A weekly ops report that checks its numbers first

A dashboard shows every number and waits for someone to look. This report picks the 5 to 10 numbers that moved past their threshold, says by how much, and points to the records behind the change. It arrives where leadership already reads, at the same time every week.

The two work well together. If you already run Looker Studio, Power BI or Metabase and people use it, keep it. The report links to the relevant dashboard view under each point, so a reader who wants more detail is one click away. If you have no dashboards, the snapshot store this build creates is a clean base for one later.

What the report adds is the sentence a COO would otherwise write by hand on Monday morning. An illustrative example of the output, with made-up figures:

New pipeline was $182,000, down from $240,000 the week before. Most of the gap is one deal: a $50,000 renewal moved to next quarter at the customer's request (HubSpot note, October 22). Excluding it, pipeline was flat. Support first response time rose from 2.1 to 3.4 hours; 60% of new tickets were tagged billing, following the invoice template change on October 20.

Every figure in that paragraph comes from a query, and every cause comes from a record the model was handed. That split is what keeps AI hallucination out of the numbers leadership reads.

The build is a fixed price from $8,000 to $20,000. The low end covers 4 sources with standard APIs, one metric catalog of up to 25 metrics, one report to one channel, and the checks described above. Each extra source adds $1,500 to $4,000; the range depends on how clean its API is. A second version for another audience, such as a board summary, is priced as an addition.

Running costs are small and billed at cost with no markup:

Upkeep is the line to budget for. An automation care plan from $1,500 a month covers monitoring, fixes and small changes, and usually covers several automations at once. To compare this against the hours it saves, use the method in how to measure AI automation ROI.

When fewer than 3 systems are involved, or when the company is small enough that the founder already knows the numbers. Under about 10 staff, the report tells people what they heard in the hallway.

It is also the wrong build in these cases:

• The numbers are disputed at the source. If finance and sales disagree on what counts as revenue, automation spreads the disagreement faster. Agree the definitions first; the metric catalog is a useful forcing document even without the build.
• Nobody needs the narrative. If leadership only wants the figures, a scheduled email with a table, built in n8n or Make with no language model, does the job for less. We will build that instead, because it is cheaper to run.
• You already have a data team with a warehouse and dbt. Then the pulls, metrics and checks, including dbt's source freshness (opens in a new tab) checks, exist. Only the driver queries and summary writer are missing, which is a single-workflow AI automation build from $4,000.

The source tools also need usable APIs. Most finance, CRM, support and project tools have them, but some plans restrict API access, and older on-premise systems may need an export step. We check every source on the scoping call, before the fixed price. For the connector work itself, see workflow integrations and n8n vs Make vs Zapier.