02 · Platform Fit

Platform fit.

Clean rooms, secure sharing, warehouse-native collaboration, walled gardens, and activation systems by job-to-be-done.

No platform is the strategy. The platform decision should follow the operating model — data footprint, governance rules, semantic needs, user workflow, and production path.

Scope this playbook → Run the 5-minute assessment →
PLATFORM FIT Your workload — its data footprint, governance rules, semantic needs, and user workflow are what should drive the platform choice. YOUR WORKLOAD data footprint governance rules user workflow Fit decision — no platform is the strategy; match the operating model to the environment that runs it best. FIT Cloud data platforms — warehouse-native collaboration and broad data + AI workloads close to where the data already lives. CLOUD DATA PLATFORMS Snowflake · Databricks · AWS · Google warehouse-native, broad data + AI workloads Specialist clean rooms — identity, media measurement, and walled-garden collaboration where match and output policy matter most. SPECIALIST CLEAN ROOMS LiveRamp · InfoSum · AMC identity + media measurement, walled gardens No platform is the strategy. Fit follows the operating model.
Three collaboration environments

Open · walled-garden · activation.

Three archetypes, not a ranking. Most enterprise workflows use a neutral home for the data, a walled-garden environment for ecosystem-specific measurement, and a destination system for activation.

  1. 01

    Open / neutral collaboration

    Match, measure, model, or activate across parties on a warehouse-native or vendor-neutral fabric.

  2. 02

    Walled-garden measurement

    Platform-owned environments scoped to a single ecosystem. Strong measurement, weaker portability.

  3. 03

    Activation & destination systems

    Where the governed output lands — audiences, suppression files, bid models, attribution feeds.

Platform deep dives

Where each environment tends to fit.

Once the output is clear, the platform decision becomes easier. These pages are not rankings. They explain where each environment tends to fit, where it does not, and what needs to be true before a POC can become production. Before choosing Snowflake, Databricks, AWS, or Google, decide whether the brand is actually solving a multi-cloud orchestration problem.

Start here

Multi-Cloud Orchestration

For brands using several data clouds, clean rooms, BI tools, media platforms, and activation systems. Start here when the real question is not which platform wins, but which platform owns which decision.

  • Data gravity
  • Output policy
  • Control plane
  • Cross-cloud workflow
Map the orchestration model
Output-led platform decision tree From the question "what output must the workflow produce?", six branches map an output to a better-fit platform or pattern. What output must the workflow produce? Raw or governed data access Snowflake · Databricks · AWS (by data estate) Joint analysis on sensitive data Clean room / DCR pattern Google campaign measurement ADH · BigQuery · GMP Installable workflow / protected IP Snowflake Native App or app workflow Lakehouse + AI/BI + agents Databricks AWS-native infra + ML AWS
  1. What output must the workflow produce?
  2. Raw or governed data access Snowflake · Databricks · AWS (by data estate)
  3. Joint analysis on sensitive data Clean room / DCR pattern
  4. Google campaign measurement ADH · BigQuery · GMP
  5. Installable workflow / protected IP Snowflake Native App or app workflow
  6. Lakehouse + AI/BI + agents Databricks
  7. AWS-native infra + ML AWS

Advisory guide, not a deterministic recommendation — the data estate, governance model, and commercial motion decide the path.

Data cloud and infrastructure environments

The collaboration runs close to where the data already lives — your data cloud, lakehouse, or hyperscaler. Strongest when the output is a governed dataset, a model feature, or an in-place query, and when data gravity and engineering ownership sit with your team.

  1. Snowflake

    Best when the workflow depends on data sharing, clean rooms, native apps, marketplace or private listings, and governed collaboration close to customer data gravity.

    • Marketplace
    • Native Apps
    • Data sharing
    • DCRs
    Open Snowflake guide

  2. Databricks

    Best when governance, discovery, open connectivity, AI/BI, semantic curation, ML, and agent-ready workflows need to operate around a lakehouse intelligence layer.

    • Unity Catalog
    • AI/BI
    • Mosaic AI
    • Open sharing
    Open Databricks guide

  3. AWS

    Best when the buyer is AWS-native and the workflow needs to sit close to S3, Redshift, Glue, Lake Formation, IAM, analytics, and ML infrastructure.

    • Clean Rooms
    • S3
    • IAM
    • SageMaker
    Open AWS guide

  4. Google / GMP / ADH

    Best when the workflow connects GA4, BigQuery, ADH, DV360, Google Ads, Meridian, and privacy-safe measurement into a marketing analytics path.

