Scheduled report reconciliation
One dependable report from two exports that disagree
You send two exports that should agree and don't, for example project status from one tool and recorded hours from another. We reconcile them with written rules, build the report, and list every record that could not be matched, so nothing is dropped to make the totals look clean.
Who it is for
Operations consultancies, implementation agencies and service businesses that rebuild the same report by hand from two tools. It fits when the inputs recur, the definitions are stable, and one person can decide what happens to records that conflict.
If your current software or provider already produces the report correctly, keep using it. This is for the reconciliation that is still left over.
What you send
- Two exports, CSV or XLSX, up to 2,000 rows combined.
- Your definitions for the five metrics the report shows, and which source wins when the two disagree.
- The totals you expect, where you have them.
- One named approver who signs off the rules and decides on unresolved records.
What you get
- Your original files, preserved unchanged.
- A reconciled dataset in which every input row can be traced.
- A change log listing every edit and the rule behind it.
- Five agreed metrics and one reusable report.
- An exception register: every record we could not match or resolve, with the decision it needs from you.
- Repeat-run instructions, so the same report can be produced again, by us or by you.
- Checks you can see: reconciled row counts and totals, plus a rerun on a fresh input to show the report repeats correctly.
- One consolidated correction round.
$1,000
For the first project: the rules, the checks and the first report.
$300 for each later run of the same approved workflow, with the same file structure, the same 2,000-row limit, the same metrics and one correction round.
Paid in full, in one payment, before work starts. We send the payment link once the scope is agreed in writing.
Lead time: 10 business days after we have complete inputs and payment.
Tell us what the report is for and how often you need it. A redacted sample of each export helps. You get a fixed scope back, or a plain no if it is not a fit.
What it does not include
- Writing to a live ERP or CRM, or maintaining an API connection.
- Bookkeeping sign-off, accounting assurance, tax or legal advice.
- Payroll decisions and predictive forecasts.
- Sensitive regulated records.
- Open-ended spreadsheet support.
- Guessed values. A missing fact stays in the exception register until your approver decides it.
- A second report, a changed source, a new business definition or another end client's workflow. Each of those is a separately scoped project.
- Rush delivery, or any promise of time or cost savings.
Synthetic example, fictional data
What the checks look like
This example was built on invented data for a fictional consultancy: a project status export with 40 projects and a time export with 300 rows for August 2026. Problems were planted in the files on purpose. Here is how each one was surfaced.
| Planted problem | What the report did |
|---|---|
| The same time entry appears twice | One copy removed, logged in the change log with the rule that removed it. The hours are counted once. |
| 4.00 hours charged to a project ID that is not in the status export | Kept in total hours, listed as unmatched, with the question: a new project or a mistyped ID? |
| 2.50 hours with no project ID at all | Kept in total hours, listed as a missing key, asking which project it belongs to. |
| A project marked Complete on 14 August, with 4.50 hours logged against it on 19 and 26 August | Counted as neither open nor completed. Flagged as a conflicting status for the approver to decide. |
| Status spelled two ways | Normalised by a stated rule, and each change logged. |
The report showed 24 open projects, 15 completed, 1,019.25 recorded hours, 6.50 unmatched hours and 1 unresolved status record. A second script worked out the same five figures a different way before the report was built, and the report is not written if the two disagree.
A fresh September pair of files then ran through the same code with no edits. An October file with the hours column renamed stopped the run: it named the missing column and wrote nothing, rather than producing a plausible report from the wrong input.
The example shows the checks on invented data. It is not a client result, and it does not measure accuracy or time saved on your files.
All services · Supplier file preparation · Campaign variations