Felix · Performance Forecaster
Forecasting
Forecasts revenue 90 days ahead, retrained nightly on your accounts.
- 90-day revenue forecasts, updated nightly
- Holt-Winters fit to your own history
- Confidence scoring on every prediction
Your agents. Each specialized. All in sync. Built for ecommerce brands.
Each agent owns a function. Together they run your marketing.
Why agents
Dashboards report. Agents decide.
The marketing stack you have today is mostly a reporting stack. Dashboards show you what happened. Attribution tools show you what they think happened. Your agency shows you a slide deck on Friday.
Cresva is built different. Agents that don't just show you the data, they make decisions on top of it. Every day. Before you finish coffee.
More on the structural shift on /why-agents →
Meet the agents
Your agents. Each one ships.
- FelixForecast
- ParkerAttribute
- SamPlan
- MaxProfit
- OliviaBrief
- DanaReconcile
- MayaRemember
- DexReport
- IrisSell
- RemyRetain
Parker · Attribution Analyst
Attribution
De-biases attribution against platform self-claims so ROAS reads true.
- Multi-touch attribution across Meta, Google, Shopify, GA4
- De-biased ROAS, no platform inflation
- Daily reconciliation against finance numbers
Sam · Media Planner
Scenario testing
Tests budget scenarios before you spend, surfaces elasticity by channel.
- Budget shifts modelled against elasticity and your own history
- Elasticity discovery across channels
- Risk-free what-if analysis before committing spend
Olivia · Creative Strategist
Creative intelligence
Reads creative DNA across your ads, flags a creative wearing out from its own click-through.
- Creative DNA extraction across all your ads
- Flags a creative whose click-through fell 20% or more against its own first 7 days while frequency rose
- New variant generation from your top performers
Dana · Analytics Engineer
Unified data
Runs the data layer. Reconciles platforms, monitors tracking, fixes gaps.
- Single source of truth across 6+ data sources
- Daily reconciliation of conversion gaps
- Tracking-health monitoring (Pixel, GA4, server-side)
Maya · Account Lead
Memory & context
Holds brand memory. Decisions, preferences, CAC ceilings, the why behind moves.
- Brand memory across all conversations
- Auto-extracts decisions, preferences, CAC ceilings
- Human-approval workflow for memory updates
Dex · Reporting Manager
Delivery & alerts
Ships the agents' outputs to the surfaces your team already uses.
- Reports to Slack, Sheets, Slides, Docs, Notion
- Anomalies in the Daily briefing
- Weekly recaps that get read
Max · Margin Analyst
Profit & unit economics
Connects every ad dollar to profit after COGS, shipping, returns and fees.
- Contribution margin per order, not blended ROAS
- Promo and discount P&L, including pull-forward
- Product profit after its cost and the ad spend of its matched orders
Iris · Commerce Analyst
AI assistant sales
Watches how AI shopping assistants meet your store, and keeps agent prices above your margin floor.
- Shopify products kept in step with the Meta catalogue
- Agent storefront, offers, bundles and negotiation
- Agent orders counted from the order records
Remy · Lifecycle Marketer
Email & retention
Runs your email flows and lists, and flags a flow that stops earning.
- Email flows and campaigns, synced every six hours
- Lists, segments and customer tags
- Subscribers from your subscription app
A Tuesday morning
What the agents do while you sleep.
An example workday, agent by agent. Every figure below is illustrative, not a measurement. The stack reads like a thread of decisions, not a stream of dashboards.
Sun · 9:14 PM
FelixForecast
Forecasts the week ahead. Sees Q4 momentum slowing in evergreen products.
Example
Mon · 7:32 AM
ParkerVerified
Reconciles weekend conversions. Catches platform overclaim before it reaches the morning brief.
Example
Mon · 8:01 AM
SamTested
Tests three budget scenarios. Recommends $2K shift from prospecting to retention.
Example
Mon · 8:26 AM
MaxPriced
Prices Sam's shift against contribution. Retention nets 18% margin, prospecting 4%.
Example
Mon · 9:47 AM
MayaRecalled
Pulls your CAC ceiling from last quarter ($47). Notes current campaign at $52.
Example
Mon · 11:30 AM
OliviaFlagged
Detects fatigue on Hook_v4. Generates 6 new variants from top-performer DNA.
Example
Mon · 2:10 PM
IrisCountered
An agent asks for more off than the margin floor allows. Counters at the floor instead.
Example
Tue · 6:22 AM
DanaPatched
Reconciles the data layer. Catches GA4 missing 4% of conversions. Patches.
Example
Tue · 6:30 PM
DexDelivered
Delivers the weekly recap to your Slack. Three decisions logged. Two alerts surfaced.
Example
Tue · 7:05 PM
RemyFlagged
Flags an email flow that drops to half its usual sends.
Example
Without Cresva
What the agents replace.
The math underneath the agents. Tools, roles, and the hours you stop spending on the dashboard job.
Tools you cancel
- Triple Whale ($219 to $749 a month, per its Shopify App Store listing)
- Northbeam (Starter listed at $1,500, per its pricing page)
- Custom attribution dashboards
- Looker or Metabase setup time (weeks)
- Spreadsheet-based scenario testing (broken every quarter)
Roles you don't need to hire
- Junior analyst running daily reports
- Growth strategist running scenarios manually
- Data engineer maintaining attribution pipeline
- Marketing operations role
Work you stop doing by hand
- Reviewing dashboards platform by platform
- Writing the Monday recap
- Checking creative performance ad by ad
- Sanity-checking attribution against the store
- Rebuilding reports when a pipeline breaks
What's underneath
Real infrastructure. Not a wrapper.
Most AI marketing tools are GPT wrappers. Cresva runs on purpose-built ML infrastructure for each agent.
Memory that compounds
Every conversation, decision, and outcome gets extracted into a brand memory store. CAC ceilings, channel preferences, what worked last quarter. Human-approval workflow on every memory write.
Custom retrieval-augmented memory store
De-biased attribution
Multi-touch reconciliation across Meta, Google, Shopify, GA4. De-biased against platform self-attribution. Daily reconciliation against your finance numbers.
Dedicated debiasing and monitoring layer
Forecasting with stated uncertainty
A Holt-Winters model with damped trend, fitted to your own history, returning prediction intervals rather than a point estimate. Updated nightly. Backtested against your last 12 months.
Prediction intervals, not a single number, retrained nightly
Where decisions land
Slack, Google Sheets, Google Slides, Google Docs, Notion. Reports formatted per surface. Anomalies arrive in the Daily briefing each morning.
Twelve delivery integrations, formatted per platform
Ready when you are
See an agent in action, or run a 7-day pilot.
Each agent has a deep dive. The pilot connects every agent to your real accounts.
Looking for something specific? See Felix forecasts → or Parker attributes → or Sam tests scenarios →.