Attribution
Back to Blog
Attribution — October 202610 min read
Attribution Models Explained for D2C Brands: Which One to Trust and When
Your Meta campaign says it drove ₹18L in revenue last month. Your Google campaign says it drove ₹14L. Your Shopify says total net revenue was ₹19L.
- Meta-attributed revenue
- ₹18L
- Google-attributed revenue
- ₹14L
- Combined platform claims
- ₹32L
- Shopify net revenue (reality)
- ₹19L
Attribution inflation₹13L
That ₹13L gap is attribution inflation, and it exists entirely because each platform is using its own attribution model to decide how much credit it deserves for each sale.
Understanding how attribution models work doesn’t just explain the gap. It tells you which channels are genuinely driving growth versus which are claiming credit for growth that was happening anyway.
What Is Attribution Modeling?
Attribution modeling is the process of assigning credit for a conversion (sale, lead, or other target action) to the marketing touchpoints that preceded it.
A customer’s journey to purchase typically involves multiple touchpoints. They might:
- 1
See a Meta Reel from your brand
- 2
Click a Google Shopping ad 3 days later
- 3
Receive an email with a discount code
- 4
Visit your site directly and purchase
Which of those touchpoints gets credit for the sale? That’s what attribution models decide, and different models give completely different answers to the same customer journey.
Six ways to split the credit
The 6 Core Attribution Models
Model 1
Last-Click Attribution
How it works: 100% of conversion credit goes to the last touchpoint before purchase.
In the example above: Direct visit gets all the credit. Meta and Google get zero.
- Where it’s used
- Shopify’s default UTM attribution. Many older analytics tools.
- Strengths
- Simple to understand. Clearly identifies the final conversion trigger.
- Weaknesses
- Systematically undervalues awareness and consideration channels (Meta, influencers, brand content) that introduce customers to the brand but don’t get the final click. Overvalues branded search and direct, channels that capture intent created elsewhere.
D2C implication: Last-click makes Google Brand Search look extremely efficient (high ROAS, high conversion rate), because it captures customers already primed to buy by Meta or organic. It makes Meta look weaker than it is.
Model 2
First-Click Attribution
How it works: 100% of credit to the first touchpoint in the journey.
In the example above: Meta Reel gets all the credit.
- Strengths
- Values brand awareness and discovery. Shows which channels introduce customers to the brand.
- Weaknesses
- Ignores every influence in the consideration and conversion stage. A customer who saw a Meta ad, waited 3 months, then was finally converted by an email gets that email attributed zero credit.
D2C implication: Useful for understanding which channels are most effective at driving first discovery, but not useful for optimising conversion efficiency.
Model 3
Linear Attribution
How it works: Equal credit distributed across all touchpoints. If there were 3 touchpoints, each gets 33.3%.
In the example above: Meta, Google, and email each get one-third.
- Strengths
- Acknowledges that every touchpoint contributed something.
- Weaknesses
- Treats all touchpoints as equally valuable regardless of when they occurred or what the customer did after seeing them. A 2-second video scroll gets the same credit as a deliberate search click.
Model 4
Time Decay Attribution
How it works: Touchpoints closer to conversion receive exponentially more credit than earlier touchpoints. The final touchpoint before purchase gets the most credit; the first touchpoint gets the least.
In the example above: Direct visit gets most credit, email gets the second most, Meta and Google split the remainder.
- Strengths
- Reflects the idea that recent influence is more directly responsible for conversion. Makes intuitive sense for short sales cycles.
- Weaknesses
- Systematically undervalues upper-funnel brand awareness campaigns that may have been the most important influence but happened weeks before purchase.
Model 5
Position-Based (U-Shaped) Attribution
How it works: 40% credit to first touch, 40% to last touch, remaining 20% distributed evenly across middle touchpoints.
In the example above: Meta gets 40%, direct visit gets 40%, Google and email split the remaining 20%.
- Strengths
- Values both discovery (first touch) and conversion (last touch), the two most strategically important moments in the customer journey.
