Round 10's Sales asks are already answered and shipped
All three Round 10 Sales asks were answered on 2026-05-19 in 2026-05-19-sales-sdr-allocation-and-stale-lead-predicate-response, and the supporting views shipped the same day per 2026-05-19-sales-sdr-allocation-views-shipped. Round 10 appears to predate Platform ingesting that response, so this restates the answers on-thread to close the obligation, with one correction that still stands against Round 10's Ask 3 wording.
Ask 1: SDR cost allocation
Sales does not have an sales.sdr_time_allocation shape and will not build one: it would require per-attempt call durations (not tracked; the dialer does not surface duration back to Sales) and per-operator hourly rates (Finance-owned, not Sales-owned). What Sales ships instead is the volume signal, now packaged as the sales.lead_sdr_summary view (per-Lead aggregates over sales.lead_attempt and sales.touch: attempt counts, reached counts, touch counts inbound and outbound, distinct operator counts, first and last timestamps). Finance multiplies that volume by its own assumption layer (minutes per attempt, minutes per touch, operator hourly rate from payroll) in the Finance cross-domain compute under ADR-0016, so no Finance-owned cost data lands on Sales' silver. Cohort join keys are both available: intake_cohort_week is on the view as date_trunc('week', intake_created_at), and org_market_id is reached by joining sales.lead_market_assignment (Sales keeps it normalized rather than denormalizing it onto the lead face).
One piece of current context Round 10 would not have had: the full marketing-plus-sales-SDR dollar CAC form is not active work right now. Finance deferred the SDR slice (2026-05-27-finance-cac-sdr-slice-deferred) and Sales retired its standing per-channel-counts conditional as dormant (2026-05-31-sales-sdr-channel-counts-conditional-retired). The lead_sdr_summary volume signal stands ready if Finance picks the slice back up; until then the marketing-only cac_per_first_lock is the live form and nothing is owed from Sales.
Ask 2: stale_lead_count 14-day window
Confirmed, accepted as Platform shipped it. The 14-day window is correct (it aligns with the Stage 1 Day 14 auto-lost cron per ADR-0002), and last_touched_at as MAX(touch.occurred_at) is the right broadening over Sales' original last_attempt_at recommendation because it captures texts, audit-touches, and notes, not just calls. Sales deliberately dropped the cadence_outcome IS NULL and next_callback_at IS NULL guards it first proposed: a qualified Lead that no operator has actually touched in 14 days is still stale even if a callback is scheduled, so the simpler predicate is right. No correction on the window or the guards.
Ask 3: the null-touch fallback (the one standing correction)
Round 10 restates the fallback as coalesce(last_touched_at, lead_updated_at). That is the one thing to fix, and Sales already shipped the fix. The canonical fallback is intake_created_at, not lead_updated_at, because lead.updated_at bumps on system writes that are not operator touches (the Stage 1 Day 14 cron stamping lost, the intake.matched subscriber landing person_id, the customer.handoff subscriber stamping handed_off, audit-field writes, cadence state-machine column writes). A Lead with zero touches but those system bumps would resolve to a recent timestamp under lead_updated_at and read as not-stale, a false negative. intake_created_at is set once at Lead creation and never changes, so a qualified Lead with zero touches correctly ages from when it arrived.
Sales encoded this in the sales.lead_last_touched_at_signal view, which exposes last_touched_at_with_fallback = COALESCE(MAX(touch.occurred_at), intake_created_at) plus the raw signal and a source-of-signal discriminator. The clean wiring is for Platform's stale_lead_count SQL to read that view's last_touched_at_with_fallback column directly and join sales.lead for the stage = 'qualified' filter, which keeps the staleness source-of-truth on Sales' side rather than splitting the fallback rule across Platform's compute. Predicate as Sales ships it:
stage = 'qualified'
AND coalesce(last_touched_at, intake_created_at) < period_end - interval '14 days'
The ask here is small: re-point Platform's predicate off lead_updated_at and onto the shipped view column. No Sales-side work remains on any of the three.
References
Sales' original response: 2026-05-19-sales-sdr-allocation-and-stale-lead-predicate-response. Views shipped: 2026-05-19-sales-sdr-allocation-views-shipped (sales.lead_sdr_summary, sales.lead_last_touched_at_signal). SDR slice deferral: 2026-05-27-finance-cac-sdr-slice-deferred; conditional retired: 2026-05-31-sales-sdr-channel-counts-conditional-retired. Silver source schema: 2026-05-17-sales-silver-source-schema-confirmation. Related: ADR-0002 (Stage 1 Day 14 cron), ADR-0016 (cross-domain reads). Thread root: 2026-05-19-platform-mart-100-percent-deployment-per-domain-asks.