Test and QA Plan

How the build is validated before go-live: three sequential test phases, edge cases, and production verification.

4 pagesPDF · Quality

FS-DOC-06Quality

Test and QA Plan

Upsell and Cross-sell Triggers

[YourCompany.com] · Sales Department · Prepared by FullSpec · [Today's Date]

This document defines the structured test and quality assurance programme for the Upsell and Cross-sell Triggers automation. It is written for the FullSpec build team and covers every test case from isolated unit checks through to a live production run. Testing proceeds in three sequential phases: no phase begins until the previous phase fully passes. The process owner joins Phase 3 to confirm end-to-end behaviour before the automation is considered production-ready.

01Testing approach

All testing follows three phases in sequence. No phase begins until the previous phase has fully passed, meaning every test case in that phase carries a confirmed pass status. Phase 1 isolates individual agent and tool behaviour in a sandbox environment. Phase 2 tests the handoffs between agents and system integrations, and exercises edge cases that could occur in real traffic. Phase 3 runs the complete flow end to end in production using a controlled test record.

Phase 1 — Unit Testing

Isolated agent and tool behaviour in sandbox

Scope

Each agent and each tool connector tested in isolation: Segment trigger parsing, HubSpot data fetch, Offer Selection Agent scoring logic, Outreach Drafting Agent email composition, Gmail send, HubSpot activity logging, Slack notification, Google Sheets append.

Environment

Sandbox only. No production data, no production API keys, no live customer records.

Credential rule

All sandbox credentials only. Production tokens must not be present in the environment during Phase 1. Sandbox credentials are stored in the build environment secret store under the prefix SANDBOX_.

Pass condition

All cases T01 through T14 return the expected output with no error state. Zero failures permitted before Phase 2 begins.

Who runs it

FullSpec team only. No process owner involvement required at this stage.

Phase 2 — Integration and Edge Case Testing

Agent handoffs, system integrations, and exception scenarios

Scope

Full agent-to-agent handoffs, data passing between Segment, HubSpot, Gmail, Slack, and Google Sheets, plus edge cases covering duplicates, missing fields, unavailable tools, and human review timeout.

Environment

Sandbox for integration tests. Edge case tests may use a staging environment with anonymised production-shaped data where available.

Credential rule

Sandbox credentials for all integration tests. If staging environment credentials are used for edge cases, they must be scoped read-only on HubSpot and have no Gmail send permission.

Pass condition

All cases T15 through T22 pass and all edge cases E01 through E06 resolve to the documented fallback state with no unhandled exceptions.

Who runs it

FullSpec team only. Results are documented and shared with the process owner prior to Phase 3.

Phase 3 — End-to-End Production Test

Full live run on production with process owner review

Scope

One complete production run traced from Segment trigger through to Google Sheets append. Includes the high-value routing branch, error log review, and final process owner sign-off.

Environment

Production. Real API credentials, real tool connections. Test record must use an internal email address and be voided immediately after the run.

Credential rule

Production credentials are used. All credentials must be stored in the production secret store only. The test Gmail recipient must be an internal address at [YourCompany.com].

Pass condition

All cases T23 through T26 pass. The process owner completes the final review case (T26) and confirmation is recorded in the FullSpec platform.

Who runs it

FullSpec team runs T23 through T25. The process owner participates in T26 to confirm output quality and approve the automation for full rollout.

Test and QA PlanPage 1 of 4

FS-DOC-06Quality

02Phase 1 — unit test cases

Sandbox credentials only. Do not use production API keys, production HubSpot contact IDs, real customer email addresses, or live Gmail send credentials during Phase 1. Every tool connection must point to its designated sandbox or test instance. Confirm SANDBOX_ prefix on all secret store entries before running any case.

