AMC vs data warehouse: what Amazon sellers actually need
Amazon Marketing Cloud is an advertising clean room. A data warehouse is your operational source of truth. Where each fits, what AMC does not cover, and how mid-market brands combine them in 2026.
Gemini
ChatGPTThe short version
AMC is an advertising clean room. A data warehouse is your operational source of truth. They solve different problems, and brands who treat AMC as a warehouse end up with unattributed inventory, missing fee lines, and reports that never reach operations. Below, what each one actually contains, where they overlap, and how mid-market Amazon brands should combine them in 2026.
Amazon Marketing Cloud gets pitched as "Amazon's data lake." It is not. It is a purpose-built advertising clean room with a specific job: give advertisers event-level access to their own campaign data, joined to Amazon shopping signals under privacy controls, so they can measure cross-tactic attribution and build audiences. That is a real job. It is a different job from running a P&L.
What each system actually is
Amazon Marketing Cloud (AMC)
A clean room hosted by Amazon Ads. SQL-driven, no visualization layer, 12-month rolling window. Contains Sponsored Ads and DSP event data, some Amazon shopping signals, and (via Advanced Instances) uploaded first-party data.
Best for: cross-DSP attribution, audience overlap, path-to-conversion, incrementality tests.
An Amazon data warehouse
A managed database (ClickHouse, BigQuery, Snowflake, or a vendor-hosted equivalent) that ingests SP-API, Advertising API, AWS Data Exchange, and settlement reports into pre-modeled tables refreshed on a schedule.
Best for: SKU P&L, fee tracking, inventory planning, PPC efficiency at product grain, weekly reporting, alerts.
What AMC covers (and what it does not)
AMC's data model is centered on ad events. Amazon publishes the schema in the AMC API documentation, and third-party ad agencies like Tinuiti's AMC breakdown maintain useful running notes on schema changes. The tables that matter for most advertisers:
Core AMC tables
- • sponsored_ads_traffic: SP, SB, SD impressions and clicks
- • dsp_impressions / dsp_clicks: DSP event stream
- • conversions: orders and units attributed to ad exposure
- • amazon_attributed_events: attributed sales windowed by tactic
- • audience_segments: anonymized audience membership
- • video_events: streaming and OLV playback signals
What is not in AMC, and where operators get tripped up:
- • No SKU-level fee lines (referral, FBA fulfillment, storage, aged inventory, LIL fee)
- • No settlement transactions, reversals, or reimbursement cases
- • No inventory snapshots, replenishment forecasts, or AWD movements
- • No S&S subscription counts or S&S sales share
- • No Search Query Performance in a joinable form
- • No COGS, landed cost, or supplier data (you would upload it, and it stays in your own instance)
- • No product-level PPC ROI joined to profit (only revenue-side attribution)
None of that is a criticism of AMC. It is a clean room for ads. The gap is real, though, and it is why brands who try to run their monthly close out of AMC end up rebuilding the reports every quarter.
The 4 operator jobs AMC cannot do
1. SKU-level contribution margin
Requires settlement fees joined to units, COGS, and fulfillment cost per unit. AMC has revenue attribution, not cost attribution.
2. Days of inventory and replenishment
Requires FBA inventory API, inbound shipments, and demand forecasts. Not present in AMC.
3. Fee anomaly detection
Requires the 40+ Amazon fee taxonomy modeled at SKU-day grain. AMC does not carry fee data.
4. Weekly operator reporting
AMC has no scheduler, no dashboarding, and a hard 12-month window. Every recurring report needs a warehouse and BI tool downstream anyway.
What Nova puts in the warehouse that AMC cannot
Nova pre-models 100+ Amazon endpoints across SP-API, Advertising API, AWS Data Exchange, and settlement reports, then ships them into a ClickHouse instance you can query directly, feed into your BI tool, or connect to Claude and other LLMs. The tables cover the operational surface AMC skips:
- • sku_pnl_daily: revenue, refunds, all fee lines, COGS, contribution margin per SKU per day
- • fee_taxonomy: 40+ Amazon fee types resolved from settlement rows
- • inventory_daily: FBA on-hand, inbound, unfulfillable, AWD; days of supply per SKU
- • ppc_efficiency_sku: product-level spend joined to profit, not just revenue
- • subscribe_save: S&S sales, S&S share of sales, active subscriptions
- • returns: reason codes, category thresholds, returns processing fee exposure
- • reimbursements: case status, recovery amount, root cause
When to run AMC alongside Nova
The two systems are complementary once a brand crosses a certain scale:
- • DSP spend over $250K per year: AMC pays back on cross-tactic attribution and audience discovery
- • Multi-brand portfolio with shared audience overlap questions: AMC audience_segments earns its seat
- • Incrementality tests for streaming or OLV campaigns: AMC is the only place these data joins exist
- • Everything else in the operator's week (P&L, PPC ROI, inventory, fees, weekly reports): warehouse territory
A common 2026 stack: AMC for media planning and DSP attribution, Nova as the operational warehouse feeding the weekly business review, and a BI tool (or an LLM assistant) on top of Nova for exploration.
Cost and effort comparison
- AMC: Included with Amazon Ads accounts that meet minimum spend. Real cost is SQL engineering time: expect 1 to 3 FTE-weeks per meaningful query, plus a downstream warehouse and BI stack to schedule and visualize outputs.
- Nova: Starts at $29/mo, with SKU P&L and 100+ pre-modeled tables live in a few hours. Extends to a full ClickHouse warehouse for operators who want raw SQL access.
- Roll-your-own warehouse: 3 to 6 months of data engineering to hit parity on the ingestion side, plus ongoing maintenance every time Amazon changes an SP-API report format.
How to decide (a 60-second test)
Ask three questions:
- Does the question involve fees, COGS, or contribution margin? → warehouse (Nova).
- Does the question involve cross-DSP attribution or audience overlap? → AMC.
- Does the question need to reach an operator on Monday morning as a scheduled report? → warehouse either way, AMC does not do this alone.
Related reading: Nova data sharing, build your own analytics with Nova + Lovable, Amazon seller analytics guide.
Frequently asked questions
Ready to Transform Your Amazon Business?
Join thousands of successful sellers who use Nova Analytics to make data-driven decisions and maximize their profits.
Continue Learning
Explore more expert insights to grow your Amazon business
Best Fivetran Alternative for Amazon Data 2026
Fivetran moves data from A to B. But for Amazon, you're paying for generic plumbing and building everything yourself. Nova delivers 200+ pre-calculated KPIs with hourly refresh. No dbt, no data engineers required.
Best AMZScout Alternative 2026 - Profit Analytics Stack
AMZScout helps sellers find products. Nova shows which ones actually make money — reconciled against Amazon SP-API. Build a research + profit-tracking stack from $29/month.
Best Sellersprite Alternative 2026 - Profit Analytics Stack
Sellersprite finds opportunities. Nova validates which ones make money — reconciled against Amazon SP-API across all 21 marketplaces, from $29/month.