Before you connect your store or upload a customer export, you may need to answer a simple question for your team or your own customers: where does this data actually go? This page answers it one kind of data at a time — what BumbleB Crunch™ stores, what it reads without copying, and what is copied when you share something. The binding versions are our Privacy Policy, Data Processing Agreement and sub-processor list.
Where it’s hosted
Everything BumbleB Crunch keeps for you — your events, files, conversations and what it makes from them — is stored in the United States: on Microsoft Azure (East US), with databases on MongoDB Atlas, which runs on Azure in East US 2 (Virginia).
A few other companies help run the service. A company that handles data for us like this is called a sub-processor, and each one is bound by written data-protection terms:
- OpenAI, in the United States, currently writes BumbleB Crunch’s answers. It receives your question and the context needed to answer it, for that and, where our sub-processor list says so, search embeddings — nothing else. Each provider’s retention is on that list; OpenAI currently keeps no prompts or answers (apart from brief retention the law may require), and our contract bars it from training on your data.
- Amazon Web Services, in Mumbai, sends our account, billing and security emails. It never receives your data, only the email address and the notice.
- Razorpay (payments) and Google (Sign in with Google, and our own email) handle account details only, never your data.
We give 30 days’ notice before adding or replacing a sub-processor. You can subscribe to changes on the sub-processor list.
Website and store events
Your clickstream — the events your website tag or Shopify pixel sends as visitors browse — goes straight to BumbleB Technologies’ own servers on Azure. The tag is served from there too, and no other company is involved. If you add the tag through your own Google Tag Manager account, that part runs under your own agreement with Google.
Each verified website’s events are kept separately, so you can delete one site’s visitor data without touching the others. Events are kept while your agreement runs, or until you delete them.
Files you upload
An uploaded file becomes a dataset — a table in your own workspace. Each fresh upload of the same file is kept as a version, so a conversation keeps reading the one it started with. Uploads are read only to answer your questions, and kept while your agreement runs, or until you delete them.
Your own data lake
When you link tables from your Postgres database or Iceberg catalog, the rows stay in your lake. BumbleB Crunch reads them where they are, with the sign-in you give it (use a read-only one), and nothing is copied in first. The tables a conversation works out from them — the results behind a chart, say — are kept in your workspace, like every answer’s.
Opening a linked table as a sheet still reads it in your lake. The one thing that copies it is publishing the sheet: the table, at the version the sheet was reading, is copied for your readers, so they never reach your lake or its sign-in details. See Open a table from your data lake.
A database behind your firewall
The on-prem sync connector runs in your own network and applies your rules to every column before anything leaves. Only what your rules let out is sent, over an outbound connection, and that is what’s stored in your workspace. A column without a rule never leaves. See Keep sensitive data inside your network.
Your conversations
Your questions and BumbleB Crunch’s answers are kept with your account, so you can come back to them. To write an answer, BumbleB Crunch sends your question and the data it needs to OpenAI, as above.
BumbleB Technologies staff don’t read your conversations to improve BumbleB Crunch. They look at one only when you ask for support, to investigate security or abuse, or when you send it to us yourself. Automated checks may look over recent conversations to label problems, without anyone reading them, and they skip data about your website’s visitors. If you use BumbleB Crunch on your own in India, you’re asked first, with a box that starts unticked.
How long conversations are kept is set out in our Privacy Policy.
If you ask BumbleB Crunch from an AI assistant such as Claude, your questions and BumbleB Crunch’s answers, including tables of your data, also go to that assistant’s provider, under your own agreement with them. See Use BumbleB Crunch from Claude.
What you publish
When you publish a story, a dashboard or a sheet, your readers see a copy, not your workspace:
- A story or dashboard shows its text, figures and the rows behind each chart, and nothing else of your data. Before you publish, What readers will see lists that data and highlights columns that look personal. Unpublish stops sharing at once. See Publish and share a data story.
- A sheet copies each of its tables, as they stand when you click Publish and share.
- A link anyone can open shows a frozen copy of what you had published when you made the link. It doesn’t refresh by itself.
Good to know
- No training on your data. Neither BumbleB Technologies nor any provider we use trains AI models on your data. The one exception is a conversation you choose to send with a bug report.
- We keep a record of how BumbleB Crunch ran — which tools it used, in what order, whether they worked — with no names, account details or values from your data. Because it identifies no one, it isn’t deleted with your account. See How BumbleB Technologies uses your data.
- We don’t currently host data in India. If a law or rule requires certain personal data to stay in India, don’t bring that data to BumbleB Crunch.
Related
- What we collect — exactly what each connection sends.
- Keep sensitive data inside your network — ways to keep raw personal data on your side.
- Read your own data lake — link tables with no copy.