Your ad platform knows what you spent. Your website knows who signed up. Neither one alone can tell you what a sign-up cost — and that’s usually the number you actually need. Everything you bring into BumbleB lands in one workspace, so a single question can span two datasets: website behavior with ad spend, store events with orders, your database with your traffic. Crunch works out how the two relate before it answers.
What you’ll need
- Two datasets in your workspace. A dataset is one table of your data: events from your site or store, a file you uploaded, or a table you linked. Not sure which route fits your data? See Which route for which data.
- A column they share — a campaign name, an order ID, a hashed email. This is how Crunch knows which row in one dataset belongs with which row in the other.
Step by step
- Find the column they share. Look at both datasets and pick the one thing they have in common. For ad spend and website visits it’s usually the campaign name; for orders and inventory it’s the product.
- Name it the same way in both, if you can. If your export says
campaignand the site saysutm_campaign, rename the column in the export before you upload it. - Say how they relate, once, at the start of the conversation — for example: “listings-search in the spend export is the same campaign as utm_campaign on the site”. This covers values that are spelled differently in the two places, which renaming a column can’t fix.
- Pick both datasets in the datasets pill before you ask. A conversation only reads the datasets it was given, so one that isn’t ticked is invisible to it.
- Ask your question as if it were one table — “what did each sign-up cost, by campaign?”
How to tell it worked
The answer puts figures from both datasets side by side — spend from one, sign-ups from the other. How this was counted under the answer shows the rules it used, so you can check the two were matched the way your team would match them. If a campaign is missing, it’s usually spelled differently in the two places: tell Crunch they’re the same and ask again.
An example
Two ad campaigns got the same budget, and the team suspects the cheap clicks aren’t the good ones.
| You ask | Crunch finds |
|---|---|
| We spent the same on our Google and Meta campaigns. How many sign-ups did each bring in, and what did each sign-up cost? | $12,000 on each. Google brought 960 sign-ups at $12.50 each; Meta brought 480 at $25.00 each. |
That answer joins the ad platforms’ spend export to the site’s sign-up events on the campaign name. Neither dataset could give it alone: one has the spend, the other has the sign-ups. The full setup is in See which ad spend comes back.
Good to know
- Once a conversation has read its data, its datasets are fixed for that conversation, so answers don’t shift under you. To add another dataset, start a new conversation.
- Hashed emails line up people across sources without the email itself ever being sent. See Hashing user IDs.
Tip: when two datasets cover different date ranges, say which window you want. Otherwise “last month” might mean a month one of them doesn’t have.
Related
- See which ad spend comes back — this example, set up from scratch.
- Datasets — picking the datasets a conversation reads.
- Hashing user IDs — line up your site and your database on a hashed email.