Skip to content

Data Health: Check That Your Cost Data Is Complete and Current

Every number in Xplorr rests on cost data arriving completely and on time. Data Health checks that assumption for each cloud account, so you find out that a bill is missing three days before you present it to someone.

Open Data Health from the sidebar. The page is read only and anyone can see it, scoped to the accounts they have access to.

Check What it looks for
Empty syncs The sync ran successfully but wrote no cost records at all
Stale data The newest cost data is older than the staleness threshold, 3 days by default
Missing days Days with no cost data inside the last 35 days
Dropped totals A day whose total fell by more than 20 percent compared with the previous sync
Provider bill Every night, whether the account’s total matches the provider’s own bill (see Does it match your bill?)

Two deliberate details that prevent false alarms:

  • Missing day detection stops a couple of days short of today, because provider billing lands a day or two late. A gap yesterday is normal, not a fault.
  • Gaps before an account’s first data date are not counted, so connecting an account today does not report 35 missing days.

Dropped totals catch the failure mode people miss most: a sync that succeeds but returns less than it did last time. A day that disappears entirely counts as a 100 percent drop. Days whose previous total was negligible are ignored as rounding noise.

Each account gets one status, in order of severity.

Status Meaning
Never synced No sync has run yet
Failing The last sync ended with an error
No data Syncs return no cost data
Stale Newest data is older than the threshold
Warning Gaps, drops, an empty sync in recent history, a total that does not match the provider bill, or cost that may be counted in two connected accounts
Healthy Current, complete, no drops

There is no numeric health score. An account is in exactly one of these states, and the summary counts how many need attention.

Accounts marked inactive are left out entirely.

Expand any row to see:

  • Last sync: result, when it finished, how many records it wrote, and whether the source was a FOCUS export or the provider cost API

  • Missing days in the last 35, listed as dates

  • Daily total drops, with the size of each drop

  • What to check, a plain list of the specific problems found

  • Provider bill, the result of the last nightly bill check (below)

  • Notes about the account, such as an AWS management account’s linked accounts

The “what to check” list is the useful part. It says things like how many days are missing, how far behind the newest data is against the threshold, or the exact error the last sync returned, rather than leaving you to infer it.

A Refresh button re-runs the checks on demand. The checks are computed when you ask for them, so there is nothing to schedule and no cached result to invalidate.

Every night at 03:30 UTC, after the 02:00 UTC cost sync, Xplorr compares each active AWS, Azure and GCP account’s stored total with the provider’s own total for the same scope and days. It checks two periods: the month to date (through the last settled day) and the previous month.

Each account row shows a badge, and the expanded row gives the detail:

Badge Expanded row says Meaning
Matches bill Matches your AWS bill (or Azure, or Google Cloud), with the date checked Both totals agree within 2 percent or 1 USD. Each period shows the Xplorr total and the bill total.
Bill mismatch Does not match your AWS bill The totals differ by more. Each period shows both totals and the gap, in money and percent. The account’s status becomes Warning.
Bill not checked Could not check against your AWS bill, with the reason The provider refused the request or returned an error, for example a missing permission.

What is compared, per provider:

Provider Xplorr total Provider total
AWS Every stored charge for the account, in its billing currency Cost Explorer GetCostAndUsage, daily, unblended cost, not grouped. A management account leaves out the linked accounts that are connected separately (see Connect AWS); a member account is filtered to itself.
Azure The same One Cost Management query of actual cost on the subscription the sync reads
GCP The same One BigQuery query of the billing export the sync reads: cost plus credits

An account already checked today is skipped, so the result is at most a day old. Refresh re-runs the other checks but shows the last stored bill check. Xplorr staff are told about every new gap over 2 percent and look into it.

The check uses the permissions the cost sync already needs; it adds no new ones.

Xplorr also looks for connected accounts of the same provider whose cost overlaps: an account whose daily cost per service equals another connected account’s, or the sum of several, and AWS management accounts that include a linked account that is also connected. When it finds one, the expanded row says so, for example that the account’s cost is also included in a management account and may be counted twice, and the account’s status becomes Warning. Xplorr staff are told as well.