Segment trigger and HubSpot data fetch (pre-agent pipeline)

Scenario

Input

Expected output

Pass criteria

T01

Valid qualifying event received from Segment

Segment track event: upsell_trigger_fired, properties: { customer_id: 'cust_sandbox_001', event_type: 'spend_threshold', value: 5000 }

Workflow execution begins; customer_id extracted and passed downstream

Execution log shows trigger received; customer_id present in payload; no error thrown

T02

Segment event with missing customer_id

Segment track event: upsell_trigger_fired, properties: { event_type: 'spend_threshold', value: 5000 } (no customer_id)

Workflow halts; error logged: MISSING_CUSTOMER_ID; no downstream steps fire

Execution halts at trigger parse step; error code MISSING_CUSTOMER_ID present in log

T03

Successful HubSpot account data fetch

customer_id: 'cust_sandbox_001' mapped to sandbox HubSpot contact ID hs_sandbox_4421

CRM record returned: account_owner, spend_tier, open_deals, last_activity_date

All four fields present in returned payload; HTTP 200 from HubSpot Contacts API

T04

HubSpot contact not found

customer_id maps to a HubSpot ID that does not exist in sandbox

Error state: HUBSPOT_CONTACT_NOT_FOUND; workflow halts; no agents fire

Error code present in log; downstream agents confirm no execution

T05

HubSpot API returns 429 rate limit response

Sandbox HubSpot mock returns HTTP 429 on first call, HTTP 200 on retry

Orchestration layer retries after 10-second backoff; second call succeeds; data fetched

Retry logic fires once; successful fetch on second attempt; total latency under 30 seconds

Offer Selection Agent (complexity: Moderate, estimated build time: 8 hours)

Scenario

Input

Expected output

Pass criteria

T06

Standard account, clear spend-tier match

spend_tier: 'mid', product_usage_signal: 'feature_ceiling_reached', account_segment: 'SMB', catalogue: Google Sheets sandbox tab loaded

Ranked offer recommendation returned: { offer_name: 'Pro Plan Upgrade', rationale: 'Customer has reached feature ceiling on current tier' }

offer_name and rationale fields present; offer exists in catalogue; response latency under 5 seconds

T07

Multiple offers equally scored, tie-break applied

spend_tier: 'mid', two offers with identical score in catalogue

Agent selects the offer with higher margin_priority flag; rationale references tie-break rule

Single offer returned; rationale string contains 'margin priority'; no error thrown

T08

No eligible offer found for account profile

spend_tier: 'low', product_usage_signal: 'inactive', no catalogue row matches all criteria

Agent returns no_offer_available flag; workflow routes to suppression state; no email drafted

no_offer_available flag present in output; Outreach Drafting Agent execution log shows no trigger

T09

Google Sheets catalogue unavailable during scoring

Sandbox Sheets connector returns a connectivity error mid-execution

Error logged: CATALOGUE_FETCH_FAILED; workflow halts; alert queued for FullSpec team

Error code present; no partial recommendation passed downstream; alert entry recorded

Outreach Drafting Agent (complexity: Moderate, estimated build time: 8 hours)

Scenario

Input

Expected output

Pass criteria

T10

Standard account, full personalisation fields available

offer_name: 'Pro Plan Upgrade', account_owner: 'Rachel Kim', customer_name: 'Acme Corp', last_purchase_date: '2024-03-01'

Draft email: subject line contains offer name; body references customer name and last purchase date; tone matches brand guidelines

All three personalisation tokens populated; no placeholder text visible; draft passes character limit check (under 400 words)

T11

High-value account routing flag set

account_tier: 'enterprise', high_value_flag: true, offer drafted

Draft queued to rep review queue; auto-send NOT triggered; Slack alert sent to account owner

Gmail send API not called; draft visible in review queue; Slack message delivered to sandbox channel

T12

Standard account, auto-send path

account_tier: 'standard', high_value_flag: false, draft complete

Email sent via Gmail sandbox; HubSpot activity log entry created; Slack notification posted

Gmail sandbox delivery confirmed; HubSpot activity record created with correct offer and timestamp; Slack message posted

T13

customer_name field missing from CRM record

customer_name: null, all other fields present

Draft uses account_domain as fallback salutation; email generated without error; flag logged: PERSONALISATION_FALLBACK

Email draft does not contain 'null' or empty salutation; PERSONALISATION_FALLBACK flag in log

T14

Gmail sandbox send fails with HTTP 503

Outreach Drafting Agent attempts send; Gmail sandbox mock returns HTTP 503

Error logged: GMAIL_SEND_FAILED; email held in retry queue; Slack alert sent to account owner noting send failure

Email not marked sent; retry queue entry created; Slack alert delivered; no duplicate draft created

Test and QA PlanPage 2 of 4

FS-DOC-06Quality

03Phase 2 — integration and edge case tests

Integration tests verify that data passes correctly across the full agent handoff chain and that each connected tool receives and processes the expected payload. All IDs continue the T sequence from Phase 1.

