# Enterprise TDM: HR and payroll data for ERP sandboxes

Dokimore documentation · Test data and credentials · Updated 2026-11-16 · 10 min read

Written for: QA leads and test managers, SDETs and automation engineers, Engineering managers · Where it applies: every market

> Enterprise TDM makes the test data an HR, payroll or ERP project needs and is honest about every number in it: employees across countries with managers, payslips whose year-to-date lines add up, prior-year payslips and a W-2 preview, scenario templates for the awkward cases (a 401(k) limit, a deceased employee, multi-state tax), a rule check after every save, and posting to SAP, Workday, Oracle, Salesforce or your own API, in sandbox and test environments only. A stored dataset gets versions, reservations, an aged copy, a scaled copy of up to a million rows, and a signed evidence record.

## The four tabs

Data Generate builds a batch from a scenario and a distribution. The HCM Lifecycle Generator builds an interconnected multi-country dataset across the whole employee lifecycle: onboarding, payroll, leaves, hikes, promotions, termination, severance, retirement. Data Posting to ERP sends a batch to a connection as a job. ERP & API Connections holds the connections. There is no plan gate on the module; the usage line shows this month's runs against the plan's allowance and says, in its own words, that nothing is ever blocked. A member with the Viewer role reads everything and changes nothing.

## Data Generate

Choose the country, then a scenario template (search by number or keyword: 401k, HSA, deceased, union, H-1B, expat, RSU, garnishment, multi-state, Quebec) or a custom distribution that must total 100%. Set the record count (capped per batch), the format (JSON, CSV or Excel), a seed (the same seed gives the same data), and whether every record needs a current-year or a prior-year paycheck. Advanced settings apply a work and home state pair to every record (the server computes the multi-state tax), a year-to-date target range for US 401(k) and HSA balances (the server solves salary, contribution rate and periods per record, and says when a target is out of reach), an annual pay range, and a company and client id.

The result shows the rule check and referential integrity of the stored batch, the payslips with their current-year and prior-year lines, a W-2 preview of the prior tax year (boxes 1, 3 and 5), a validation score out of 100 that reads "NOT safe to post" when a blocking issue was found, and a statistical profile computed only when you ask for it. A notice lists the regulatory figures that change next year and which one is still in use.

## The HCM Lifecycle Generator

Ten sections describe the population: the total, a seed and the as-of date (the dataset's "today"); the employment-status split; countries, subdivisions, cost areas and remote staff; the management span; salary floor and ceiling, currency and hikes; job roles; pay frequencies and payslip months (1 to 48); service bands; leave rules, global and per country; the employer shown on each payslip; and the earning, deduction and tax code catalogues with your own custom codes. A scenario library fills the form with one click. The result lists the entity links (how many payslips, earnings lines and deductions fan out per employee), a sample employee with every linked record, and the integrity checks.

Send Dataset to Client API posts the whole dataset as JSON to a saved connection or a custom URL. The target environment is required and production is refused; you tick "I have reviewed the preview and integrity checks" yourself, every time; a dataset with failing integrity checks is never sent until you accept the failures by name. Auto-send runs the same gate for a pipeline. A configuration can be saved in the browser or exported as a JSON file for a teammate.

## Stored datasets: versions, reservations and copies

- Reserve a dataset for a test run, from 4 hours to 14 days, with a note everyone in the workspace can read; until the reservation ends the quota and the expiry leave it alone.
- Save a dataset as the next version of a named series, and restore an earlier version as the newest one. When the workspace quota is full, the oldest unreserved dataset makes room.
- Age a copy: every date moves by the same amount, month-ends stay month-ends, pay dates keep their weekday, under any-day, US-business-day or weekday rules. The original is unchanged.
- Scale up a copy to as many as 1,000,000 rows in the main table, with new ids, links that stay inside each copy and new unique values; the same seed and size give the same copy.
- Mask an extract: upload CSV, TSV or JSON files (16 MB in all), choose a rule per column, and get a dataset in which personal data is replaced and the files still join as before. A kept column that still looks like personal data needs a rule or your acknowledgement. The masking key is made for the one run and destroyed. The masked copy is pseudonymised, which under privacy law is still personal data.
- Download as CSV, Excel (one sheet per table), JSON, a SQL script for PostgreSQL (CREATE TABLE and COPY in one transaction), a signed evidence record that anyone can verify offline with the command-line tool, or a readable summary of the same evidence.

## A check on every dataset

Every saved dataset is checked against the rule catalogue. A warning never blocks anything. A failed error rule blocks posting from that dataset until the data is fixed or a new batch is generated; a job can be posted anyway with an explicit override, which is recorded. Results and the evidence record carry positions and field names, never a value.

## Posting to an ERP sandbox

A connection names the system (SAP, Workday, Oracle HCM, Salesforce, or your own REST, SOAP or GraphQL endpoint), the environment (sandbox, staging, UAT or development; production is blocked), the base URL, the authentication (username and password, bearer token, API key or OAuth client credentials), custom headers and protocol settings, and can be tested before use. Credentials are encrypted before they are stored.

A job posts newly generated records or a stored dataset, as the entity you name, and moves through pending, queued, in progress, posting, posted, verifying, verified and complete, or failed, retrying, dead letter and cancelled. Retry a job, cancel it, or delete it (records the system already accepted stay there). Dead-lettered records keep their payloads and can be re-sent, with a warning that a record whose first attempt ended without a clear answer may already exist. Read-back reconciliation fetches what the system stored and compares it field by field with what was sent. Every job has an audit trail and a landed CSV.

## From a pipeline, and from plain words

A CI/CD API key lets a pipeline generate a dataset, read it and post it to a sandbox connection without a browser session. A key acts as the member who created it, expires after 30, 60 or 90 days, and can do nothing else: connections, keys, the audit log and every delete need a signed-in person. The qax-tdm command-line tool is one file with no installation: generate, reserve, age, mask and post from any pipeline, with retries when the server asks it to wait and a clear exit code. For SAP calls, a small connector runs next to your SAP system with your own licensed SAP Java Connector (Java 21 or newer), connects out to the platform, and is tested with SAP's own echo function, which changes nothing.

The configuration assistant takes a sentence such as "500 employees in Texas and Ohio, 20% leavers, paid biweekly" and fills the form; the rules decide, a model inside the platform is used only for wording the rules could not read, the fields it filled are marked as suggested, and you review the form before you click Generate.

## Questions people ask about this

### Is a masked extract still personal data?

Yes. It is pseudonymised: names, identifiers and contact details are replaced and the joins stay intact, but the result is still personal data under privacy law and is handled as such.

### Can I post to a production system?

No. A connection must name a sandbox, staging, UAT or development environment, and a production target is refused at the connection and again at the send.

### What stops a bad dataset reaching the sandbox?

The rule check that runs after every save. A failed error rule blocks posting from that dataset; a warning never does. An explicit override posts it anyway and is recorded.

This guide is about test data, the credentials vault and enterprise test-data management: https://dokimore.com/product/test-data

Source: https://dokimore.com/documentation/enterprise-test-data-management
