One system expanded into multiple domains for marketing agencies

Agencies keep asking me the same question.

"Can you build us a system that does X?"

X is always different. Instagram reports. Amazon competitor analysis. LinkedIn outreach drafts. TikTok shop intelligence. Meta ads library intelligence.

The engine is always the same.

Scrape. Store. Analyze. Output.

Here is what I learned building custom automation for agencies.

1. Build the data layer first.

Supabase with client_id on every table. New request comes in. I already have the schema. Add a new source. Same database.

2. Keep the interface stupid simple.

Telegram bot. One command. No dashboard. If they have to learn a new UI, they will not use it.

3. Batch the AI calls.

One prompt. 30 items. Structured JSON. Not 30 separate calls. Cost drops 70%.

4. Deliver where they already work.

Email. Telegram. Slack. Google Slides. Do not make them open another app.

5. The first request is never the last.

Cookie Monster started as Instagram competitor reports. Same engine now runs Amazon ASIN analysis, LinkedIn message drafting, and TikTok shop research.

The agencies that win are not the ones with the most tools.

They are the ones with the fewest steps between data and decision.

5 comments

One system expanded into multiple domains for marketing agencies