Scenario

Input

Expected output

Pass criteria

T15

Segment to HubSpot to Offer Selection Agent full handoff

Valid Segment event fires; HubSpot fetch succeeds; payload passed to Offer Selection Agent

Offer Selection Agent receives complete payload including spend_tier, product_usage_signal, and account_segment

Agent execution log shows all three fields present on receipt; no data loss between steps

T16

Offer Selection Agent to Outreach Drafting Agent handoff

Offer Selection Agent returns { offer_name: 'Pro Plan Upgrade', rationale: '...' }

Outreach Drafting Agent receives offer_name and rationale and uses both in draft

Both fields referenced in generated draft; no null tokens; agent execution sequence confirmed in log

T17

Auto-send path: Outreach Drafting Agent through to Google Sheets append

Standard account, high_value_flag: false, full draft generated

Gmail send confirmed; HubSpot activity logged; Slack posted; Google Sheets row appended with customer, offer, date, account_owner

All four downstream actions confirmed in execution log; Google Sheets row present and correctly formatted

T18

High-value routing path: draft queued, rep approves, send fires

high_value_flag: true, rep approves draft in review queue within session

Gmail send fires after rep approval; HubSpot log created post-approval; Slack alert references approval event

Gmail send timestamp is after approval event timestamp; HubSpot log references rep approval; no premature send

T19

HubSpot activity log entry format validation

Auto-send path completed for sandbox customer record

HubSpot activity record contains: activity_type: 'upsell_outreach', offer_name, trigger_event_type, send_timestamp (ISO 8601)

All four fields present; timestamp format valid; activity type matches defined enum in HubSpot property schema

T20

Slack notification content and delivery

Send completed for sandbox account owned by 'Rachel Kim' in sandbox Slack workspace

Slack message delivered to correct channel; message contains offer_name and HubSpot record URL

Message received in sandbox channel; offer_name matches sent offer; URL resolves to correct HubSpot sandbox record

T21

Google Sheets append: correct column mapping

Completed run for sandbox customer 'Acme Corp', offer 'Pro Plan Upgrade', account_owner 'Rachel Kim'

New row in tracking sheet: Column A = customer name, Column B = offer, Column C = send date (YYYY-MM-DD), Column D = account owner

Row present; all four columns populated correctly; no row appended to wrong sheet tab

T22

End-to-end latency benchmark

Valid Segment event fired at T+0 in sandbox

Gmail send or review queue entry created within 90 seconds of trigger

Execution log timestamps show total elapsed time under 90 seconds from trigger receipt to send or queue entry

Edge case tests cover failure modes, data anomalies, and timeout scenarios that could occur under real traffic conditions.

Scenario

Input

Expected output

Pass criteria

E01

Duplicate Segment event for the same customer within 24 hours

Two identical upsell_trigger_fired events for cust_sandbox_001 fired 10 minutes apart

Second event detected as duplicate; suppressed; no second email drafted or sent; deduplication log entry created

Only one execution proceeds; dedup log shows suppression of second event; single HubSpot activity entry

E02

HubSpot record exists but account is marked churned or closed

HubSpot contact fetched; lifecycle_stage: 'churned'

Workflow halts after CRM fetch; no offer selected; no email sent; suppression logged

Execution stops at eligibility check; suppression record created; no downstream agents fire

E03

Offer catalogue in Google Sheets is empty or unreachable

Offer Selection Agent attempts to read Google Sheets; sheet returns 0 rows or HTTP 503

Agent returns CATALOGUE_UNAVAILABLE; workflow halts; FullSpec team alerted via internal error channel