    • GA4
    • BigQuery
    • ADH
    • Meridian
    Open Google guide

Specialist, identity, and walled-garden measurement environments

The specialist environments are where ownership, neutrality, identity dependency, media signal gravity, and activation rights matter most. These are not interchangeable clean rooms — each has a different ownership model, data-movement model, and output policy.

  1. Amazon Marketing Cloud

    Best when the job centers on Amazon Ads, Amazon DSP, Amazon retail / commerce signals, custom SQL analysis, audience creation, and privacy-safe media measurement.

    • AMC
    • Amazon Ads
    • DSP
    • Retail media
    Open AMC guide

  2. Meta Advanced Analytics

    Best when the job centers on Meta media measurement, first-party signal enrichment, custom attribution, campaign overlap, path-to-purchase, lift, Advantage+ insight, and privacy-safe optimization inside the Meta ecosystem.

    • Meta AA
    • CAPI
    • Lift
    • Attribution
    • Advantage+
    Open Meta AA guide

  3. InfoSum

    Best when the collaboration requires a neutral, privacy-enhancing clean room model built around non-movement of data, decentralized processing, strict permissions, and multi-party analysis without centralizing raw data.

    • Non-movement
    • PETs
    • Neutral clean room
    • Multi-party
    Open InfoSum guide

  4. LiveRamp

    Best when identity resolution, RampID-enabled matching, governed clean room collaboration, activation, measurement, and cross-cloud interoperability are central to the workflow.

    • RampID
    • Identity
    • Activation
    • Clean Room
    Open LiveRamp guide
Specialist comparison — role maps, not rankings
Platform Primary role Ownership implication Best first use Main watch-out
InfoSum PET-led non-movement collaboration WPP / GroupM Multi-party planning, activation, and measurement without raw data movement Architectural neutrality and ownership neutrality are different questions
LiveRamp Identity-led collaboration, activation, measurement Publicis (agreement pending approvals) RampID-enabled activation or measurement across partners Identity dependency and agency neutrality must be explicit
Amazon Marketing Cloud Amazon-centered media / retail-media clean room Amazon Ads Amazon Ads, DSP, retail / commerce measurement and audiences Not a general enterprise collaboration layer
Meta Advanced Analytics Meta-specific, partner-mediated walled-garden measurement Meta (partner-mediated) Meta media measurement, CAPI signal, attribution, lift, Advantage+ insight Not a self-serve product or general clean room; access is permissioned

Meta AA, Google ADH, and AMC are not interchangeable clean rooms. Each is strongest when the decision depends on its own media signal gravity and privacy rules.

Capability fingerprint — illustrative
Specialist capability fingerprint — illustrative A 6-axis radar comparing LiveRamp, InfoSum, Amazon Marketing Cloud, Meta Advanced Analytics across Identity, Activation / reach, Measurement, Neutrality (2026), Cloud breadth, Privacy / PETs. Each platform peaks on a different axis, but all three score low on neutrality in 2026. LiveRamp: neutrality 2 of 5; InfoSum: neutrality 2 of 5; Amazon Marketing Cloud: neutrality 1 of 5; Meta Advanced Analytics: neutrality 1 of 5. Identity Activation / reach Measurement Neutrality (2026) Cloud breadth Privacy / PETs
  • LiveRamp
  • InfoSum
  • Amazon Marketing Cloud
  • Meta Advanced Analytics

Illustrative advisory shorthand, not a vendor benchmark — the point is the shape, not the score. All four sit low on neutrality: LiveRamp (Publicis, pending), InfoSum (WPP-owned), AMC and Meta AA (walled gardens). Score them against your own use case.

This is not a ranking. These are role maps. No environment is “best” — each fits a different output, data-gravity position, and governance model. Most real programs use more than one. Start from the decision the data must improve, then place the platform.

Start with Multi-Cloud Orchestration

Capability patterns are representative. Validate current product availability, regional support, preview status, account requirements, and privacy controls against official documentation.

Output-led decision rules

Design backward from the output.

  1. 01

    If the output is insight only

    Warehouse-native collaboration or a clean room with strong query / aggregation.

  2. 02

    If the output is activation

    Pair the clean room with a CDP or DSP destination — activation rights matter.

  3. 03

    If the buyer is Google-centric

    ADH / PAIR for measurement; the wider data product still needs a neutral home.

