AI Attribution for D2C Brands Selling on Amazon, Nykaa, and Flipkart

Your GA4 dashboard shows ₹30,000 in AI-referred revenue last month. A user on ChatGPT asked for the best vitamin C serum, your brand was recommended, they clicked through, and bought. Clean attribution. You can see it.

But three other users on the same day read the same ChatGPT recommendation, closed the chat, opened Amazon, searched your brand name, and bought on Amazon. GA4 shows none of that.

For pure DTC brands selling only through their own website, the how to track AI referral traffic in GA4 guide covers everything you need. But for D2C brands with marketplace presence — which in India means almost every brand — GA4 AI attribution is structurally incomplete by design. It cannot see what happens after a user leaves your site for Amazon, Nykaa, or Flipkart.

This guide builds the full attribution model: how to measure AI's true revenue impact across your DTC site and your marketplace channels, why this problem is especially acute for Indian D2C, and how to present this data to investors and leadership in a way that's defensible.

https://cdn.sanity.io/images/en9d2pb2/production/07f5f69d66643c0ba0a17602e0d4771001fc9cd5-1200x630.png

---

The Attribution Problem No One Is Talking About

Traditional attribution models — last click, first click, data-driven — were built for a world where customers journey through trackable touchpoints on your own properties. A user sees a Facebook ad, clicks to your website, browses, abandons, gets retargeted, returns via email, and buys. Every step has a cookie. You can reconstruct the whole path.

AI attribution breaks this in two distinct ways.

Break #1: The referrer stripping problem. When ChatGPT sends a user to your website, the referrer header is often stripped — particularly on mobile apps and when users open links in new tabs. GA4 records these sessions as direct traffic, not AI referral. This is well documented and widely discussed. Typically 20–30% of AI-originated sessions land in GA4 as direct.

Break #2: The marketplace exit problem. This one is less discussed but more consequential for most D2C brands. A user reads a ChatGPT recommendation, doesn't click the link at all, and instead opens Amazon or Nykaa directly. They search your brand name. They buy. This journey is completely invisible to GA4 — there's no session on your site, no referral to track, no conversion to attribute.

For a brand selling exclusively on their own website, Break #1 is the main concern. For a brand selling on Amazon, Nykaa, and Flipkart alongside their DTC channel — which describes the vast majority of Indian D2C brands — Break #2 is where the larger revenue impact lives.

The result is what we call the AI attribution gap: the difference between what GA4 shows as AI-influenced revenue and the true brand revenue driven by AI recommendations. For Indian D2C brands in beauty and wellness, this gap can be 5–8× the GA4 figure.

---

How AI Recommendations Actually Drive Marketplace Purchases

To understand why the gap is so large, trace the actual user journey.

Journey A (GA4 sees this): User asks ChatGPT "best collagen supplement for women in India" → ChatGPT recommends Brand X with a link → User clicks link → Lands on Brand X website → Buys → GA4 records as chatgpt.com / referral, conversion attributed.

Journey B (GA4 misses this): User asks ChatGPT same question → ChatGPT recommends Brand X → User closes chat, opens Amazon app → Searches "Brand X collagen" → Buys on Amazon → GA4 records nothing. Amazon records a sale, but attributes it to Amazon organic search.

Journey C (GA4 partially sees this): User asks Perplexity same question → Perplexity recommends Brand X → User reads the response, doesn't click → Later that evening, types "Brand X" into Google → Clicks organic search result → Lands on Brand X website → Browses but doesn't buy → Returns two days later via direct URL → Buys → GA4 credits direct traffic, last click. The Perplexity recommendation gets zero credit.

Journeys B and C are not edge cases. Based on behavioral data from Indian D2C brands that have implemented the full 5-layer attribution model, Journeys B and C together account for the majority of total AI-influenced revenue — often 60–75% of the true total. This is the revenue that standard GA4 reporting permanently misses.

The pattern is especially pronounced in India because Indian consumers are deeply marketplace-habituated. They discover brands across multiple surfaces — Instagram, YouTube, WhatsApp, AI chatbots — but their default purchase environment is Amazon, Nykaa, or Flipkart, not the brand's own website. An AI recommendation is a discovery event. The purchase happens wherever the consumer is most comfortable transacting.