When the latest sync of an account failed, the top bar shows a sync error badge with the number of failing accounts. It counts only active accounts you can see, so a member is never shown an error on an account they cannot open.

Click the badge to see, for each failing account:

  • the account and its provider
  • what went wrong, in plain words
  • when the sync failed
  • when the account last synced successfully

Retry sync starts a new sync of that account. It needs the same permission as Sync now on the Cloud Accounts page, and the same limit applies: an AWS account syncs its costs at most once an hour, so a retry inside that hour is refused with a message saying when it can run. If a retry fails again, the common causes below cover the usual errors.

After you replace an account’s credentials, its old failure stops counting and the account shows as pending until the new sync finishes.

When an account is failing, empty or stale, a banner appears on other pages linking here. It lists at most three accounts and can be dismissed for the rest of your browser session.

Treat it as a warning about the numbers on the page you are looking at, not as a notification you can safely clear and forget.

Syncs do not run on the page that asks for them. A scheduled run, a Sync now, or connecting a new account all place a job on a queue, and a worker takes them one at a time per account while running several accounts at once.

When anything is queued, running or failed, a Sync queue panel appears above the accounts table. It disappears when the queue is empty, because a permanent “nothing wrong” panel is noise.

Failed syncs are listed with the account, the window the sync covered, how many attempts it made, and the provider error that ended it. Each has a Retry button, which puts the job back on the queue.

This is not the same information as the table below it. A sync that fails before it starts writes no sync record at all, so it can never appear in the table. That includes:

  • an account removed or deactivated between the job being queued and running
  • a worker that never picked the job up
  • a job still waiting its turn
  • two syncs of the same account colliding, where the second waits rather than being dropped

Those are exactly the failures that used to be invisible. If an account looks healthy in the table but its data is not moving, the queue panel is where to look.

Sync now appears on every account row. It queues a sync immediately and returns a job id, so you can watch it move through the panel rather than wondering whether anything happened. A manual sync takes priority over a scheduled one already waiting for the same window.

If the queue panel says the queue is unavailable, Xplorr could not reach the queue itself. That is not the same as “nothing is wrong”: the account checks below are still accurate, but nothing can be said about work in flight.

Failing

Almost always credentials or permissions. A rotated key, an expired secret, a deleted role, or a permission that was narrowed. The exact provider error is shown in the expanded row.

No data

The connection works but there is nothing to read. Common causes are a billing export that was never enabled, a BigQuery table that is empty, or an account with genuinely no spend.

Stale

The sync is not completing, or the provider has stopped publishing. Check whether the last sync succeeded; a stale account with successful syncs usually means the provider side is behind.

Missing days

Often transient, from a sync that failed on a particular day. Running a sync again for that period usually fills the gap, because re-syncing a day replaces it.

Bill mismatch

Look at which period differs and by how much. Late billing data usually fills in within two days. A gap that stays points to a partial export, or to cost counted in two connected accounts, such as an AWS management account and one of its members connected separately; the expanded row says when Xplorr found the second.

Dropped totals

Investigate rather than dismiss. A large drop can mean a provider restated a bill, an account was removed from an organization, or a sync only partly succeeded.

  1. Open the account from the expanded row and confirm the credentials still work.
  2. Fix whatever the error names, usually a permission.
  3. Run a sync for the affected period. Admins can trigger syncs, limited to 10 per hour.
  4. Come back and refresh to confirm the status cleared.

A newly connected account syncs within a few minutes and then on the regular daily schedule, so Never synced on a brand new account is expected for a short while.

Does this page change anything? No. It only reports. Syncs are triggered from Cloud Accounts.

Can I change the staleness threshold? It is a platform setting, 3 days by default, rather than a per organization option.

Why does it only look back 35 days? It is long enough to catch a real gap and short enough to stay fast. Older gaps are unlikely to be fixable from the provider anyway.

Who can see this? Everyone, scoped to the accounts they have access to.