  4. 04

    If the product is a model

    Model-as-a-service or DCR-resident model; raw data should not move.

  5. 05

    If the product is a workflow

    Native app or workflow inside the buyer's stack; the UI carries the value.

  6. 06

    If multi-party trust is the unmet need

    Specialist clean room (InfoSum, LiveRamp Clean Room) before warehouse path.

DCR solution-fit matrix

Capabilities by platform.

Use this matrix as a planning tool, not a vendor scorecard. 15 capability rows × 6 platforms. Validate against current product documentation and the buyer's actual environment.

Capability SnowflakeDatabricksLiveRampInfoSumAWS Clean RoomsADH / BigQuery
Deterministic ID matching Possible Custom joins Strong Strong Strong Ecosystem-specific
ID graph enrichment Possible Possible Strong Partner-led Possible Limited
Multi-party matching Possible Partner-led Strong Strong Strong Limited
Audience enrichment Possible Possible Strong Partial Possible Limited
Media activation Connector-led Partner-led Strong Not ideal Custom Google stack
Measurement & attribution SQL-led Model-led Strong Strong Strong Strong (Google)
Consent & privacy enforcement Policy-dependent Unity Catalog Built-in Policy-dependent Custom Google rules
Lookalike / segmentation Snowpark / ML Strong ML Strong Not ideal ML integration Platform-led
Automated refresh / scheduling Strong Strong Strong Strong Strong Campaign-based
Data export Strong Strong Strong Limited API / custom Aggregated only
Discovery & catalog Horizon + tagging Unity Catalog Partner-managed Limited Glue / Lake Formation Google catalog scope
Business semantic layer Semantic views (emerging) Genie / Unity metrics Partner-led Not native BYO (QuickSight) SQL templates only
Lineage & observability Account usage + Horizon Unity lineage + MLflow Partner-managed Audit logs CloudTrail / CloudWatch Job logs only
Agent / tool exposure Cortex + Snowpark APIs AI Functions + MCP Partner integrations Not yet SDK integration SQL API only
Evaluation & monitoring Cortex Analyst evals (emerging) MLflow + Lakehouse Mon. N/A N/A BYO (SageMaker) Limited
Delivery

Share, clean room, native app, or API?

The delivery shape is decided by the buyer's stack, governance posture, and the kind of output that creates value.

  1. 01

    Governed share

    Snowflake share, S3, BigQuery export, DCR-mediated table. Buyer enriches their own model.

  2. 02

    Clean room

    Match, query, model inside a governed environment. Raw data does not leave.

  3. 03

    Native app / workflow

    UI, templates, permissions, repeatable execution inside the buyer's stack.

  4. 04

    API

    Per-call scoring, model-as-a-service, embedded enrichment.

Google Marketing Platform

ADH, PAIR, DV360, BigQuery.

The Google stack is best framed as a measurement and ecosystem-specific environment — not a general clean room. ADH / PAIR are most useful when the buyer's question is fundamentally about Google or YouTube exposure.

ADH
Privacy-safe Google / YouTube measurement environment with strict aggregation and output rules. Best for closed-loop measurement against Google media.
PAIR (DV360)
Privacy-safe advertiser-publisher first-party reconciliation inside DV360. Not a general analytics clean room.
BigQuery
Neutral warehouse — the home for data that lives outside ADH. Pair with ADH for the measurement step.
ADH 7-step POC pattern
Hash match keys → BigQuery staging → match inside ADH → strip keys → matched exposure → measurement → output.
Watch-outs

Platform-specific things to know.

  1. Snowflake

    Strong neutral fabric, but media activation and clean room workflows still need partner connectors.

  2. Databricks

    Strong on ML, semantic, and agentic surfaces; matching + cross-party activation are partner-led.

  3. BigQuery / ADH

    Best when the measurement question is Google/YouTube-centric — not a general clean room replacement.

  4. AWS Clean Rooms

    Strong for matching + measurement; activation and semantic layer are BYO.

  5. LiveRamp

    Identity + activation are mature; semantic and agentic layers are partner-led.

  6. InfoSum

    Federated by design; activation, export, and semantic surfaces are limited.

Right environment chosen? Package it for the buyer.

Platform fit is half the puzzle. The other half is whether you're selling a data product, model, app, or workflow — and how the budget moves through Land → Bridge → Anchor.

Scope this playbook → Next: Use Cases → Back to overview →