How it works
One version of the truth.
- We connect your systems.
- We keep one model of your business: products, stock, orders, costs and customers.
- Your people and your AI agents ask the same model.
/0.1
Your systems, connected and maintained.
Connect loads data from the systems you already run into the Observatory. Read-only access wherever the provider allows it. Every grant is visible in your own tools and can be removed there.
- Managed integrations We build and maintain the connections, so your team does not run pipelines or fix broken feeds.
- ERP and custom systems We map complex ERPs, legacy accounting software and custom feeds. ERP mapping takes two to three weeks from production access.
- Read-only by default Read-only access wherever the provider allows it. Every grant is visible in your own tools and can be removed there.
- Secrets in a vault Credentials are kept in an encrypted, self-hosted vault.
Secrets in an encrypted, self-hosted vault.
/0.2
One model of products, stock, orders and customers.
The Observatory is the linked model of your company. Dataloader loads your ERP, webshop, advertising, analytics and finance data and links it into one model. Dashboards, AI assistants and agents all read from it.
- Linked business records Source identifiers are linked to business records such as products, orders and stock. Stock from the ERP, sales from the webshop and spend from Google Ads meet on the same product.
- KPI Register 150 KPIs across 11 domains, each with a definition, formula, owner and source of truth. Your KPIs are rated High, Reasonable, Low, Very low, No data, based on your connected data.
- Provenance on every value Every value is tagged as observed, declared, estimated, derived or generated, with its source and time. You can always see where a number comes from.
- Built for scale 4,000+ non-personal API fields mapped across 7 sources and 230+ tables. 79,065 products in one linked catalogue for a single customer.
79,065 products in one linked catalogue for a single customer
/0.3
Catch errors before they cost money.
94 automated checks on content gaps and data anomalies, with a completeness score per product, per category and per vendor. Intent-based enrichment then writes product data for the questions buyers ask, structured for AI search (GEO). Every generated value stays tagged as generated.
- Catalogue audit and issue detection 94 automated checks for product data gaps, plus AI detection of site issues, GA4 tracking problems, and Google Ads or Merchant Center errors.
- Cost price checks We spot cost price mismatches between the ERP and the webshop, and products in the wrong lifecycle stage.
- Intent-based enrichment for SEO and GEO We identify the intent for your site, then restructure and enrich the catalogue around it: categories, attributes and product texts, maximised for SEO, GEO and AI assistants. Thousands of products at once, tagged as generated.
- Review and Syncback Issues are listed for review. Approved values are written back to your systems where agreed.
Wrong cost prices caught early
/0.4
Your data, ready for your AI.
Work with your data in Claude, ChatGPT or your own agents. Ask questions, run analysis and build dashboards in Dataloader or as Claude artifacts, all from the same model, through the Dataloader MCP server and API.
- Dashboards in Dataloader or as artifacts Ask for a dashboard in plain language. Publish it in Dataloader for your team, or keep it as a Claude artifact you can share. Both read the same model.
- MCP server Claude, ChatGPT and your own agents read the Observatory model through MCP.
- Provenance in every answer Values come back with their source, ingest run and time, so an answer can be checked.
- EU language-model policy Language models run with providers in the EU or EU-hosted regions, receive only redacted, pseudonymised data, and are contractually barred from training on your data. Generated values are tagged so they stay distinguishable from observed ones.
Language models are contractually barred from training on your data.
/0.5
Put the model to work.
Fleets are workflows and AI agents that run on the Observatory model: the Labelizer, enrichment, the product advisor and AI search. What they do is agreed per customer.
- The Labelizer Labels each product family by ROAS from your ad accounts and GMROI from stock and margin: Scale, Keep, Watch, Fix or Dead.
- Product advisor An agent on your product pages that answers shoppers from the product model.
- AI search Search on your webshop that reads the same product model.
- Built per scope Workflows and agents are agreed per customer, and run on the Observatory model.
Labelizer, the product advisor, AI search and enrichment run in production on customer data.