Tines Academy › Lesson

Keep the data fresh

Add weekday updates, show when data is missing or out of date, and learn how to investigate a failed refresh.

Objective: Define how a dashboard handles loading, missing data, and failed updates, and use run information to investigate problems.

Say what’s missing

You’ve loaded the feedback and built the dashboard. Now you’ll decide what people should see while data is loading, when no responses match their filters, or when an update fails.

A week with responses but no negative ratings should show 0%. A week with no responses shouldn’t report a percentage. Make the difference clear so Maya’s team doesn’t mistake missing data for real results.

Make the state of the data clear

Say how fresh the data is

Maya’s team leads use the follow-up list throughout the week, so the dashboard will load new survey responses every weekday at 7:00 UTC. Each update picks up responses rated since the last successful refresh. If an update fails, the next successful one should fill the gap.

Until then, keep the last good data visible, show when it was refreshed, and warn that it’s out of date. Mark any affected weeks as possibly incomplete until the missing responses have loaded.

The current week is incomplete for a different reason: more feedback is still arriving. Show it separately in a “This week so far” panel, leaving the weekly chart for completed weeks.

Scheduled updates run only when the workflow is live and active. Schedules use UTC, so 7:00 UTC may be a different time where you are. Tell Tines 3B if you want to use a different time. You’ll test the update manually on the draft in this lesson.

Add the update and the missing-data cases

Choose the tab that matches your setup, then send the prompt in workflow chat.

Add an update that runs every weekday at 7:00 UTC. Load survey responses rated since the last successful update. If there’s no saved data, load the last 12 full weeks and the current week so far. Apply the existing preparation rules to new responses.

If an update fails, keep the last good data and show an out-of-date warning. Display the time of the last successful refresh. Mark affected weeks as possibly incomplete until the missing responses have loaded.

Add a "This week so far" panel showing the current week's negative-rating percentage and both counts. Follow the existing product area and channel filters, but not the full-weeks period filter. Make clear that this week is still in progress.

Show clear messages while data loads, when no feedback is available, and when data can’t be read. Don’t use zero to represent missing data.

Add a draft-only test switch for no data, unreadable data, slow loading, and a failed update. Label it "Test only" and include a way to turn the simulations off.

Use temporary simulations without deleting or changing saved data, source records, or shared connectors. Hide the switch on live and prevent the simulations from being triggered there, even without the switch.

Test the update and simulated states where possible. Check that a failed update preserves the last good data and that a later successful update fills the gap without counting responses twice.

Report what you tested, what passed or failed, and anything you couldn't verify. Identify the step I should run to test an update and explain how to use the test switch.

Keep the preparation rules, access restrictions, and existing filter behavior unchanged. Leave the workflow as a draft.

Add an update that runs every weekday at 7:00 UTC and generates a few made-up responses dated today. After a failed update, also generate responses for the days missed since the last successful update. Apply the existing preparation rules to all new responses.

If an update fails, keep the last good data and show an out-of-date warning. Display the time of the last successful refresh. Mark affected weeks as possibly incomplete until the missing responses have been added.

Add a "This week so far" panel showing the current week's negative-rating percentage and both counts. Follow the existing product area and channel filters, but not the full-weeks period filter. Make clear that this week is still in progress.

Show clear messages while data loads, when no feedback is available, and when data can’t be read. Don’t use zero to represent missing data.

Add a draft-only test switch for no data, unreadable data, slow loading, and a failed update. Label it "Test only" and include a way to turn the simulations off.

Use temporary simulations without deleting or changing saved data, source records, or shared connectors. Hide the switch on live and prevent the simulations from being triggered there, even without the switch.

Test the update and simulated states where possible. Check that a failed update preserves the last good data and that a later successful update fills the gap without counting responses twice.

Report what you tested, what passed or failed, and anything you couldn't verify. Identify the step I should run to test an update and explain how to use the test switch.

Keep the preparation rules, access restrictions, and existing filter behavior unchanged. Leave the workflow as a draft. Keep the "Test data" label visible and don't invent ticket links.

Test the update

Read the build summary and test results in workflow chat. Find the update step and the instructions for the test switch. Note anything Tines 3B couldn’t verify.

Run an update

Run the update step identified in the build summary. Ask Tines 3B to show you where to run it if you’re unsure.

Once it finishes, open Browser. Check that the last successful refresh time has changed and the dashboard still shows the expected data.

A successful update may find no new responses. The totals don’t have to increase.

Check the current week

Find “This week so far.”

It should show the current week’s percentage and counts, with a clear note that the week is still in progress.

Try a product area or channel filter. The panel should follow it. Changing the full-weeks period should leave this panel’s dates unchanged.

Try a failed update

Use the test switch to simulate a failed update.

Check that:

The last good data stays visible.

An out-of-date warning appears.

The last successful refresh time stays unchanged.

Any affected weeks are marked as possibly incomplete.

Review the other messages

Check the reported results for no data, unreadable data, and slow loading. Use the test switch to try any that weren’t verified.

The page should explain what’s happening. Missing data shouldn’t appear as 0%.

Turn the simulations off

Return the test switch to normal and run another update.

Check that the warning clears after a successful update. Review the test report for evidence that missed responses were added without counting any twice. If that wasn’t tested, leave the check marked as unverified.

If something doesn’t match

Tell Tines 3B which test you tried, what you expected, and what you saw. Ask for a focused fix, then repeat the affected check.

Turn off any remaining simulations and ask it to record unfinished checks in Readme.

Find out why an update failed

If the dashboard warns that its data is out of date, start with the update run.

Open the workflow’s Monitor tab and select the latest update run.

If it failed, open the failed step’s details and read the error.

Use Investigate in chat to ask Tines 3B for help. Check what happened before changing anything.

For a connector problem, open Connectors from Home and check your ticket system connector’s status. If it shows “Error,” read the details and ask whoever manages the connector for help.

What to do if something still feels fuzzy

You added an update and tested it on the draft using your ticket system or made-up data. You also simulated a failed update and checked how the dashboard warned people. The weekday schedule will run once the workflow is live and active. Keep unfinished checks in Readme for the next lesson.

Build it safely: You checked that a failed update preserved the last good data and showed when it was last refreshed, so people could recognize that the numbers might be out of date.