Your stack, your channels, a more persuadable audience.
atlas is the audience intelligence system behind every enterprise engagement. It works with the data you already hold, resolves it to people and households, scores whose behavior will change and why, and returns the answer to the channels you already buy. Nothing to replatform.
Your CRM record
No reason. No timing. No message.
Leaves your environment as
- Email
········ - Phone
········ - Name and ZIP
········
SHA-256, hashed on your side. Raw records stay put.
Back in your CRM
Behavioral layer
Lead with
Delivered to
- CRM and email
- DSP and social
- Warehouse and BI
Architecture
Hashed keys cross. Your raw records stay put.
atlas sits between your systems and your channels. Only hashed keys cross the match boundary. Scores come back to where your teams already work, and the outcomes you agree to share return every day to check and retrain the models.
Your environment
CRM or CDP
Customer file: email, phone
Data warehouse
Share or export, hashed keys
Site and app
Pixel, first-party events
Media stack
Seed audiences, DSP and social
Outcome systems
Conversions, funded accounts, enrollments
atlas · audience intelligence system
Resolve
270M+ U.S. adults resolved · one key per person and household
- Hashed email and phone
- UID 2.0
- Household key
Gravity · model engine
Calibrated 0 to 1 · versioned · digital activity refreshed daily, graph rebuilt monthly
- Propensity
- Motivation
- Decision style
- Receptivity
- Persuadability
- Publish gate
Organize and evaluate
Audience builder, holdout split, baseline checks
- Thresholds
- Neighbor separation
- Holdout split
- Cohort cards
Your channels
CRM and email
Scored records and a next-best message
DSP and social
Hashed audiences, seeds and suppressions
Warehouse and BI
Scores joined to your own models
Clean room
Match, score and evaluate in place
Outcome labels return
Conversions feed the daily retrain
Runs where your data already lives.
Runs on
- Databricks
Delivers to
- Meta
- Google Ads
- TikTok
- Snowflake
- LiveRamp
- Snapchat
- X
- Microsoft Advertising
- Amazon Ads
- Yahoo DSP
Connects with
- Salesforce
- HubSpot
- Marketo
- Braze
- Mailchimp
- Twilio
- Segment
- mParticle
- Google BigQuery
- Amazon S3
- Shopify
- Google Analytics
And any DSP, clean room or platform that takes hashed files or UID 2.0.
What it does
Six ways we move a number you already report.
Each is scoped to one decision, measured on a holdout you control and read out on day 60.
- 01
Find who isn’t in your file
Score 270M+ U.S. adults for readiness on your decision and deliver the people most likely to act whom you aren’t already reaching.
MeasureCost per acquired customer against your current audiences
- 02
Say the thing that moves them
Group the audience by motivation and attach a message hypothesis to each group. Your creative team tests the variants.
MeasureConversion-rate lift by motivation group, in a controlled test
- 03
Reach them when they’re ready
Readiness concentrates in windows of the day and week. Media and outreach run in those windows instead of the category default.
MeasureCost per outcome against your current flighting
- 04
See the change inside your base
Recognize changing needs within existing customers: a consumer becoming a business, a patient whose regimen is shifting.
MeasureAdditional product relationships and cross-sell rate
- 05
Catch switching before it happens
Comparison behavior that comes before a switch triggers outreach while there is still a decision to influence.
MeasureRetained relationships or balances against a control
- 06
Prove it against your own model
Held-out evaluation of your current model with and without Moonbrush signals, then controlled activation on the agreed outcome.
MeasureAUC and top-decile lift, then incremental outcomes on your holdout
The engagement
Sixty days to an answer you can audit.
Every engagement starts as a proof of concept on your own data. If it doesn’t beat the baselines on your holdout, you’ll know on day 60.
- Week 0
Agree the decision, the data and the success criteria. Paperwork includes the data use agreement and the acceptance criteria.
- Weeks 1–2
Match keys and attach scores in the clean room or your warehouse. Your match rate is reported before anything is built.
- Weeks 3–8
Activate in your channels against a holdout you control. Outcomes return as they happen.
- Day 60
Read-out on lift and impact, with the validation report. Then an annual program, if the numbers earn it.
Integration
We connect to your warehouse, CRM, clean room or media stack, matching your DSP’s spec when it has one.
Data onboarding
Your file is matched under one identity key, and you see the match rate before any model is built.
Applied intelligence
A Moonbrush team builds and validates the models with you, and writes the cohort cards and message hypotheses.
Enablement
Your analysts get the scores, the validation report and view-only dashboards or API access to keep using them.