October 14, 2026. These are prepared experiment specifications, not claims that product tests or customer interviews have already happened. Prices, effort and decision thresholds are planning assumptions.
Evidence gates
- Identify a real recurring task and its owner. Preserve a concrete recent incident rather than a vague expression of interest.
- Compare current paid tools, native features and a manual workaround on the same task.
- Produce one bounded result using authorized or synthetic data. Record setup time and exceptions.
- Offer that result with a defined scope and price. Interest and a stated budget are not completed payments.
- Measure repeat use and support before offering a subscription. Two similar paid pilots justify another experiment, not a market-size forecast.
No outreach, purchases, payment collection or customer-data processing has been performed as part of this catalogue.
Files and data
Attachment routing: test different files with the same name, repeated delivery, duplicate content, invalid files, unavailable storage and revoked access. Require an outcome per message, an error log and safe retries. Compare a native flow with Mailparser before writing a multitenant service. A $199 setup experiment covers one agreed workflow; $29/month support only works if recurring work is small.
Supplier CSV: use leading-zero SKUs, duplicates, unknown products, blank versus zero inventory, decimal commas, multiple locations and missing variant dependencies. Show original row, rule and proposed change. Never silently resolve ambiguous identifiers. Initially export a reviewed file rather than update a live store.
Supplier prices: compare 330 ml against 5 L, single units against cases of twelve, per-kilogram against per-pack prices, currency and unknown tax treatment. Separate exact matches, human-confirmed mappings and unresolved cases. A high aggregate matching percentage must not hide a dangerous pack-size substitution.
Commerce
Mixed cases: test twelve mixed bottles, ten cards plus one print, and 10 kg multiples across product pages, catalogue add-to-cart, classic cart and Blocks. Compare the existing Woo Min/Max extension first. Configuration success validates an implementation service, not a reason to clone the plugin.
B2B quotations: check required information, selected items, quantities, variants, acknowledgement and one destination. A quote request must not silently become a paid order. Compare an existing RFQ app plus theme configuration before custom development.
Feed mismatch audit: compare feed value, visible landing-page price and structured data for the exact variant and market. Include scheduled discounts, currency, cached values and plugin upgrades. Compare Google diagnostics and an existing feed validator; report causes rather than promise ad approval.
Download permissions: distinguish editing an existing file from adding a new file, migrations and subscription entitlement. Start with a read-only audit and a reviewed repair plan. Preserve existing access and logs. No bulk regeneration of permissions without a backup and explicit approved scope.
Agency operations
Tracking QA: on an authorized test property, remove an event, rename it, duplicate it, delay it and change consent. Include a normal day with zero conversions. Compare GA4 custom insights. Different attribution definitions are not automatically tracking failures.
WordPress update review: compare specified journeys and screenshots before and after a staged update. Separate dynamic content from regressions. Existing visual-regression products are the baseline; a screenshot difference alone is not a confirmed defect.
Access handoff: record account owner, intended recipient, recovery owner, permission scope and transfer acknowledgement. Keep credentials out of the research catalogue. Verify ownership and authority before actual changes. A handoff template may be adequate; a new credential vault is outside this experiment.
Report delivery: a control copy proves arrival at the control mailbox, not delivery to every client or that anyone read it. Test missed schedules and stale data separately. Start with a register of expected reports before writing advertising connectors.
Small-business operations
Invoice follow-up: use synthetic paid, partially paid, disputed and overdue invoices. Payment matching can lag. Require status review before sending a reminder and compare built-in invoicing reminders. Sending messages is a separate authorized action, not a research test.
Manually arranged appointments: check an agreed time slot, payment-link expiry, successful payment, cancellation, reschedule and duplicate callbacks in a payment sandbox. Compare existing scheduler workflows. A public online-booking deposit feature may not cover staff-created appointments; verify the precise platform/region and avoid promising universal support.
Certificate renewals: test date formats, bulk import, ownership changes, reminder delivery and export. This is administrative tracking, not verification of authenticity or legal eligibility. Compare the customer's calendar, HRIS and existing reminder tools first.
What to record
Use a simple experiment log: date, segment, last real incident, current solution, bounded offer, price, objection, payment status, setup hours, recurring support and outcome. Keep sensitive customer data in the customer's authorized environment, not in published research files.
After five suitable conversations without a willingness to pay for a concrete result, change the segment or the offer. Stop when a short native configuration solves the task, support exceeds the price, or the proposed product cannot safely distinguish its critical failure cases. These thresholds are internal decision rules, not statistical proof.