---

The Indian D2C Context: Why This Matters More Here

India's D2C landscape is structurally different from the US, Europe, or Southeast Asia in one critical way: pure-play DTC is rare. Most brands that identify as D2C operate across multiple channels simultaneously.

According to RedSeer's 2024 analysis of India's D2C market:

• ~55–60% of D2C brand revenue flows through Amazon India, Nykaa, and Flipkart combined
• Only 20–30% of brand revenue comes through the brand's own DTC website
• The remaining 10–20% flows through offline retail, modern trade, and quick commerce (Blinkit, Swiggy Instamart, Zepto)

This channel mix is not a choice brands are making reluctantly — it reflects where Indian consumers actually shop online. Marketplace trust, cash-on-delivery availability, easy returns, and established payment flows make Amazon, Nykaa, and Flipkart the path of least resistance for most online purchases.

Nykaa's position in beauty and wellness is particularly dominant. Nykaa accounts for more than 40% of online beauty and personal care category sales in India. For D2C beauty brands, Nykaa is often the primary revenue channel — not a secondary one. When ChatGPT recommends a beauty brand to a user in India, the most likely purchase location is Nykaa, not the brand's own website.

Flipkart's role in mass-market categories mirrors this dynamic for value-segment apparel, consumer electronics accessories, and personal care. Flipkart's user base skews toward Tier 2 and Tier 3 cities, where D2C brand websites often have lower conversion rates due to payment friction and trust concerns.

Amazon India's pan-category presence means that almost any D2C brand — beauty, wellness, nutrition, apparel, home goods — has meaningful Amazon revenue. Amazon Brand Analytics is therefore one of the most valuable cross-reference tools for AI attribution validation.

The implication for AI attribution is direct: if 60% of your brand's revenue is happening on platforms that GA4 cannot see, then GA4's AI attribution number represents approximately 40% of the true picture at best — and that's before accounting for dark social and branded search halo effects.

---

The US DTC Context: Comparison

For US DTC brands, the picture is similar in structure but different in degree.

Amazon US accounts for approximately 35–45% of product purchase intent conversion for CPG and personal care categories, according to eMarketer 2024 data. This is meaningfully lower than India's 55–60% marketplace dependency, but still substantial.

US DTC brands face the same AI attribution gap, but the gap is somewhat narrower:

• US DTC websites have higher standalone conversion rates — consumer trust in brand websites is stronger
• Subscription and DTC loyalty programs are more mature in the US, meaning more repeat revenue flows through the brand's own properties
• Amazon's share of specific high-consideration categories (skincare, supplements, apparel) is lower than in India because US consumers are more willing to transact directly with brands

A US DTC brand in beauty and wellness might see a marketplace spillover multiplier of 1.5–2×, versus 2–2.5× for an equivalent Indian brand. The 5-layer combined multiplier for US brands typically lands at 4–5×, versus 6–8× for Indian D2C.

The methodology is identical — the inputs differ based on your actual channel mix.

---

Step-by-Step: Building the Marketplace Spillover Calculation

This is the most actionable section of this guide. Here's exactly how to calculate the marketplace spillover for your brand.

Inputs You Need

Before you calculate anything, gather these numbers for the same time period (use the last 90 days for statistical reliability):

• GA4 AI referral revenue — sessions from chatgpt.com, perplexity.ai, gemini.google.com, claude.ai, and similar, with their associated conversion revenue
• DTC website total revenue — all revenue from your brand's own website in the same period
• Total brand revenue — DTC website + Amazon + Nykaa + Flipkart + any other channels combined for the same period

The Calculation

Step 1: Calculate your DTC revenue share

DTC Share = DTC Revenue ÷ Total Brand Revenue

Example: DTC revenue ₹4,00,000 / Total brand revenue ₹10,00,000 = 0.40 (40% DTC share)

Step 2: Apply to GA4 AI revenue to get total AI-influenced brand revenue

Total AI-Influenced Revenue = GA4 AI Revenue ÷ DTC Share

Example: ₹50,000 GA4 AI revenue ÷ 0.40 DTC share = ₹1,25,000 total AI-influenced brand revenue

Step 3: Validate the implied marketplace figure

Implied AI Marketplace Revenue = Total AI-Influenced Revenue - GA4 AI Revenue

Example: ₹1,25,000 - ₹50,000 = ₹75,000 implied marketplace revenue from AI influence

This ₹75,000 is the revenue that happened on Amazon, Nykaa, and Flipkart because of AI recommendations — invisible to GA4 but real and attributable through this model.

