• value (needs that actually reach you): structured form with a stated posture > registration-flow field > named inbox with named people
• compliance cost: registration-flow field > structured form with a stated posture > named inbox with named people
• efficiency: registration-flow field > named inbox with named people > structured form with a stated posture

Value and effort run in one order, so no pair strictly dominates: the instrument that surfaces more needs costs strictly more triage. The field outranks the inbox on effort even at ten minutes to add, because effort here means the months of triage it generates, not the setup - asking in the flow multiplies who discloses, which is exactly what makes it valuable and exactly what makes it work.

The compliance axis disagrees with the other two, which is the point of splitting them: the rung that wins on efficiency also tops compliance cost. A registration-flow field deposits health and disability data into a third-party ticketing platform's records by default, where retention is the platform's choice, not yours. The structured form triggers a heavier review upfront - a retention rule, a named owner, a published turnaround - and that review is what makes the data deletable again.

• Registration-flow field (efficiency default) - one question in the flow, routed to a named owner. Surfaces needs a passive inbox never hears. Add a retention rule stated on the form, and delete at the debrief.
• Named inbox with named people - a dedicated address answered by named individuals, not a role alias. Promote over the field when Q6 leaves the jurisdiction's data rules unanswered.
• Structured form with a stated posture (starved rung) - tops value (people disclose when they see what happens to it) and effort (posture, owner, turnaround all written and honoured). Promote when requests already arrive in volume, an institutional partner requires it, or Q6 found a statutory duty - route that to counsel first.

If Q9 left nobody to answer requests through to the event, delete all three and design every provision to work walk-up instead.

Deleted rather than ranked at any staffing level: routing requests to the general registration desk with no accessibility owner. It is not the cheap rung, it is the unowned one: the request is received, nobody is accountable, and the attendee learns this on the day.

See references/accessibility-page-and-request-flow.md for the request-channel copy and the five-point data-handling posture to decide before the channel opens.

Before Menu 4: the slot itself

Childcare is a provision to fund. The time slot decides whether caregivers need it at all, and it costs nothing to change while the date is still open.

• Organizers report that evening events lose women mostly through childcare, and that experienced practitioners also skip evenings for family reasons.
• The same reports describe women's share rising with daytime or lunch slots, and with broadly uniting topics: a product-themed conference, a workshop co-run with an HR meetup.
• Read this as a cheaper move than any rung below, not as a substitute for them: a full-day conference still needs childcare, and some audiences can only come in the evening.

Hand the slot question to samber/dev-event-organizer-skills@event-date-selection (§ Time of day) while it is movable. Hand topic breadth and co-hosting with a non-technical community to samber/dev-event-organizer-skills@event-format-selection. Once the slot is fixed in the evening, Menu 4's childcare rungs are the remaining lever.

Menu 4 - Economic access

Ticket price itself is not yours: the need-gated scholarship rung belongs to samber/dev-event-organizer-skills@event-ticket-pricing, which owns its application mechanic and reviewer. This menu is what access costs beyond the ticket.

Ranking (default, not a law - re-rank against Q1, Q3, Q4):

• effort (organizer hours, vendor coordination, payment/eligibility handling, reversibility): travel and registration grant > subsidized childcare > complimentary childcare > the ticket-side need-gated rung
• value: travel and registration grant > complimentary childcare > the ticket-side need-gated rung > subsidized childcare
• compliance cost: complimentary childcare == subsidized childcare > travel and registration grant > the ticket-side need-gated rung
• efficiency: the ticket-side need-gated rung > complimentary childcare > subsidized childcare > travel and registration grant

The two childcare rungs tie on compliance cost only: both place minors in the event's care (vendor vetting, staffing ratios, insurance, background checks), the heaviest and least reversible review on this menu since an incident can't be undone. The grant's review is financial - identity confirmation, receipts, disbursement, possible tax reporting - heavy but reversible edition to edition.

They do not tie on effort. A subsidized fee looks cheaper and is not: see references/economic-access-and-childcare.md for the sourced cost ratio and the vendor due-diligence checklist. Small budget relief, real administrative load - that is why subsidized loses to complimentary on three axes and is still the right answer when the budget line, not organizer hours, is what binds.

• The ticket-side need-gated rung (efficiency default) - point at the sibling, ask the interview question, spend nothing here. Broadest reach, smallest cost removed.
• Complimentary childcare - onsite, free, RSVP required. Unblocks an otherwise fully excluded population. Promote when Q3 shows a caregiving population in the audience.
• Subsidized childcare - the same provision behind a published flat fee against real per-child-day cost. Last on value - any fee filters out the families the program was for. Promote when the budget can't absorb the full cost.
• Travel and registration grant (starved rung) - tops value (travel is the largest excluded cost outside the host city) and effort (an application cycle, review panel, disbursement). Name its access limit: receipts-only disbursement excludes anyone who can't front the cost. Delete outright, not rank, when Q4 found no entity able to disburse money.

