Skip to content
⌘K
InboxAsk a human

Troubleshooting Export Failures

Updated by Avi

Widget Exports are deprecated. Existing widget exports keep running for now, but don't set up new ones: use Export Pipelines instead. To move an existing widget export, see Migrating from Legacy Exports to Export Pipelines.

This article helps you find out why an export did not arrive, arrived incomplete, or looks different from what you expected. It covers Export Pipelines and the legacy Data Exporters and Widget Exports.

Check the Run

For Export Pipelines, every run and its result is in Clarisights. There are no email or Slack alerts, so start here.

  1. Click Export pipelines in the left sidebar. If a pipeline needs attention, a banner above the pipeline list names it with one of these:

    • Last run failed: the most recent finished run failed. Click View runs to open its runs.

    • Destination deleted: a schedule sends files to a destination that was deleted. Click Fix schedule to pick a new one.

  2. Open the pipeline and go to the Runs tab. Each run shows its status (Running, Succeeded or Failed), the version, the schedule, when it started, the date range it covered, and the number of files and rows. The tab covers the last 30 days.

  3. Click a run to see its details: the destination and folder, and for each extract the number of rows, the file size and the file name.

  4. For a failed run, read the error message in the details, fix the cause, then click Retry with the same dates (or Retry). This opens Run now with the failed run's dates filled in.

The Runs tab of a pipeline with succeeded and failed runs
The details of one run, with the rows and file size of each extract

Note: A run with 0 rows succeeded, but no data matched its date range. The file is still written, with only the header row.

Quick Checks

  1. Is it time yet? Check the schedule's time and timezone, and allow time for the export to finish. See Export Scheduling & Timing Guide.

  2. Are you looking in the right folder? With the default file name template, Export Pipeline files are stored under the destination's path prefix, then the pipeline name, the version (v1, v2, ...) and the extract ID. A new version writes to a new folder. The run details on the Runs tab show the exact folder.

  3. Did the bucket permissions change? A changed bucket policy, IAM role, trust policy, service account or storage account key is the most common cause of failed exports. Run Test connection on the destination to check (see Access problems below).

  4. Check the metadata file. Each Export Pipeline data file has a .meta.json file next to it with the exported date range and the list of columns.

Common Problems

Access problems: use Test connection

When runs fail with an access, sign-in or permission error, test the destination:

  1. Click Export pipelines in the left sidebar and open the Destinations tab.

  2. Click Manage on the destination, then Test connection. It checks that Clarisights can read, write and overwrite files there.

  3. If the check fails, its title tells you what to fix, for example Access denied, Sign-in failed, Couldn’t assume the role, Impersonation failed, Bucket not found or Wrong region. For Amazon S3, the check can also ask you to run a command in your AWS account and paste its output to confirm write access.

  4. Fix the permissions, test again, then retry the failed run from the Runs tab.

Every check and its fix is in Configuring Export Pipeline Destinations & Permissions. The common causes per provider are below.

1. Amazon S3: Access Denied

Symptom: No files arrive in your S3 bucket, and the run fails with an access or "assume role" error.

Common causes:

  • The role or user is missing s3:PutObject or s3:PutObjectAcl on the bucket. Both are needed.

  • The trust policy of your IAM role no longer lets the Clarisights AWS account arn:aws:iam::747179778889:root assume it.

  • The External ID in the destination does not match the one in your trust policy.

  • The Role ARN in the destination is wrong. The format must be arn:aws:iam::ACCOUNT_ID:role/ROLE_NAME.

  • A bucket policy denies the upload, even though the IAM permissions are correct.

  • The bucket is encrypted with your own KMS key, and the role or user cannot use that key (kms:GenerateDataKey).

What to do: Fix the policy, then run Test connection on the destination and retry the failed run.

2. Google Cloud Storage: Permission Denied

Symptom: No files arrive in your GCS bucket, and the run fails with a 403 or permission error.

Common causes:

  • The service account Clarisights writes as (the Clarisights service account, or your own service account with Impersonation or a JSON key) does not have the Storage Object User role on the bucket, or another role with storage.objects.get, storage.objects.list, storage.objects.create and storage.objects.delete.

  • When impersonation is used, the Clarisights service account is missing the Service Account Token Creator role on your service account.

  • The role only allows creating files (for example Storage Object Creator). Clarisights also needs storage.objects.delete to overwrite a file, and the connection test checks it.

3. The export has fewer rows than expected

  • Widget exports contain at most 100,000 rows. Rows beyond that are left out without a warning. If your file has exactly 100,000 data rows, narrow the date range, remove dimensions, add filters, or move to an Export Pipeline, which has no row limit.

  • Data Exporters can split a large file into several parts (.gz.1, .gz.2, and so on). Make sure you read every part.

  • Export Pipelines deliver one file per extract, with no row limit. Compare the row count on the Runs tab with your file, and check the extract's filters, the pipeline's data sources, and the date range in the .meta.json file.

4. Numbers don't match what I see in Clarisights

  • Rounding: Exports contain unrounded values, while Clarisights rounds numbers for display.

  • Date range and timezone: Compare the same dates in the same timezone. Export Pipelines use the schedule's timezone, which can differ from the one you use in your reports.

  • Late data: Channels keep updating recent days. An export that ran early in the morning may not include the latest updates. Re-export the last few days (for example with a date range of the last 3 or 7 days) to pick them up.

  • Grouping: An extract is grouped by exactly the dimensions you chose. Compare it with a widget that uses the same dimensions, filters, and data sources.

5. The file doesn't match the preview

The preview in the extract editor is a quick check, not a copy of the file:

  • Dates: The preview's date options use your company timezone. A schedule uses its own timezone, so the same preset can cover different days.

  • Currency: If you set a preferred currency for yourself, the preview shows amounts in that currency. The file doesn't follow personal currency settings.

  • Columns and values: The preview shows metric names and formatted values. In the file, metric columns use metric keys (the .meta.json file maps them to names), and values are unrounded.

6. Duplicate files, or a new folder appeared

When a new version of an Export Pipeline is published, the previous version keeps exporting for 30 days. During those 30 days, every run sends two sets of files, one per version. With the default file name template, each version writes to its own v<N> folder. If your template has no {manifest_version}, files from both versions land in the same folder and can overwrite each other.

Update your downstream processing to read the new version before the old one stops. The Versions tab shows how many days the old version has left. Legacy Data Exporters work the same way with their v<version> folder.

7. Runs don't start

If the Runs tab shows no run at the expected time:

  • The pipeline is paused: the pipeline page says so, and no schedule runs. Resume it from the ⋯ menu.

  • The schedule is paused: its status on the Schedules tab is Paused. Click Resume.

  • The schedule is waiting: its status is Waiting because the pipeline was never published. Publish the draft.

  • The destination was deleted: the schedule shows Destination deleted, and runs are skipped without a record. Edit the schedule and pick another destination.

  • A data source was disconnected: check that the pipeline's data sources are still connected in Clarisights.

  • A custom metric or custom dimension changed: if one that an extract uses was renamed or deleted, start a new draft, edit the extract (fields that are no longer available show in amber), pick the current field, save it to the draft and publish. Publishing starts a new version, so the 30-day overlap above applies.

Contacting Us

If you can't find the cause, contact support (click ? at the bottom right of Clarisights and choose Send us a message, or email support@clarisights.com) with:

  • The pipeline name and the run's start time (or the widget or channel, for legacy exports)

  • The error message shown in the run details