The Full Worked Example Calculation Table

Step	Metric	Value	Formula
Input	DTC Revenue (90 days)	₹4,00,000	From GA4 / Shopify
Input	Total Brand Revenue (90 days)	₹10,00,000	DTC + Amazon + Nykaa + Flipkart
Input	GA4 AI Referral Revenue	₹50,000	GA4 Traffic Acquisition report
Step 1	DTC Revenue Share	40%	4,00,000 ÷ 10,00,000
Step 2	Marketplace Spillover Multiplier	2.5×	1 ÷ 0.40
Step 3	Total AI-Influenced Brand Revenue	₹1,25,000	50,000 ÷ 0.40
Step 4	Implied Marketplace AI Revenue	₹75,000	1,25,000 - 50,000
Step 5	% of AI Revenue That Was Invisible	60%	75,000 ÷ 1,25,000

This is a conservative calculation. It assumes your AI traffic converts on marketplaces at exactly the same rate as your overall channel mix — which is likely an underestimate, since AI-referred users typically show higher purchase intent.

---

How to Cross-Reference AI Attribution with Marketplace Data

The spillover calculation above is a model estimate. To validate it and build confidence in the numbers, you need to cross-reference against actual marketplace data. Here's what each platform provides.

Amazon Brand Analytics: Branded Search Volume Correlation

Amazon Brand Analytics (available to brand-registered sellers) provides weekly branded search volume data — how many users searched your brand name on Amazon.

The cross-reference method:

• Export your weekly AI mention volume from your brand visibility tracker or similar tool
• Export Amazon branded search volume by week from Brand Analytics
• Plot both on the same timeline
• Look for correlation with a 3–7 day lag (users see AI recommendation, buy on Amazon within a week)

A strong positive correlation — AI mention spikes followed by branded search spikes on Amazon — is the clearest validation signal available that AI is driving real marketplace purchase intent. This transforms your spillover calculation from a model assumption into an empirically supported estimate.

Nykaa Brand Dashboard: Referral Source Tracking

Nykaa's seller/brand dashboard provides traffic source data for your brand page. In particular:

• Direct traffic to your Nykaa brand page (users who typed your brand name directly)
• Search-driven traffic (users who found you via Nykaa's internal search)
• External referral traffic

Monitor weekly direct and branded search traffic on Nykaa. Spikes in this data that correlate with AI mention peaks are strong signals of AI-driven marketplace discovery. Nykaa's dashboard does not expose granular referrer data from external sources like ChatGPT, but the branded search volume proxy works well given the platform's scale in beauty and wellness.

Flipkart Seller Analytics: Direct Link vs. Search Traffic

Flipcart's seller analytics portal distinguishes between:

• Search traffic (users who found your listing via Flipkart search)
• Direct link traffic (users who navigated to your listing via a URL)
• Flipkart ad traffic

AI attribution signal on Flipkart is typically visible in the branded search category — users who search your exact brand name rather than a generic category term. Weeks with high AI mention volume should show corresponding branded search lift on Flipkart with a similar 3–7 day lag.

Marketplace Platform Data Comparison

Platform	Data Available	Attribution Signal	Lag to AI Mention	Reliability
Amazon Brand Analytics	Weekly branded search volume, Search Query Performance	Branded search spike after AI mention	3–7 days	High — granular and reliable
Nykaa Brand Dashboard	Traffic sources, branded search, direct visits	Direct + branded search spike	3–7 days	Medium — less granular referrer data
Flipkart Seller Analytics	Search vs. direct traffic split, keyword-level data	Branded keyword search spike	3–7 days	Medium — improving with Flipkart Commerce Cloud
Your DTC site (GA4)	Full session + conversion data, referrer details	AI referral session spike	Same day	High — but only covers DTC channel

---

The 5-Layer Model Applied to Indian D2C: Full Worked Example

The marketplace spillover calculation above addresses one layer of the attribution gap. But there are four additional layers of AI revenue that GA4 misses. Together, they produce the full multiplier.