Deleted rather than ranked: an unpublished "email us and we'll see" hardship promise with no budget behind it.

The people attendees turn to

Menu 3 shrinks what needs a written request; this is its human counterpart. At a large conference, make a team trained in handling inequality - discrimination, exclusion, unequal treatment, an access need nobody anticipated - physically identifiable for the whole event, a distinct t-shirt being the usual mark. An attendee then reaches the right person without first explaining to a stranger why they need help.

Scope it by Q11, not by ambition:

• Large conference (multi-day, or big enough that attendees cannot see the organizers): staff the dedicated team, on every day and floor the event uses.
• Meetup or small event: skip the dedicated team. The organizers are the contact - name them and point at them in the opening, and say the same on the event page. Matching shirts at a single-room evening add a uniform, not reach.

Three conditions make the team exist rather than just wear a colour:

• Announce it before the event or at the start: event page, accessibility page, the pre-event attendee email, and the opening remarks - that the team exists, what it is for, and how to recognize it. A mark nobody was told about is decoration.
• Brief before identifying. An identifiable team without training is worse than none: the attendee who finally asks gets a shrug and stops asking. Minimum briefing: every provision and where it is, how to listen without demanding justification, what a member may resolve alone, and the handoff path - provision owner, organizer on duty, code-of-conduct response team. Recruit and brief through samber/dev-event-organizer-skills@event-volunteers.
• Give it a handoff, not a desk of its own. Every member knows who owns each provision and how to reach the organizer on duty; the team routes, it does not hold requests nobody else sees.

Not the code-of-conduct response team

samber/dev-event-organizer-skills@event-code-of-conduct already makes physically identifiable on-site responders the default reporting channel for incidents, and owns that ranking. The inclusion team is broader and earlier: proactive help with access and unequal-treatment friction, most of which never becomes a report. Its small-event answer does not touch the response team, which that skill staffs at any size.

• Can share: people (a trained responder can also serve here), one briefing session, and the shirt. One "ask us" mark is easier to announce than two colours attendees must decode - a default, not a law.
• Must not be conflated: intake. A member who hears a code-of-conduct violation hands it to the response team under that skill's runbook and confidentiality promise; they do not investigate, decide or record it themselves. A shared shirt therefore requires every wearer to have that skill's staff briefing - a reporter cannot tell which wearer may take a report.
• Must not be conflated: purpose. If only responders wear the mark, attendees read it as "for harassment only" and bring no access problem to it. When the mark is shared, the opening announcement states both purposes.

The digital layer

Web accessibility is a separate discipline and transfers here only to the event's own digital surface - the website, registration and CFP forms, the schedule, the livestream player, and published program materials. It says nothing about physical or sensory provisions on site, and must never be presented as covering them.

Target WCAG 2.2 level AA on those surfaces, structured by the four POUR categories, with captions on published media and contrast minimums met on the schedule and forms. The checklist and the audit workflow are in references/digital-layer-checklist.md. A perfect automated score is not conformance; keyboard-exercise the registration flow before claiming anything.

Workflow

1. Run the interview. Fix the funding pole, venue status, known needs, entity, owner, effort ceiling.
2. Decide the provision set from menus 1, 2, and 4, plus the dietary commitment - before the venue is signed.
3. Write the accessibility criteria sheet and hand it to event-venue-sourcing as a sourcing input; it checks candidates against your sheet.
4. Declare the set publicly on one page, provision by provision, with what is covered and what isn't: references/accessibility-page-and-request-flow.md.
5. Open the request channel chosen in Menu 3, with owner, lead time, and retention rule stated on the form. Answer every request, including declines.
6. Route delivery: captioning to event-production, with the captioner's jargon brief sent ahead by the accessibility owner; dietary commitments to event-vendor-sourcing; a speaker's declared need to event-speaker-experience; placement and wayfinding to event-attendee-experience.
7. Staff it and say it on the day: name who owns each provision, and open with what exists and where. At a large conference, brief and mark the inclusion team, and announce how to recognize it; at a small event, the organizers name themselves as the contact.
8. Review after the event: requested, delivered, declined and why. Delete the request data on the schedule published.

If your harness has persistent memory, record the provision set as published, every request and how it was answered, the venue's real access facts, and which provisions went unused. Next edition provisions against that record instead of a template.

