Data Health and reliability reports

Run Data Health analyses to grade financial book quality, schedule recurring runs across the workspace, and surface cash vs accrual basis.

Data Health grades a company's financial books across four reliability categories and produces an AI-generated summary with prioritized findings. Each run records the accounting basis (cash, accrual, modified cash, or unknown) used during analysis.

Overview

You can run Data Health on any company you have access to from Data Health in the workspace navigation. Each run scores Statement Integrity, Earnings Reliability, Solvency and Liquidity, and Working Capital Cycle on an A+ to F scale, then combines them into an overall letter grade. You can also schedule recurring runs across all companies, internal companies only, shared companies only, or a specific list.

Where to find it

The dashboard lives at Data Health in the workspace navigation. Per-company reports open at Data Health > [Company].

The dashboard has two tabs:

Companies — every accessible company with the latest letter grade, last run, next scheduled run, and months of books on file.
Schedules — visible to Owners and Admins only.

How to run Data Health on a company

Go to Data Health.
Click Run now in the page header, or click the run action on a row in the Companies tab.
Pick the target company in the picker.
Confirm the Target month in the run modal. The default is the latest month with accounting data.
Select a Data window: 36 months (standard, 1 token) or 60 months (deep, 2 tokens).
Review the Company context panel. If completeness is at 0, you must acknowledge running without context or add it first.
If the run will exceed your token balance or the company is shared, check the overage acknowledgment box.
Click Use N tokens and run or Run and charge $X.XX.

The first run for a target month also generates a reliability report before scoring, which typically completes in under 60 seconds. The page polls every 5 seconds and refreshes automatically when the run completes. You can close the modal during processing — the report appears on the company page when ready.

How to read a company report

Open a company from the Companies tab to see its report at Data Health > [Company].

The report has the following sections:

Hero card — overall letter grade, AI-generated summary, and book status pills (GL synced, Period closed, Integrations).
Category breakdown — four cards, each with a category letter grade, description, and an info icon that opens the scoring rubric.
Analysis summary — sorted findings with severity (critical, important, informational), description, impact, recommended action, and root cause hint.
Report rail (right side, large screens) — list of past runs you can switch between, plus a context summary.

A grade of — (N/A) means the company has fewer than 12 months of bookkeeping data. Twelve months is the minimum required to grade a company.

Letter grade ranges

Grade	Tone
A+, A	Success (green)
B+, B	Neutral (blue)
C+, C	Warning (amber)
D, F	Destructive (red)

Book status indicators

Each report shows three book-status pills above the AI summary:

GL synced — last sync recency from the connected accounting system. Red if the tenant is inaccessible, amber if the sync is stale, green if fresh.
Period closed — the period lock date from the accounting system. Amber if older than 60 days.
Integrations — count of active integrations. Red if zero are connected, amber if some are disconnected, green if all are connected.

Cash vs accrual basis

Each run records the company's accounting basis as cash, accrual, modified_cash, or unknown. The basis is read from the connected accounting system or the company context form. The basis affects how findings are interpreted — for example, missing AR/AP under cash basis is a data characteristic, but under accrual basis it is flagged as a finding.

How to add company context

Open the company in Data Health > [Company].
Click Add company context (or Edit context if context exists).
Fill in systems, business model, and close process fields.
Click Save context.

A completeness score of 80% or higher produces findings that account for the company's systems and close process. The score appears as a progress bar on the context page and as a percentage on the run modal.

How to schedule recurring runs

You must be an Owner or Admin to manage schedules.

Go to Data Health.
Click the Schedules tab.
Click New schedule.
Enter a Name.
Pick a Cadence:
- Monthly — set the day of month (1 through 28).
- One-off — set a single fire date.
Pick a Scope:
- All companies — every company in the workspace.
- Internal only — companies your workspace owns.
- Shared only — companies shared with your workspace.
- Specific — pick one or more companies from the list.
Pick a Data window: 36 months or 60 months.
Pick a Budget behavior:
- Skip on insufficient — skip companies that lack tokens at fire time.
- Charge overage — run regardless and bill overage at next invoice.
- Halt if over — stop the schedule before any overage is incurred.
Click Preview to see the target count, estimated tokens, and estimated overage in cents.
Click Create schedule.

The Schedules tab shows the next fire time, last fire time, and last fire outcome for each schedule. You can edit or delete schedules from the row actions. Deleting a schedule is permanent.

Permissions

Action	Owner	Admin	Member
View dashboard	Yes	Yes	Yes
View company report	Yes	Yes	Yes
Run Data Health manually	Yes	Yes	Yes
Edit company context	Yes	Yes	Yes
Manage schedules	Yes	Yes	No

Access to Data Health is gated at the organization level. If your organization does not have Data Health enabled, the Data Health page returns a 404.

Limitations

A company needs at least 12 months of bookkeeping data to receive a graded report.
Only one run can be in progress per company at a time.
A 36-month run costs 1 token; a 60-month run costs 2 tokens.
Tokens reset on the 1st of each month and do not roll over.
Shared companies always run as overage — they do not draw from the workspace token allocation.
Overage runs require an active subscription with overage allowed; otherwise the run is blocked.
Schedule day of month is capped at 28 to avoid month-end edge cases.
One-off schedule dates must be in UTC.
A run that fails on infrastructure refunds the token automatically. Failures from the AI analysis itself do not.
The accounting basis falls back to unknown when the connected system does not report it and no value is set in company context.

Estimated tax

Macro-Economic Correlations