Particular Audience Particular Audience
Talk to us

AI for search, personalisation and retail media

A different store for every customer.

Bring what each shopper wants together with what your business needs.

Search, recommendations, retail media and promotions across your digital store and AI shopping experiences.

Retail Media Reporting & Forecasting — API, MCP and RevenueOS (Beta) | Particular Audience

One retail media reporting record, used three ways: the Reporting API, the open-source Reporting Tool and DiscoveryOS MCP. RevenueOS Forecasting (Beta, Lighthouse engine) turns that history into low, expected and high ranges.

Reporting & Forecasting

One reporting record, three ways to use it: pull it through the Reporting API, report on it with the open-source Reporting Tool, or ask about it from an AI assistant with DiscoveryOS MCP. RevenueOS then turns that history into forecasts.

RevenueOS Forecasting is in Beta. Ranges get better as reporting history builds up.

Retail Media Reporting API: The source record: campaign, ad set and SKU delivery and spend, pulled with your own credentials. Does not forecast or invoice.

Retail Media Reporting Tool: Open-source tool that extracts Reporting API metrics into your own reports. Does not change campaigns.

DiscoveryOS MCP: Ask about campaign performance from an AI assistant, within the access you already have. Does not launch or edit campaigns.

RevenueOS Forecasting: Turns reporting history, bookings and pipeline into low, expected and high ranges. Does not book, commit or change floors.

Retail Media Billing Extract: Monthly invoice-ready file: delivered spend for performance lines, booking for Fixed Tenancy. Does not send invoices.

Inventory forecast: Impressions, clicks or spend over 30, 90 or 180 days, with low, expected and high ranges, a back-test once there are 90 days of history, and data-cleaning checks. Not booked or invoiced revenue.

Pipeline scenarios: Campaign budgets by month, weighted by deal stage, against your annual target, with best and worst cases.

Delivery pacing: Projects each campaign's finish and any shortfall. Campaigns without delivery reports show unknown, never zero.

Placement availability: Weekly bookings and holds against recorded capacity and allowed overbooking. Missing capacity shows unknown.

Finance view: MRR, ARR, invoicing and expected cash over 6, 12 or 18 months.

Lighthouse, rebuilt: RevenueOS now runs the rebuilt Lighthouse forecasting engine. It blends several models (seasonal regression, gradient-boosted trees, year-over-year and seasonal-naive), weights them by how well each scores on rolling back-tests, and returns low, expected and high ranges.

Being folded into RevenueOS: Inventory per placement and device, with booked and available impressions; Share of voice and scroll reach in sellable inventory; Revenue ranges for CPC, CPM, CPA and fixed deals, with budget caps; Accuracy checks against history, beating a simple seasonal baseline.

PA provides: Campaign, ad set and SKU delivery and spend (Reporting API); Placement formats, prices and inventory rules; Bookings and holds already in RevenueOS; Measured click and win rates; The models, back-tests and data-cleaning checks.

PA depends on the retailer for: Site traffic per placement and device (for example Google Analytics); How far shoppers scroll on each page; Recorded placement capacity and overbooking limits; Annual targets and deal stages; Enough history: back-tests start at 90 days.

Where an input is missing, the result shows unknown rather than a guess.

The forecast screens are self-serve once the data is connected. During Beta, PA onboarding connects the Reporting API and your site analytics and checks placement capacity with you.

Inventory forecasts are projections of reported delivery and spend: not booked or invoiced revenue.

PA only alerts AdOps about pacing and never changes bid floors or caps. Make-goods are the retailer's decision and cost, and nothing books or commits on its own.

Is forecasting self-serve? The forecast screens are self-serve once the data is connected. During Beta, PA onboarding connects the Reporting API and your site analytics and checks placement capacity with you.

Is a forecast booked revenue? Inventory forecasts are projections of reported delivery and spend: not booked or invoiced revenue.

What happens if we have no capacity or traffic data? Where an input is missing, the result shows unknown rather than a guess. Availability and inventory stay unknown until capacity and traffic are connected.

Does PA act on pacing shortfalls? PA only alerts AdOps about pacing and never changes bid floors or caps. Make-goods are the retailer's decision and cost, and nothing books or commits on its own.

What is Lighthouse? Lighthouse is PA's retail media forecasting engine. Its rebuild now runs inside RevenueOS, and its planning screens are being folded into RevenueOS during Beta.

Where does reporting data come from? The Retail Media Reporting API: campaigns, ad sets and SKUs. The Reporting Tool, DiscoveryOS MCP and RevenueOS all read that same record.