Let’s talk
GUIDE

How do you build a report that updates itself when the data changes?

Keep the data in one place that people already update, write the report as a set of rules that reads that data, and run those rules on a schedule or every time the data changes. Add checks that stop the report when an input is missing or does not add up, so it never shows a wrong number quietly. Nobody edits the report by hand; they edit the data.

The hard part is not the chart. It is agreeing what each number means, where it comes from, and what should happen when a source is late or wrong.

Updated Sep 28, 2026

How it works

  1. Pick the source of truth

    One spreadsheet, database or system export per number. If the same figure lives in two places, decide which one wins before building anything.

  2. Define each metric in writing

    Name it, write the formula and the date range, and say which source feeds it. This is what makes the report the same next month.

  3. Choose when it rebuilds

    A schedule (every morning, every Monday) or an event (a new row, a new file). Event-driven reports are fresher; scheduled ones are easier to predict and to check.

  4. Add checks that can stop it

    Missing file, empty period, totals that do not match the source, a date outside the range. When a check fails, send the old report with a warning, or nothing, rather than a wrong one.

  5. Deliver it where people already look

    A dashboard, an email or a single message. A short line people read beats a full dashboard nobody opens.

What it costs to run

The main costs are the dashboard tool and whatever runs the schedule. The list below is list prices read today; the systems the data comes from have their own plans.

ItemList priceSource
Google Data Studio (previously Looker Studio), self-service reportsNo chargeGoogle Cloud, Data Studio: Business Insights Visualizations (pricing section)Read on Sep 28, 2026
Google Data Studio Pro (project subscription)US$9 per user per project per monthGoogle Cloud, Data Studio: Business Insights Visualizations (pricing section)Read on Sep 28, 2026
n8n Cloud Starter, to run the rebuild on a scheduleUS$20 a month billed annually, 2,500 workflow executions. n8n notes that a daily schedule uses 30 or 31 executions a monthn8n, n8n Plans and PricingRead on Sep 28, 2026

Third-party list prices, read on the date shown. They are not Betterlane prices.

What we learned building it

From Betterlane's own knowledge base and its own lead handling, both running in September 2026. Not a client system.

The page is a view; the files are the truth
Our knowledge base page is regenerated from plain markdown files by one script. Nobody edits the page itself, so it cannot drift from the files.
The build refuses to publish when something does not add up
It fails, and writes nothing, when an entry is missing a field, cites a source that does not exist, links to an entry that does not exist, names a file that is not there, or keeps a source that nothing cites. The last good page stays up.
Every claim carries a confidence grade
Each entry ends its evidence with a grade, A, B or C, and the build rejects one without it. A reader can see how sure a line is, not only what it says.
One line a day beats a dashboard
Our lead-handling workflow sends a single line at 20:00 Dubai time: leads, answered, and the median time to the first answer. It is short enough to read every evening.

When it is not worth it

  • If the report is read once a quarter and takes an hour to prepare, keep preparing it by hand.
  • If nobody agrees yet on what the numbers mean, automating the report only repeats the disagreement faster.
  • If the source data is typed in irregularly, fix how it is captured first; a report cannot be fresher than its inputs.

Questions

Schedule or real time?

Most small businesses need a schedule: daily or weekly. Real time is worth it only when someone acts on the number within the hour.

Can I keep using spreadsheets?

Yes. A spreadsheet is a fine source of truth if one person owns it and the columns do not change without notice.

What happens when a source is late?

The report should say so. A check that stops the rebuild, or marks the gap, is better than a number built on half the data.

Should a model write the summary?

It can draft the words, but the figures should come from the calculation, and the summary should keep calculated numbers apart from interpretation.

Want a report that rebuilds itself from the data you already keep?

See the service: Automated reportingLet’s talk

All guides