Autonomous Data Platform,
built for Growth.

Ember runs your warehouse and tells you in Slack what to test next.

How it works

Signups are paused. Join the waitlist for access.

Warehouse / coreLast sync 6:10 AM · next 18:10

Warehouse

27 tables from 4 sources · 1.9 GB

Postgres13 tablesSupabase6 tablesStripe7 tablesSlack1 table

Tables24

core.customersPostgres · Stripe12,480 rows6:10 AM

core.subscriptionsStripe3,912 rows6:10 AM

events.checkout_eventsPostgres12.4m rows6:08 AM

Metrics3

metrics.trial_to_paid_7dPostgres · Stripe11.4% · 7-day avg6:12 AM

metrics.refund_rate_30dStripe1.7% · 30-day6:12 AM

metrics.trial_first_session_dailynewdbt6,412 rows9:58 AM

Backed by

Built by engineers from

You have the data.Nobody has time to read it.Ember does.

A data team forthe whole company.

Ember works from your Slack, with its own warehouse behind it. Here is what a month with it looks like.

Ember learns what you care about.

It reads your warehouse and finds the numbers that matter for your checkout on its own. Say one thing in Slack to add anything it missed.

Cayden Le9:14 AM

I own mobile checkout. Trial → paid is the number I watch, and refund rate is the guardrail.

EmberAPP9:15 AM

Got it. I’ll keep an eye on both, starting today:

Trial → paid is 11.4% over the last 7 days, from subscriptions and checkout_events
Refund rate is 1.7% over the last 30 days, from refunds

I’ll message you here when I find something worth trying.

Ember brings you something to test.

It reads your warehouse every day and says nothing until it finds something worth acting on. On day 23 it did: iOS trials with notifications on convert at 18%. The rest convert at 8%.

23 days read. One proposal.

Every proposal comes with its evidence.

Cayden asked the obvious question: are the engaged trials just the ones turning notifications on? Ember checked only trials that came back three or more times in week one. The gap held.

EmberAPP9:41 AM

Worth a test: iOS trials that turn notifications on convert at 18.2%, everyone else at 7.9%. Only about a third ever get the prompt, since the app asks after the paywall.

Cayden Le9:42 AM

Isn’t that just the engaged people turning them on?

EmberAPP9:44 AM

Some of it, yes. If I only look at trials that came back 3+ times in week 1, it’s 21.3% vs 13.1%. Smaller gap, but it’s still there.

Ember proposes the experiment.

It turns the finding into a test it can size: ask for notifications before the paywall, on half of iOS trials, nine days to see a two point move. You approve in Slack and it starts.

EmberAPP9:52 AM

Since most trials never see the prompt, moving it ahead of the paywall is a cheap thing to try. I’d run it on half of iOS before changing it for everyone:

Proposed experimentAsk for notifications before the paywallHalf of iOS trials, about nine days to see a 2 pt move in trial → paid. I’ll stop it if refunds pass 3%.

ApproveEditNot now

Cayden Le9:58 AM

Approved. Go.

EmberAPP9:58 AM

It’s live as EXP-016. I’ll post the readout on Sep 12, or sooner if it’s clear.

Ember builds the tables it needs.

Ember needed a table of what each trial did in its first session, and there was none. It wrote the dbt model and ran the tests. Tomás approved it before it went live.

app_eventsPostgres · 31.6m rows · synced 6:10 AMsubscriptionsStripe · 3,912 rows · synced 6:10 AM

Written by Ember · approved by Tomás Reyestrial_first_session_dailydbt model · 6,412 rows · 8 tests passed · built 9:58 AM

Live 9:58 AMIn production

Ingest anything,analyze everything.

Connect the tools you already use. Ember loads them into a warehouse it runs for you and answers every question from there.

Have a data warehouse already? Embrasure Enterprise

Autonomous Data Platform,built for Growth.