frequently Asked Questions
What's the single most important attribution metric to start with?
Attributed revenue or conversions by channel, reported alongside the model used to generate it. It's the metric every other decision downstream depends on, and stating the model prevents disputes later.
Why does the same channel show different CPA in different reports?
CPA is derived from attributed conversions, and attributed conversions change depending on the attribution model applied. Different reports using different models (or different lookback windows) will legitimately produce different CPA for the same channel and spend.
What does a high assisted-conversion ratio actually tell you?
It signals a channel is playing a supporting, awareness-building role in the funnel rather than typically closing the sale — useful evidence for justifying continued investment in channels that look weak under last-click reporting alone.
Should I track marketing-sourced and marketing-influenced pipeline as one number?
No. Keep them separate. Blending them inflates marketing's apparent contribution and tends to create friction with sales and finance teams once the definitions are questioned.
Is customer lifetime value really an attribution metric?
It becomes one once you segment it by acquisition channel — doing so reveals whether cheaper-to-acquire channels are actually bringing in lower-value customers, which pure acquisition metrics like CPA can't show on their own.
How do I know if my attribution data quality is good enough to trust?
Check the percentage of sessions and conversions with missing source/medium data. If it's above roughly 10-15%, prioritize fixing tracking and tagging before drawing conclusions from any downstream metric.
Terry has 30+ years in software and SEO. He’s the founder of Salterra Digital Services and SEO Spring Training, host of the Roundtable SEO Mastermind, and lead instructor at SEO University — teaching the exact tactics his team uses on client work.
Continue learning
Ready to master this?
This guide is one lesson from the Analytics, Measurement & Attribution course. Get every lesson, framework and checklist — plus the full 38-course catalog — inside SEO University.
Practitioner-focused training across the full digital marketing stack — from technical SEO to conversion optimization and the AI search era. By Salterra Digital Services, since 2011.