Skip to main content

Documentation Index

Fetch the complete documentation index at: https://fromhazel.ai/docs/llms.txt

Use this file to discover all available pages before exploring further.

​
General

Not at all. Hazel is designed for non-technical users. Ask questions in plain English, just like you’d ask a colleague. No spreadsheets, no formulas, no query language required. Learn more about chatting with Hazel.
ChatGPT and Claude are general-purpose AI tools. Hazel is purpose-built for consumer brands — it’s already connected to your data, understands ecommerce metrics natively, and remembers your brand between conversations. You don’t need to upload CSVs or explain what ROAS means. See how Memory works.
Hazel is built on Claude by Anthropic, one of the most capable AI models available.
Yes. Hazel uses enterprise-grade encryption in transit and at rest. We never use your data to train AI models. Your data stays yours — isolated per organization. Read more about our security practices.

​
Data & integrations

Every 4 hours. When you first connect a source, Hazel backfills your historical data — this can take a few hours to a few days depending on the source’s rate limits and data volume. After connecting, Hazel handles all data cleaning and setup — no manual configuration needed. Browse available integrations.
Hazel pulls as much history as each platform provides. For Shopify, that’s typically your entire order history. For ad platforms, it varies but usually goes back 1–2 years.
Yes. Each store connects separately and you can query across all of them. Learn how to connect a source.
Reach out via your shared Slack channel (current customers) or contact us at support@fromhazel.ai. We add new integrations regularly and prioritize based on customer requests.
There are a few reasons why Hazel’s numbers might differ slightly from what you see in your native dashboards:
  • Identity resolution — Hazel links customers across platforms using email, phone, and other identifiers. This means metrics like new vs returning customer revenue may differ from what Shopify reports, because Hazel can recognize a “new” Shopify customer who previously purchased under a different email or through a different channel. Hazel’s numbers are often more accurate because of this.
  • Google Analytics sampling — GA4 uses sampling to estimate values, especially for larger date ranges. The numbers displayed in your GA dashboard are estimates, not exact counts. Hazel uses the same sampled data from GA’s API.
  • Sync timing — Hazel refreshes data every 4 hours. Depending on when you ask, there may be a few hours of lag compared to a real-time dashboard.
  • Data transformations — Hazel normalizes and cleans data across sources (e.g., handling refunds, cancellations, currency conversions). This can produce slightly different totals than raw platform data.
If you notice a significant discrepancy, reach out to your Hazel team — we’re always improving data accuracy and want to investigate any gaps.
Yes. You can upload CSV files directly in the chat by clicking the attachment button. Hazel will analyze the data alongside your connected sources. There are file size limits, so for very large files, reach out to your Hazel team and we can help process the data.Excel (.xlsx) and PDF files are not currently supported for direct upload — export to CSV first.
It depends on the platform and your data volume. Most integrations finish within a few hours. High-volume Shopify stores or platforms with years of event-level data (like Klaviyo) can take 1–3 days. You’ll get an email when your data is ready. You can start asking questions about partially synced data while the full backfill completes.

​
Features

Citations are available in Deep Thinking mode. They show a simple equation of the calculation and the source(s) used — so you can verify any answer and build trust in the analysis. Learn how citations work.
Ask gives you answers in seconds and handles most questions — including ones that pull from multiple data sources. Deep Thinking is for when you want Hazel to spend an hour or more doing thorough, all-angles research on a big strategic question. Learn more about Ask vs Deep Thinking.
Yes. You can share any thread, and scheduled reports will be able to be delivered to Slack channels or email distribution lists.
Yes. Deep Thinking runs on Hazel’s servers, not your browser. You can close your laptop, switch tabs, or go offline entirely. Hazel will email you when the analysis is complete, and you can review the full results anytime by opening the thread.
This means your thread has reached the conversation length limit. Start a new thread for your next question — Hazel performs best when threads stay focused on a single topic. If you frequently hit this limit, try breaking your work into separate threads by topic (e.g., one for marketing analysis, one for subscription churn).
There are two types. Org Memory is shared across everyone in your organization — when one person adds a definition, everyone benefits. User Memory is personal to you and only affects your conversations. Learn more about Memory.

​
Account & billing

Hazel charges a flat annual fee with no per-seat or per-connection charges. Contact us for pricing details.
Yes, at no extra cost. Hazel’s flat pricing means you can invite your whole team without worrying about per-seat fees. Learn about team management.
Book an intro call to learn about trial options.