1stli connects every source, answers in plain English, and exposes one MCP endpoint — so ChatGPT, Claude, and your own agents all run on the same trusted numbers.
how did paid acquisition compare to organic in q2, by repeat revenue?
Organic search drove 2.2× higher 90d LTV than paid channels. Display fell out of profitability; reallocate to organic + email.
| channel | new | repeat £ | 90d LTV | ret. |
|---|---|---|---|---|
| organic | 4,812 | £284k | £2,140 | |
| 2,988 | £167k | £1,590 | ||
| paid social | 3,201 | £198k | £1,720 | |
| display | 887 | £41k | £980 |
WITH channel_revenue AS (
SELECT ga.source AS channel,
COUNT(DISTINCT o.customer_id) AS new_customers,
SUM(o.revenue) AS repeat_rev,
AVG(l.ltv_90d) AS ltv_90d
FROM ga4.sessions ga
JOIN shopify.orders o USING (customer_id)
JOIN klaviyo.ltv l USING (email)
WHERE o.created_at BETWEEN '2024-04-01' AND '2024-06-30'
GROUP BY ga.source
)OAuth your stack in. We keep it fresh and queryable.
Same endpoint for your team and your agents.
We plan across sources and write the joins.
Tables for humans. SQL for engineers. Both, every time.
One MCP token. Your agents query like a function.
show me Q2 acquisition channels ranked by 90-day LTV, with retention curves and total repeat revenue.
Organic search and email lead on retention. Display has fallen below your blended payback target — 2.2× LTV gap vs. display.
| channel | new | repeat £ | 90d LTV | retention |
|---|---|---|---|---|
| organic | 4,812 | £284k | £2,140 | |
| 2,988 | £167k | £1,590 | ||
| paid social | 3,201 | £198k | £1,720 | |
| display | 887 | £41k | £980 |
SELECT ga.source, SUM(o.revenue) AS repeat_rev
FROM ga4.sessions ga
JOIN shopify.orders o USING (customer_id)
WHERE o.created_at BETWEEN '2024-04-01' AND '2024-06-30'
GROUP BY ga.source ORDER BY repeat_rev DESCTwo paths exist today. Neither works.
Plug an agent into ten MCP servers and its context blows up. Joins are guessed. Numbers come back confidently wrong.
dbt, Cube, LookML — define a metric for every question, in YAML, before anyone gets an answer.
Every source wired in. Built-in AI querying in plain English. One MCP endpoint that ChatGPT, Claude, and your own agents all call — same data, same answer, everywhere.
OAuth your stack in 90 seconds.
218 native sourcesPlain English. We write the SQL.
no modeling requiredOne MCP endpoint. ChatGPT, Claude, or your own agents.
mcp-ready"We deleted four reporting tools the week we onboarded. Our LTV question, answered in 18 seconds, used to take a sprint."
"As an agency we run 22 clients in 1stli. Same prompt, twenty-two workspaces, twenty-two answers. The economics broke our spreadsheet."
"The SQL transparency is what sold engineering. Our agents call it like a function and we can audit every join. Blended CAC fell 28%."
Three sources free. No card. No sales call.