Here is the complete model applied to a real-world scenario: an Indian beauty and wellness brand (comparable to a MyMuse-category brand) with ₹30,000/month in GA4 AI referral revenue.

For a deep dive on how MyMuse became ChatGPT's #1 brand in its category, the methodology there validates the layer multipliers used in this model.

The 5-Layer Model: Full Calculation

Layer	What It Captures	Multiplier	Revenue After Layer	Notes
L1: GA4 Direct	Tracked AI referral sessions that converted on DTC site	1.0×	₹30,000	Read directly from GA4
L2: Dark Social Adjustment	AI-originated sessions that arrived in GA4 as direct (referrer stripped)	×1.25	₹37,500	20–30% of AI sessions mis-attributed as direct
L3: Branded Search Halo	Users who searched brand name on Google/Nykaa/Amazon after AI exposure	×1.40	₹52,500	Branded search lift validated via GSC and Amazon BA
L4: Marketplace Spillover	AI-influenced purchases on Amazon, Nykaa, Flipkart	×2.50	₹1,31,250	60% DTC share → 2.5× multiplier (40% DTC / 60% marketplace)
L5: LTV Uplift	Higher 90-day repeat purchase rate from AI-referred cohorts vs. paid	×1.615	₹2,11,718	AI-referred customers show 15–25% higher repeat rate
Combined	Full true AI-influenced brand revenue	~7.06×	₹2,11,718	Validated on MyMuse-like 90-day dataset

The 7.06× combined multiplier means that for every ₹1 that GA4 shows as AI-referred revenue, this brand's true AI-influenced revenue across all channels is approximately ₹7.06.

Conservative vs. best-case sensitivity:

• Conservative estimate (lower multipliers at each layer): ~4.5× → ₹1,35,000
• Central estimate (model above): ~7.06× → ₹2,11,718
• Best case (higher marketplace share, strong LTV): ~9× → ₹2,70,000

https://cdn.sanity.io/images/en9d2pb2/production/a4fdb43915514079a0b09009c803120c2a68b7bc-902x878.png

For the full methodology behind the GA4 undercount problem and how each layer is validated, see the AI revenue attribution GA4 undercount guide.

---

Category-Specific Multipliers: Beauty vs. Electronics vs. Apparel vs. Food

The 7.06× multiplier above is calibrated for beauty and wellness — the highest-multiplier category for Indian D2C. Other categories have different channel mix, purchase cycle dynamics, and LTV profiles that change the combined multiplier significantly.

Category Multiplier Comparison Table

Category	Typical DTC Share	Marketplace Share	Key Marketplace	L4 Multiplier	Combined Multiplier	Notes
Beauty & Wellness	25–35%	65–75%	Nykaa dominant	2.5–4×	6–9×	High-intent AI queries; Nykaa's category dominance drives high spillover
Nutrition & Supplements	30–40%	60–70%	Amazon dominant	2.0–3×	5–8×	Amazon is trusted channel; auto-subscribe behavior raises LTV
Apparel & Fashion	40–55%	45–60%	Flipkart + Myntra	1.5–2×	4–6×	Higher DTC share for premium; Myntra not covered by basic model
Electronics & Accessories	20–30%	70–80%	Amazon + Flipkart	3–5×	5–8×	Price-sensitive category; Amazon heavily dominant; lower AI CVR
Food & Gourmet	50–65%	35–50%	Amazon + QC	1.3–1.8×	3–5×	Higher DTC share; quick commerce adds complexity; shorter LTV
Pet Care	35–50%	50–65%	Amazon dominant	1.5–2.5×	4–7×	Growing category; high repeat purchase; AI recommendations gaining traction

How to apply this table: Find your category, read the typical DTC share, verify it matches your actual channel mix, then use the combined multiplier range as your starting estimate. Refine using your actual DTC share in the spillover formula from Section 5.