No offer passed downstream; error logged with timestamp; alert delivered

E04

Outreach Drafting Agent produces draft exceeding 400 words

Agent generates a draft of 520 words due to verbose account context

Draft is automatically truncated to 400 words with an ellipsis appended; flag logged: DRAFT_TRUNCATED

Sent draft is 400 words or fewer; DRAFT_TRUNCATED flag present; no content after ellipsis is sent

E05

Slack API unavailable when notification is attempted

Slack API returns HTTP 503 at notification step

Notification failure logged; workflow does not halt; email send and HubSpot log complete as normal; retry attempted after 5 minutes

Gmail send and HubSpot log confirmed despite Slack failure; retry log entry present; no workflow halt

E06

High-value account in review queue not actioned within 48 hours

Draft queued for rep review; no action taken for 48 hours

Escalation Slack alert sent to sales manager; draft remains in queue; no auto-send triggered; timeout logged

Escalation message delivered to manager channel at T+48h; draft not sent; timeout entry in execution log

Test and QA PlanPage 3 of 4

FS-DOC-06Quality

04Phase 3 — end-to-end production test

Important before running Phase 3: The Gmail recipient for all test cases must be an internal address at [YourCompany.com]. Do not use a real customer email address at any point. The HubSpot contact record created for this test must be flagged as TEST_RECORD and deleted or voided by the FullSpec team immediately after T26 is complete. The Google Sheets row appended during this test must also be deleted before the sheet is used for live reporting. Confirm with the process owner that a named internal test address is available and that the HubSpot sandbox-to-production credential swap has been reviewed before proceeding.

Production credentials are now active. Confirm that all SANDBOX_ prefixed secrets have been replaced or scoped out before executing any case in this phase. The FullSpec team is responsible for credential verification prior to T23.

Scenario

Input

Expected output

Pass criteria

T23

Full production run: standard account, auto-send path

Live Segment event fired for internal test contact (spend_threshold crossed); HubSpot production record for test contact pre-created with spend_tier: 'mid', high_value_flag: false

Offer selected from live catalogue; personalised email drafted and sent to internal test address; HubSpot activity logged; Slack notification posted to sales channel; Google Sheets row appended

Email received at internal test address within 90 seconds of Segment event; HubSpot activity record present with correct offer and ISO 8601 timestamp; Slack message in sales channel; Google Sheets row matches expected format across all four columns

T24

Full production run: high-value account, rep review path

Live Segment event fired for internal test contact with high_value_flag: true

Draft queued in rep review queue; auto-send not triggered; Slack alert sent to account owner noting review required; HubSpot activity record created with status: 'pending_review'

No email delivered to internal address at time of queue entry; review queue item visible; Slack alert delivered; HubSpot status field set to 'pending_review'

T25

Error log verification across full production run

Review execution logs from T23 and T24 in the production logging environment

All steps present in log with status, timestamp, and duration; no unhandled exceptions; retry events (if any) correctly attributed; SANDBOX_ credentials absent from log entries

Log contains an entry for every named workflow step; no error status on any step; no sandbox credential string appears in any log field

T26

Process owner review and approval

Process owner reviews: email received at internal test address from T23, HubSpot activity log entry, Slack notification, Google Sheets row, and rep review queue entry from T24

Process owner confirms: email tone and personalisation are correct, offer name is accurate, HubSpot log is complete, Slack message is clear and actionable, Google Sheets row is correctly formatted. Process owner approves the automation for full rollout. Approval is recorded in the FullSpec platform confirmation log.

Process owner confirms all five outputs meet expected quality. Approval recorded in FullSpec confirmation log. No separate sign-off block required. Automation cleared for full rollout.

Once T26 is confirmed, the FullSpec team will void the test HubSpot record, remove the Google Sheets test row, and archive this QA plan with the confirmation timestamp. The automation is then eligible for the pilot rollout defined in the Launch Plan. Contact support@gofullspec.com with the confirmation reference if any issue arises during the voiding step.

Test and QA PlanPage 4 of 4

Test and QA Plan

Test and QA Plan

01Testing approach

02Phase 1 — unit test cases

03Phase 2 — integration and edge case tests

04Phase 3 — end-to-end production test

More documents for this process