Why it matters
Analytics are only as reliable as the data behind them.- Stale data misleads: A “last 7 days” analysis is useless if the most recent 3 days haven’t synced
- Proactive visibility prevents bad decisions: Catching a sync issue before a board meeting is better than discovering it afterward
- Scoping saves time: Knowing which domains are ready helps you focus on answerable questions
Data Health tells you whether the pipeline is working. Attribution Health tells you whether your tracking is capturing marketing touchpoints. Both matter. A table can be perfectly fresh but still show 40%
(direct) / (none) if UTM tracking isn’t set up properly.What we check
When data updates
Key tables refresh daily. Our SLA guarantees fresh data through the previous day based on your reporting timezone.- Through yesterday: Complete and reliable. This is what we guarantee.
- Today’s data: Incomplete. Do not use current-day data for analysis.
Real-time isn’t the goal. We optimize for accuracy over speed: data is correctly transformed, deduplicated, and enriched before it reaches your dashboard or warehouse.
Platform-specific timing
Some sources take longer than the previous day before their numbers settle.Why 14 days?
Key tables (orders, customers, sessions, and ad spend) should have fresh data every day. That is the normal operating state for an active e-commerce business. The 14-day threshold exists mainly for secondary tables that may not see daily activity:- A new subscription program might not have orders every day yet
- Refunds tables depend on actual refund volume
- Some niche integrations only fire on specific events
If a core table such as orders, customers, or ad performance is stale, that is almost always a pipeline issue. If a secondary table is stale, check whether you would expect activity before treating it as a broken sync.
Common scenarios
What to do if data is degraded
1
Check specific tables
Identify which tables are stale. Is it one platform or multiple?
2
Scope your analysis
Avoid date ranges that depend on stale data. If orders haven’t synced since Jan 15, don’t analyze Jan 16–20.
3
Check Attribution Health
If data is fresh but results look wrong (e.g., high
(direct) / (none)), the issue may be tracking, not pipeline.4
Escalate if persistent
If staleness persists beyond 24–48 hours, reach out to your SourceMedium team. There may be an integration issue requiring admin attention. See When to contact SourceMedium for what to include.
What the dictionary flags establish
The data dictionary publishes coarse table observations. Read each one for what it records:
Identify a dictionary row by dataset, table, column and its actual store scope. Table-level flags repeat on every column row of a table: read them once rather than adding them up, and treat disagreeing repeats as an inconsistency to investigate. A row for another store, or a global row, is not evidence about the selected store.
A dashboard with only a record date should say “Latest recorded order date: October 22. Pipeline freshness unavailable” rather than “All current”. A latest record date, a query completion time, the metric catalog’s
catalog_updated_at, one row per calendar day, and the 14-day flag are none of them a load-success or completeness statement. Low activity alone does not prove a broken pipeline, and ingestion or transformation success alone does not prove the data reached your warehouse when delivery is a separate stage. Where that evidence is unavailable, a dashboard should show “Unknown” and keep showing the results it can support.
Before reconciling a number
If a dashboard value or warehouse query looks different from another system, check Data Health first. Freshness is only one cause; for the full cross-tool checklist, see Why would external reports not match the SourceMedium dashboard?.Data Health is a readiness check, not a full reconciliation. It can tell you whether data is fresh and available. It cannot prove that two reports use the same metric definition, date basis, or filters.
When to contact SourceMedium
Reach out to your SourceMedium team when:- A core table such as orders, customers, or ad performance remains stale after the expected refresh window.
- A connected platform is empty when you expect recent activity.
- Multiple core domains are stale at the same time.
- Data Health looks healthy, but a discrepancy remains after checking metric definitions, filters, and date basis.
- You need to know whether a freshness issue affects one table, one platform, one store, or all data.
Example questions
You can ask about data health in natural language:- “How is my data health?”
- “Can I trust my last 7 days of data?”
- “Which tables are fresh?”
- “What data do I have available?”
- “Are my tables up to date?”
- “When was my orders data last updated?”
Data Health vs Attribution Health
Related resources
Attribution Health
Diagnose and improve tracking coverage for marketing attribution.
Data Freshness
How Seller Central and Amazon Ads settle after the previous day.
Why (direct) / (none) happens
Common causes of missing attribution and how to fix them.
Data Architecture
How SourceMedium structures and transforms your data.