Cites your own notes
The narrative points at the exact document and date — no guessing which change caused the dip.
Allocate budget, set targets, and optimise autonomously across Google, Microsoft and Meta — all from one engine that understands marginal ROAS.
New · GA4, Shopify and HubSpot revenue routingLink Google, Microsoft and Meta Ads in one OAuth flow. ROASt pulls campaigns, budgets, and conversion actions automatically — no CSVs, no API keys, no manual setup.
ROASt scans every account on connect and auto-fixes broken conversion setups. Anything it can’t fix gets flagged for review with a one-click resolve.
The Strategic Mix Allocator fits Hill response curves per platform. Given a total budget, where should each pound go? These curves feed the live nightly allocator.
Marginal-ROAS scoring redistributes campaign budgets every night. Cards run, recommendations stage, you review and approve.
Approved changes flow back to Google, Microsoft and Meta via their APIs. No copy-paste, no spreadsheet exports, no manual edits.
Each push is observed against actual outcomes. The engine’s confidence in your portfolio compounds with every cycle.
Flume is your ROASt Labs assistant. Ask about performance, compare campaigns, generate reports, or manage portfolios — all in natural language. Or hand the controls to your own LLM via MCP — ROASt becomes the data layer and safety net, your LLM the controller.
Upload your landing pages, brand briefs, dated promos, and client notes. When an anomaly lands, ROASt cross-references the drop against everything it knows — events, platform changes, creative refreshes, URL shifts — and writes a one-paragraph narrative you can paste into a client reply.
The narrative points at the exact document and date — no guessing which change caused the dip.
Every Monday at 7am client-local: KPIs, what changed, what to act on. One page, ready to forward.
Connect Claude, ChatGPT, Gemini, or any MCP client to ROASt’s 24 read-only tools. Have an agent run your Monday check before 9am.
The engine is pure math and platform-agnostic, no LLM involvement. Adaptors are trained on platform behaviour and apply nuance and safeguarding to campaign scoring.
The widest auction surface, with the most learning-phase quirks. Search, Performance Max, and Shopping each get their own ceiling.
Native learning stages, dwell windows, and creative cycles — respected in their own language, not papered over.
Smaller auction volume means more caution, more compounding, and a harder watch on budget that isn’t spending.