Pravin DurganiSay hi

Free checklist

Is your reporting pipeline ready to automate?

Twelve questions. If you tick four or more as problems, automation is the cheaper option.

Most reporting pipelines aren't broken because the tools are wrong. They're broken because nobody has written down what they actually do. Run through these twelve before you buy anything.

Pravin DurganiUpdated June 2026pravindurgani.com/checklist

Data flow basics

Where the numbers come from, and what happens between source and send.

  1. Same export, same shape, every week?

    Healthy: one named source, identical columns, predictable timing — you could schedule it without thinking.

  2. Can you point to the file or dashboard the report comes from?

    Healthy: one link, one owner, no "it depends which version Sarah sent last".

  3. Does anyone manually copy, paste, or clean data before the report goes out?

    Healthy: zero manual steps between source and final output — or the steps are scripted and logged.

  4. How long does the report take from data-grab to send?

    Healthy: under an hour, end-to-end, including QA — and the hour is mostly waiting, not doing.

People & process

Who owns the work, and what happens when they're not at their desk.

  1. One person owns it, or do three people fight over it?

    Healthy: one named owner, one named backup, and everyone else knows to ask them.

  2. What happens when that person is on holiday?

    Healthy: someone else runs it from a checklist, nothing slips, no panicked Slack messages.

  3. Does the report ever go out wrong — and how do you catch it?

    Healthy: a sanity check fires before send, errors are logged, and someone reviews the log.

  4. Is the same insight rewritten every period, or does it stay live?

    Healthy: recurring observations are templated; the writing time is spent on what actually changed.

Risk & compliance

What happens when someone asks where a number came from.

  1. Do you send raw client or customer data to ChatGPT or other cloud LLMs?

    Healthy: sensitive data stays on a controlled environment, and you can name which model touched what.

  2. Where does the report live after it's sent — inbox, drive, board pack?

    Healthy: one canonical location, versioned, searchable, and someone owns the retention policy.

  3. If a regulator asked "how was that number produced", how long to answer?

    Healthy: a few minutes — you can trace any number on the report back to its source query and timestamp.

  4. Is there a written runbook, or is it tribal knowledge?

    Healthy: a runbook a new hire could follow on day one, kept up-to-date alongside the code.

If this hit a nerve —

Book a free 20-min fit call.

No deck, no pitch. We talk about which of the twelve are the painful ones, and whether automation is actually the right answer. By the end you know whether I'm the right fit — or who is.

Pravin Durgani · London · pravindurgani.com