- Weaknesses
- Still somewhat arbitrary. The 40/40/20 split isn’t derived from actual conversion influence data.
D2C implication: A reasonable balanced model for brands that want to give appropriate credit to both acquisition (Meta prospecting) and conversion (branded search, retargeting) channels.
Model 6
Data-Driven Attribution
How it works: Uses machine learning to assign credit based on actual impact, comparing the conversion rates of customer paths that included a specific touchpoint versus those that didn’t.
In the example above: Credit is assigned proportionally based on what the model calculates each touchpoint actually contributed to the probability of conversion.
- Where it’s used
- Google Ads (available when sufficient conversion data exists), Meta (in development).
- Strengths
- Most accurate model when sufficient data exists. Accounts for actual influence rather than assumed rules.
- Weaknesses
- Requires large conversion volumes (Google recommends 3,000+ conversions/month for reliable data-driven attribution). A black box, you can’t fully audit how credit is assigned.
Why Attribution Models Produce the Shopify-Meta Revenue Gap
The most practical implication of attribution modeling for D2C brands: why your combined platform-attributed revenue doesn’t match Shopify.
The problem: Each platform applies its own attribution model to the same customer journey. A customer who touched both Meta and Google before purchasing gets counted by both.
Meta default
7-day click, 1-day view
If the customer clicked a Meta ad anytime in the 7 days before purchase, or even just saw it in the 24 hours before, Meta attributes the conversion.
Google default
30-day click
If the customer clicked any Google ad in the 30 days before purchase, Google attributes the conversion.
For the same purchase, both platforms can simultaneously attribute full conversion value, creating the 30–80% combined attribution inflation that separates platform-claimed revenue from real Shopify net revenue.
The only solution
Use a platform-independent metric as your aggregate efficiency benchmark. MER (Total Shopify Net Revenue ÷ Total Marketing Spend) ignores all attribution models and gives you the real relationship between marketing spend and real revenue.
Attribution Modeling vs Incrementality Testing
Attribution modeling answers
Which touchpoints were in the path to conversion?
Correlation along the journey.
Incrementality testing answers
Which touchpoints actually caused the conversion?
Causation, measured by holding a channel out.
These are different questions.
A customer who sees a Meta ad, then purchases 6 days later through branded Google Search, Meta gets attribution credit. But would they have bought anyway through Google Search even without seeing the Meta ad? If yes, Meta’s attributed credit overstates its actual incremental value.
Incrementality testing (pausing a channel for 2 weeks and measuring the revenue impact) gives you the answer attribution models cannot: does this channel generate genuinely new demand, or does it mostly capture demand created by other channels or organic intent?
For most Indian D2C brands:
| Channel | True incrementality | Why |
|---|---|---|
| Google Brand Search | 30–60% | Captures a lot of organic intent |
| Meta prospecting | 60–80% | Higher, drives genuinely new awareness |
| Email retargeting | 40–65% | Mixed, some organic repurchases, some genuine influence |
Run incrementality tests quarterly on your highest-spend channels. The results will likely change your budget allocation significantly.
What Attribution Model Should D2C Brands Use?
The honest answer: none of them perfectly.
Each model is a simplification of a complex multi-touch reality. None of them can perfectly isolate how much each touchpoint actually caused the conversion.
The practical framework:
Campaign-level optimisation
Platform-native attribution
Use platform-native attribution (Meta 7-day click, Google last-click) consistently. Don’t change attribution settings frequently, the goal is a stable, consistent signal for the algorithm, not perfect accuracy.
Cross-channel budget allocation
MER
Ignore individual channel ROAS numbers for allocation decisions. MER is the only metric that doesn’t depend on any attribution model.
Strategic channel assessment
Incrementality tests
Quarterly holdout tests on your highest-spend channels tell you more about real channel value than any attribution model.
Campaign profitability
POAS from Shopify revenue
Not platform-attributed ROAS. Connect your actual net revenue, COGS, and returns to get the real profitability picture.