Stated gaps

Say these out loud rather than inventing an answer:

• Dietary and childcare practice each now draw on two independent organizations rather than one, and the catalog states the range where the two sources diverge - see references/provision-catalog.md and references/economic-access-and-childcare.md.
• Quiet-space sizing has no established seat-to-attendee ratio or square-footage rule anywhere sourced. One unconference series reports an operational observation - roughly 5% of attendees in the quiet room at any one time - but states it as its own experience, not a codified standard; treat any stricter-sounding figure seen elsewhere as a vendor estimate. Decide sizing against the room you have and say it was self-set.
• The trigger threshold for interpretation is self-set: nothing fixes the attendance or evidence level at which on-request provision stops being enough, so state the level you chose and why.
• Jurisdictional duties are out of scope by construction. Route Q6 to counsel.

Failure modes

• Announcing a provision nobody is staffed to run. A quiet room that is locked, a nursing room whose location nobody at the desk knows, a captioning line with no operator. Name the owner in step 7 before the provision reaches the page.
• Collecting disability disclosures with no handling posture. The request channel opens, disclosures arrive, and nobody has decided where they live, who reads them, or when they are deleted. Decide it before the channel opens, not after the first request.
• Treating the templated full page as a floor at community budget. It is the top of a range most published events do not reach. Copying it produces a page of promises the budget cannot fund.
• Captioning part of the programme without saying which part. Sourced practice scopes captioning to the highest-attendance days; the failure is not the scope, it is discovering the scope in the room.
• Deciding accessibility after the venue is signed. Almost every physical provision is a property of the building. Once the contract is signed, the menu has already been chosen for you.
• Reading production capability as coverage. A capable A/V team, or an automatic caption feed, is not a captioning provision until someone states, budgets and publishes it, with the automatic feed labelled as such.
• Sending the captioner in cold. A human captioner given no programme or glossary mis-renders product names, acronyms and library names all day. Fix: the accessibility owner sends the pack before the event (§ Menu 2).
• Marking a team nobody briefed. Shirts go on without training, an attendee asks about a provision or an unfair moment, and gets a shrug. Fix: brief before the mark goes on, with a written handoff path (§ The people attendees turn to).
• Letting a request channel substitute for walk-up design. Every provision that requires disclosure filters out the people least willing to disclose. Shrink the set that needs asking first.
• Funding childcare for an evening slot nobody questioned. The slot was the cheaper lever. Fix: ask Q10 before Menu 4, while the date is still movable.
• Publishing a grant program without its access limit. A reimbursement-only grant excludes anyone who cannot front the cost. Say so on the page rather than letting the applicant discover it.

Measurement

Everything below is self-set - pick two or three and write down before the event what each would change:

• Requests received against requests met (self-set): the ratio, and the reason for every unmet one. A channel with zero requests is not evidence of zero need; it is usually evidence the page was published too late or the channel was not trusted.
• Provisions used against provisions funded (self-set): a quiet space nobody entered and a childcare booking nobody took are different findings from an over-subscribed one, and both belong in the debrief.
• Lead time from page publication to the travel-booking window (self-set): a page published after attendees have booked reached nobody it was for.
• Composition by time slot (self-set, only where gender or caregiver status is collected with consent): share of women and of caregivers per slot type across editions.
• Requests carried forward (self-set): needs that recur edition to edition should become walk-up provisions rather than repeat requests.

Invocation examples

• "We're running a 400-person community conference in a university building - what accessibility should we actually commit to, and what should we not promise?"
• "Should we caption our talks? We have a good video team and no captioning budget."
• "Can we use automatic captions instead of paying a stenographer for our conference?"
• "Someone asked for a sign-language interpreter three weeks out. What do we do, and what should we have published?"
• "Write our accessibility page. We have a quiet room, step-free access and no childcare."
• "Almost no women come to our evening meetup. Should we fund childcare?"

Expected output: a provision set and the request channel, presented section by section for validation, every item labelled sourced or self-set.

• Provision set - each provision with its depth rung, who owns it, what it costs in effort, which sibling delivers it, and what the page will say about it.
• Request channel - its posture.

References

• references/provision-catalog.md - every sourced provision with its origin and date: the templated full page item by item, the historical 2019 page, FOSDEM's thin page including its stated absences, human real-time captioning with the captioner's jargon brief and the limits of automatic captions, and the dietary categories with their fallback paths.
• references/accessibility-page-and-request-flow.md - how to structure the page, how to word a provision and its scope honestly, the request-channel copy, the data-handling posture, and what to announce on the day.
• references/economic-access-and-childcare.md - the childcare vendor due-diligence checklist with its sourced staffing ratios and vetting standards, the subsidy ratio, and the grant program's mechanics including its access limit.
• references/digital-layer-checklist.md - the WCAG 2.2 AA checklist for the event's own website, forms, schedule and livestream, labelled digital-layer-only throughout.

Sibling skills referenced throughout: event-venue-sourcing, event-production, event-vendor-sourcing, event-attendee-experience, event-speaker-experience, event-code-of-conduct, event-volunteers.