MTurk Exit Kit · public launch sample · updated 29 August 2026

MTurk closes 30 September. Leave with a clean evidence trail.

A fixed-scope, vendor-neutral archive and workflow-mapping service for small research teams and legacy requesters. Preserve what the workflow was, what remains unsettled, and what must be tested before choosing a replacement.

Use the free 12-control close-out boardRequest free readiness triageInspect fictional delivered exampleDownload fictional archive worksheet

30 Sep 2026MTurk permanent closure and HIT submission stop
30 Oct 2026requester approval/rejection and bonus window ends
28 Jan 2027transaction-history access ends
0real participant records, migrations or equivalence claims
Evidence boundary: Amazon’s Closure FAQ establishes the dates above and says the MTurk Worker type in SageMaker Ground Truth and Amazon Augmented AI will no longer be available from 30 September 2026. It does not establish that any other platform has an equivalent worker pool, qualification model, sampling frame, consent process, price, API or quality outcome. This service is operational documentation and test planning—not IRB/ethics, privacy, labour, legal, procurement or research-methods advice.

Three clocks, one owner checklist

By 30 September

Freeze each active HIT type, qualification rule, batch state, reward basis, response target, code dependency and consent/IRB artifact pointer. Record submitted versus unsubmitted work separately.

By 30 October

Assign an owner for pending approvals/rejections and any intended bonuses. Amazon says completed HITs left without action are auto-approved after the standard 30-day window.

By 28 January

Export and retain transaction-history evidence under the team’s own authorised retention rules. The pack records export state and location pointers; it does not ask for worker identities or raw participant data.

Fictional requester: before a replacement decision

EX-MT-01 · Northbridge Behaviour Lab

Wholly fictional composite. It does not describe a customer, university, study or real participant dataset.

archive_ready_mapping_hold
Workflow observed

Two survey-screening HIT types and one follow-up batch are represented in the sample worksheet. Reward and response counts are synthetic placeholders.

Archive state

HIT-purpose, qualification logic, approval/bonus owner, transaction-export state, code dependency and ethics-artifact pointers are separated from participant data.

Mapping hold

No replacement platform is selected. Worker-pool fit, screening semantics, sample comparability, API shape, payment model and pilot acceptance rules remain unknown.

Smallest safe next test

Use a synthetic questionnaire and a separately approved small pilot plan. Compare operational states and data fields; do not claim methodological equivalence from one pilot.

Archive first: 24 explicit fields

Field groupExamples in the worksheetWhy it existsSafe state
Identity & purposerequester_record_id, workflow_name, hit_type_id_or_redacted, task_purposeNames the operational unit without requiring a public HIT or participant identity.Known / redacted / unknown
Qualification & batchqualification_logic, batch_state, target_response_count, actual_response_countStops a platform label from being mistaken for a transferable sampling rule.Documented / partial / unknown
Money & settlementreward_basis, bonus_rule, approval_window, pending_review_countAssigns closure actions without providing payroll, tax or labour advice.Owner assigned / hold
Evidence & sensitivitytransaction_export_state, consent_or_irb_pointer, pii_or_sensitive_data_flag, worker_id_handlingRecords pointers and sensitivity state; raw data is out of scope for the public sample.Pointer only / restricted / unknown
Code & replacementcode_or_api_dependency, candidate_replacement_route, equivalence_claim_state, acceptance_testTurns “move platforms” into a bounded mapping and test backlog.Candidate only / no equivalence claim
Ownershipowner, next_action, unknowns, source_dateMakes every unresolved item reviewable before the deadline.Named role / due date / hold

Download the two-row fictional CSV → · Open dated source hashes and claim boundaries →

Replacement routes are questions, not recommendations

Research-participant platforms

Prolific and CloudResearch Connect have public researcher-facing participant-recruitment pages. The mapping sprint can record their public workflow fields and prepare owner questions.

Not established here: availability for a named geography or study, qualification equivalence, representative sampling, IRB acceptability, API parity, price, response quality or migration success.

Human-review / data-labelling workflows

AWS maintains public SageMaker Ground Truth and SageMaker AI surfaces; the latter currently carries an end-of-support notice that includes Augmented AI. Separately, the MTurk FAQ says the MTurk Worker type ends on 30 September.

Not established here: current availability for a named customer, whether an existing job can use another workforce type, whether schemas carry over, or whether a managed/private workforce fits the requester’s budget, methods or data boundary.

Buy the bounded evidence step

US$250 · Archive Readiness Pack

  • Up to three requester workflow/HIT-type rows
  • 24-field archive worksheet and closure-owner checklist
  • Unknown/restricted-data register
  • One factual/source-link correction round

Customer exports and stores authorised records. We do not receive raw participant responses, worker identity data, payment details or account credentials.

US$750 · Assisted Mapping Sprint

  • Everything in the Archive Readiness Pack
  • One requester workflow mapped to up to two candidate routes
  • Field/qualification/API gap matrix
  • Synthetic or redacted small-pilot plan with stop rules

Prices are validation hypotheses, not proven market rates. No worker recruitment, live migration, platform purchase, data transfer or research-equivalence guarantee.

Start with the free close-out boardRequest free readiness triageSee the fictional US$750 delivery shapeAsk a scope question

Source register: six first-party public pages returned HTTP 200 on 27 August 2026 and were captured/hash-checked: Amazon MTurk homepage and Closure FAQ; AWS SageMaker Ground Truth and SageMaker AI; Prolific for Researchers; and CloudResearch Connect for Researchers. Platform marketing and support-status statements are treated as vendor statements, not independent performance evidence. No login, account creation, participant contact, customer-data access, worker-data collection, API call to a requester account or external outreach occurred.