---

How to Present This to Investors and Leadership

The challenge with AI attribution models is credibility. Any multiplier that produces a number 7× larger than what GA4 shows requires careful framing — otherwise it looks like you're manufacturing favorable data.

Here is how to present this analysis in a way that's defensible and compelling.

Frame the Conservative Case First

Always lead with the number your audience can immediately verify: GA4 AI referral revenue. This is the floor — it's undeniably real. Then build up from there with each layer explicitly justified.

Sample framing for an investor update:

"GA4 shows ₹30,000 in direct AI referral revenue last month. This is what we can directly track and verify. However, GA4 structurally cannot see purchases that happen on Amazon and Nykaa after AI discovery — and given our channel mix (40% DTC / 60% marketplace), we estimate the actual AI-influenced brand revenue is between ₹1,35,000 (conservative) and ₹2,10,000 (central estimate). The conservative estimate uses a simple DTC share adjustment only. The central estimate adds dark social correction and branded search halo, both of which are validated by Amazon Brand Analytics branded search spikes that correlate with our AI mention peaks."

This framing:

• Anchors on a verifiable number
• Explains the structural reason for undercounting (not a modeling trick)
• Offers a range rather than a single optimistic figure
• Points to external validation (Amazon BA branded search correlation)

Model Sensitivity Analysis Table

Presenting a sensitivity table makes the analysis look rigorous rather than cherry-picked.

Scenario	DTC Share Assumption	Dark Social Layer	Branded Search Layer	Marketplace Layer	LTV Layer	Total Revenue	Multiplier
Ultra-conservative	40%	None	None	2.5× only	None	₹75,000	2.5×
Conservative	40%	×1.20	None	2.5×	None	₹90,000	3.0×
Central (base case)	40%	×1.25	×1.40	2.5×	×1.615	₹2,11,718	7.06×
Optimistic	35%	×1.30	×1.50	2.86×	×1.80	₹2,97,000	9.9×
Best case	30%	×1.35	×1.60	3.33×	×2.0×	₹4,32,000	14.4×

For most purposes, present the Ultra-conservative and Central scenarios side by side. The ultra-conservative scenario only applies a single marketplace multiplier and no other layers — it's the absolute minimum credible estimate. The central scenario applies all five layers at validated midpoints.

What to Watch For: Red Flags That Signal Model Error

If your central estimate multiplier is above 12×, check your inputs. Common mistakes:

• Double-counting marketplace revenue (brand total already includes DTC site sales, so don't add them again)
• Using monthly revenue figures for one channel and quarterly for another
• Applying an LTV multiplier to the already-multiplied figure instead of the L1 base

If your multiplier is below 3×, check that you've correctly identified all marketplace channels. Brands sometimes forget Myntra, Meesho, Jiomart, or quick commerce platforms (Blinkit, Swiggy Instamart) in their total brand revenue figure.

---

Tracking Improvements: What Better Attribution Looks Like

The model above is the best approximation possible with standard tools. But there are concrete investments that improve attribution accuracy over time.

UTM Tagging for Creator and Publisher Programs

If your brand works with content creators, publishers, or affiliate partners who mention you on AI-indexed content, add UTM parameters to every tracked URL:

https://www.yourbrand.com/products/vitamin-c-serum
  ?utm_source=creator
  &utm_medium=content
  &utm_campaign=q1-2026
  &utm_content=chatgpt-cited-article

When a creator's article gets cited by ChatGPT or Perplexity and a user clicks through, you'll see the UTM parameters in GA4 — giving you the most direct attribution possible for AI-cited content.

Dedicated Landing Pages for AI-Cited Content

Create unique landing page URLs for content specifically written to appear in AI answers. When AI platforms cite these URLs, traffic to these pages becomes a reliable proxy for AI referral volume — even when referrer headers are stripped.

Example: /pages/best-vitamin-c-serum written specifically to rank in Perplexity answers gets 300 sessions in a week. Even if GA4 shows those sessions as direct, you know from the URL that they came via AI-cited content.