Attribution Windows: Why They Matter as Much as the Model
Attribution model (last-click, linear, etc.) gets most of the attention. Attribution window, how far back the model looks for touchpoints, has equally large impact on reported ROAS.
| Source | Default window |
|---|---|
| Meta | 7-day click, 1-day view |
| 30-day click | |
| Shopify UTM | Last click, no defined window |
For D2C products with any consideration period, anything above ₹500, anything where the customer compares alternatives, the purchase decision may take 1–3 weeks. A campaign that introduced the customer 10 days before purchase gets zero credit under a 7-day click window, even if it was the primary conversion driver.
What to do
In Meta Ads Manager, compare 7-day click vs 28-day click attribution for each campaign. If significant volume falls in the 8–28 day window, your campaigns are reaching customers earlier in the consideration cycle than the default window shows. This doesn’t mean Meta is more effective than reported, but it means the 7-day ROAS understates the actual contribution.
Use this data to understand your product’s consideration cycle, not to inflate your reported ROAS for optimisation decisions.
The Attribution-Proof Metric Stack
For D2C brands, the pragmatic answer to attribution complexity is building a metric stack that triangulates rather than relying on any single attribution model:
- 1
Level 1: MER (macro, weekly)
Total Shopify net revenue ÷ total marketing spend. Attribution-proof aggregate. Tells you if overall marketing efficiency is improving or declining.
- 2
Level 2: POAS per campaign (campaign, weekly)
Real profitability per campaign using Shopify revenue, COGS, and returns. Not platform-attributed ROAS. Not any attribution model.
- 3
Level 3: Incrementality test results (strategic, quarterly)
True causal impact of each channel. Tells you which channels are genuinely driving growth.
- 4
Level 4: Platform ROAS (diagnostic, weekly)
Used for in-platform decisions, creative testing, bid strategy, audience comparison, not for business-level decisions.
This stack is less elegant than “one attribution model to rule them all.” It’s more accurate. And accuracy is what budget allocation decisions actually need.
Conclusion
Attribution models explain how platforms divide credit for conversions. None of them perfectly capture real causal influence. All of them produce some version of attribution inflation when multiple channels run simultaneously.
Understanding the models matters not because you need to choose one and trust it completely, but because understanding what each one does helps you know:
- why your platform numbers don’t match Shopify,
- why cross-channel ROAS comparison is unreliable,
- and why a platform-independent metric like MER is essential.
Attribution tells you the story each platform wants to tell. Incrementality, MER, and POAS tell you what actually happened.
Frequently Asked Questions
What is attribution modeling in marketing?
Attribution modeling assigns credit for a conversion to the marketing touchpoints that preceded it. Different models (last-click, first-click, linear, time-decay, position-based, data-driven) distribute credit differently, producing different views of channel effectiveness from the same customer journey data.
Why does my combined Meta and Google revenue exceed my Shopify revenue?
Because both platforms use their own attribution models and claim credit for the same purchases. A customer who touched both Meta and Google before purchasing appears in both platforms' conversion data. This attribution overlap is structural, the only solution is measuring aggregate efficiency through MER (Shopify net revenue ÷ total marketing spend) rather than comparing individual platform ROAS.
What is data-driven attribution and when should D2C brands use it?
Data-driven attribution uses machine learning to assign credit based on actual impact, comparing conversion rates of customer paths that included a touchpoint versus those that didn't. It's the most accurate model but requires 3,000+ monthly conversions for reliable results. Available in Google Ads automatically at sufficient scale.
How does Flable AI address attribution uncertainty?
Flable uses Shopify net revenue as the ground truth for all profitability calculations, bypassing platform attribution entirely. POAS and CM2 per campaign are calculated from actual Shopify orders, COGS, and returns data, not from what platforms claim to have driven.
Measure What Actually Happened
Flable AI calculates POAS and CM2 per campaign from real Shopify orders, COGS and returns, bypassing platform attribution entirely.
Book a Live Demo → Backed by Microsoft & NVIDIA
