Automation
Testing Marketing Workflows Before Launch
Test a gambling marketing workflow by trying to make it send to people it must not reach. Seed accounts in every consent and exclusion state, change their state mid-flow, break the decision service on purpose and check that each send was stopped and logged.
In short
- A test passes when the workflow suppresses or holds the right accounts, not when emails arrive.
- Seed accounts for every row of the decision order, plus mid-flow changes.
- Drill failures: decision service down, delayed events, manual uploads.
- Keep the results as evidence and rerun them after every change.
Scope and limits. A test plan template. It uses internal test accounts with addresses you control; never test on real customers.
Why negative tests matter most
The Gambling Commission's guidance on self-excluded customers lists failures operators reported, including "members of staff not following policies and procedures" and "technical vulnerabilities in control systems". These are the failures a happy-path test never finds. The plan below spends most of its effort on sends that must not happen.
Seeded accounts
Create one test account per row, with addresses and numbers you control:
| Account | State | Expected outcome |
|---|---|---|
| T1 | Self-excluded, opted in to everything | Suppressed on every channel |
| T2 | Objected to all marketing | Suppressed on every channel |
| T3 | Age not verified, opted in | Suppressed |
| T4 | Opted in to betting email only | Betting email sent; casino email and all texts suppressed |
| T5 | Opted in, consent record missing wording | Held for review |
| T6 | Opted in to everything, verified, no flags | Eligible for human review only; sent after a recorded approval |
| T7 | Safer gambling flag set, opted in | Suppressed; protective messages still sent |
The same order of checks can be explored interactively in the consent-state explorer, and its CSV export can be filed with your test record. Account T7 tests the flag described in safeguarding overrides and protective messages.
Mid-flow changes
- Start T6 in a multi-step flow.
- After step 1, self-exclude T6. Expected: no further marketing; flow exits.
- Repeat with a casino email opt-out between steps. Expected: casino emails stop, betting emails continue if permitted.
- Repeat with an opt-out by text reply. Expected: texts stop; check whether the reply was meant to cover all marketing.
Measure how long each change took to reach every tool, following propagating preference changes. Compare the result with the rule, such as the two-day limit in code 3.5.3 for Great Britain, and with your own target.
Failure drills
- Decision service down. Expected: sends are held, not sent.
- Event consumer offline. Expected: replay on restart, then sends resume.
- Manual upload. Upload a file that includes T1. Expected: T1 is suppressed by the send-time check.
- Old segment. Reuse a segment built before T1 was excluded. Expected: suppressed.
Product classification
Send a campaign tagged as both betting and casino. Expected: only accounts opted in to both products on that channel receive it, matching the per-product rule in code 5.1.12 for Great Britain, as explained in how to classify a campaign by product.
Record the evidence
For each test, keep the account, the state, the expected and actual outcome, the reason code and the time. Rerun the plan after any change to templates, segments, tools or rules. The same scenarios can be used during a vendor trial with the marketing automation platform criteria. The integration checks that come before this plan are in the CRM integration checklist.
Who should sign off
A workflow test is not complete until someone other than its builder has read the results. A simple pattern:
- The builder runs the plan and records results.
- A second person from marketing operations checks the results against the expected outcomes.
- A compliance owner signs off the safeguarding rows, T1, T2 and T7, before launch.
Keep the signed record with the workflow, so the next change starts from a known state.
Questions
How often should the full plan run?
After every change to a workflow, a tool or a rule, and on a regular schedule even when nothing has changed, because data and integrations drift.
Can we test with real customer records?
No. Use internal accounts with addresses and numbers you control, so a failed test never reaches a real person.
Sources
- Gambling Commission, Ensuring self-excluded customers do not receive your marketing materials, last updated April 13, 2022.
- Gambling Commission, LCCP 3.5.3 Self-exclusion - remote SR code, no date shown.
- Gambling Commission, LCCP Condition 5.1.12 - Direct marketing preferences, no date shown.
Every source was read on October 28, 2026. How we choose sources: data sources.
Safer gambling support is listed on the safer gambling page.