Marketplace Creator Program Links

Both Amazon (Amazon Associates) and Nykaa have creator/affiliate programs that generate trackable links. If you can incentivize creators to include your Amazon or Nykaa product links alongside their DTC links, you get partial attribution for the marketplace journey.

This doesn't solve the core problem (organic marketplace searches from AI discovery), but it improves measurement for the creator-driven portion of your AI attribution.

Publisher Code Partnerships

Work with publishers and media partners who appear in AI answers to include your product links with publisher-specific codes. When a piece of coverage on Healthshots or Femina gets cited by Perplexity, publisher-coded links in that article pass attribution data you can track.

What Full Attribution Looks Like

When all tracking improvements are in place, a brand running this full system sees:

Attribution Method	What It Captures	Coverage
GA4 AI referral (baseline)	Direct clicks from AI platforms to DTC site	20–30% of true total
GA4 direct (UTM-tagged URLs)	Referrer-stripped AI sessions via recognizable URLs	+5–10%
Creator/publisher UTM links	AI-cited content with tracked URLs	+5–15% (depends on program size)
Amazon Associates creator links	Marketplace clicks from creator content	+3–8%
Modeled marketplace spillover	Estimated organic marketplace AI revenue	Remainder
Amazon BA branded search correlation	Validation signal (not direct attribution)	Cross-reference only

Even with all improvements in place, some portion of AI-influenced marketplace revenue remains modeled rather than directly tracked. This is a structural limitation of selling through third-party marketplaces — they don't share customer journey data with brands.

The goal of tracking improvements is to shrink the modeled portion and grow the directly tracked portion over time. Each quarter of better tracking data also improves the calibration of your spillover model.

---

Practical Action Plan: Getting Started This Week

If you're reading this and want to get started without rebuilding your entire analytics stack, here's a prioritized action list.

This week:

Pull your GA4 AI referral revenue for the last 90 days (sessions from chatgpt.com, perplexity.ai, gemini.google.com)

Pull your DTC revenue and total brand revenue (DTC + all marketplaces) for the same 90 days

Calculate your DTC share and run the spillover calculation from Step 2 above

This gives you your first AI attribution estimate in under two hours

This month:

Export Amazon Brand Analytics branded search data for the same 90 days

Plot AI mention volume (from your brand visibility tracker) against branded search weekly data

Document any visible correlation — this is your primary validation evidence

Build the 5-layer model using your actual channel mix inputs

This quarter:

Implement UTM tagging for any creator or publisher partnerships

Set up a custom GA4 segment for AI referral traffic with a pinned report

Create a monthly reporting template that shows GA4 AI revenue, modeled total, and key validation signals

Present the sensitivity analysis table to leadership with conservative and central estimates clearly labeled

For automated AI traffic monitoring that does much of this work continuously — tracking AI mentions, mapping them to GA4 sessions, and flagging branded search correlations — the AI traffic decoder is designed specifically for this workflow.

---

The Bottom Line

GA4 is not broken. It tracks what it can see, and it sees your DTC website accurately. The problem is that for Indian D2C brands, your DTC website is only 30–45% of your business. The other 55–70% happens on Amazon, Nykaa, and Flipkart — where GA4 has no visibility.

When ChatGPT or Perplexity recommends your brand, the majority of the purchase intent that recommendation generates converts on marketplaces, not on your site. This isn't a hypothetical — it's the structural reality of how Indian consumers shop online.

Building a proper AI attribution model for your brand means accepting that some of your most valuable marketing activity will never be perfectly tracked. But "not perfectly tracked" is not the same as "unmeasurable." The spillover calculation, the 5-layer model, and the Amazon BA cross-reference together produce estimates that are rigorous enough to inform investment decisions, compelling enough to present to investors, and specific enough to guide your answer engine optimization strategy.

The brands that build this attribution discipline now will have 12–18 months of historical data and model calibration by the time AI becomes every analyst's top question in a board meeting. Start with the spreadsheet calculation. Refine it with marketplace data. And consider the AI traffic decoder for continuous monitoring without the manual overhead.

Book a demo to see how Asva AI's full attribution stack handles the Indian D2C marketplace problem automatically.

For more on how to approach AI visibility for e-commerce categories broadly, read our guide on AI visibility for e-commerce.