Skip to content
Sprigr

Sprigr Analytics

Sprigr Analytics gives agents one tool, analytics_query, that runs a read-only SQL statement over your company’s conformed analytical store, the joined, cleaned tables built from your connected data such as orders, order line items, customers, products and variants. Every query runs as the person asking: the platform resolves their analytics role, binds the query to the build delivered for that role, runs it there, and audits the call. That is what lets a store manager, a sales rep and an executive ask the same agent and each see only what their role permits.

Prerequisites

Section titled “Prerequisites”
  • A conformed analytical store for your company, with analytics roles assigned to the people who will ask questions.
  • Admin or Owner role in your Sprigr Team organisation to install the app.

Installing

Section titled “Installing”

  1. Install the app

    Sign in to team.sprigr.com, open Apps in the sidebar, find Sprigr Analytics under Available, and install it. No secrets are requested. See Installing Apps.

  2. Check your access

    Ask your agent “What analytics tables can I query?” It reads the store’s schema for your role: every table in your build with its columns and types, row counts, the partition field and the months held.

  3. Ask a question

    Ask in plain language. The agent reads the schema first, writes one read-only SQL statement, names a month range when the question has one, and reports the rows with the tables and months it touched.

How queries are scoped

Section titled “How queries are scoped”
  • Read-only, one statement. Only a SELECT or WITH statement is accepted; anything else is rejected before it runs.
  • Role-bound. The role comes from the identity stamped on the call, never from the request, and the query runs against the physical build delivered for that role. A table outside your role’s build is refused with the list of tables you can use.
  • Month ranges keep it fast. Partitioned tables such as orders and line items read only the months named; unpartitioned tables such as customers and products are read whole. A query that names too many months or tables is refused with a budget message rather than running slowly.
  • Row caps. Results return up to 200 rows by default (2,000 at most); a truncated result is a prompt to aggregate or filter rather than page.
  • Audited. Every call is recorded with the role and build it ran against.

Common use cases

Section titled “Common use cases”
  • Sales questions without a dashboard. Revenue, orders and average order value by month, channel or product, straight from chat.
  • Role-appropriate answers. A store manager asks about their store and gets their store; an executive asks the same question and gets the company.
  • Cross-source joins. Orders joined to products and customers in one query, with the conformed keys already in place.
  • Scheduled briefs. A weekly workflow, running as a nominated person, posts the numbers that person is allowed to see.

Troubleshooting

Section titled “Troubleshooting”

“No assignment” Your company scopes analytics by role and you do not hold one. Ask an admin to assign you an analytics role.

“No delivered build” Your role’s build has not been delivered yet. Sprigr delivers builds as part of setting up the store.

“Table not in slice” The table is not in your role’s build. The refusal lists the tables you can use.

“Budget exceeded” Narrow the month range or name fewer tables. Aggregate earlier in the query for large joins; there is a 45-second limit.

A column does not exist Ask the agent to read the schema again rather than guess column names; types matter too (timestamps and numeric money columns).

Next steps

Section titled “Next steps”