Cold-Start Instance: Salem (Cherry City / MG10)

The market-specific parameters for the Cold-Start Fill Playbook. The playbook holds the method; this holds the dates, numbers, and named specifics for Salem.

Situation. Cherry City is not a fresh cold-start. It opened in October 2025 and grew to 28 of 49 occupied (57%) through two move-in batches (14 at the October launch, 10 in March); the organic months between and since are thin, and move-ins have now fallen to roughly zero (2 in May, 0 so far in June) while occupancy slips from a May peak of 30. It is a stalled lease-up in recovery, and the diagnosis below (validated against the occupancy view) shows the binding constraint is the front door: move-in flow has dried up. Churn is real but secondary, lumpy (spikes in February and April) and averaging ~1.4/month, not a steady bleed.

Parameters

Parameter	Value
Market / building	MG10 Cherry City, Salem, OR
Timing	post-open (~8 months in); stalled lease-up in recovery
Open date	2025-11 (presales from 2025-10)
Market depth / tier	C-tier, low-volume (Salem); LTV and absolute demand smaller than Portland
Brand-cold baseline	Cold. No scene reputation, no waitlist of consequence (2 as of June).
Monthly budget envelope	$2,000–3,000 / month
Member LTV (this market)	Use Portland ~$16K until measured; likely lower given C-tier
Target occupancy + date	90% (44 of 49 lockouts) ASAP; no fixed refinance deadline

The number (back-calc)

Source: analytics.occupancy_monthly, location MG10 Cherry City, as of 2026-06 / 2026-07.

Step	Value	Source
Total lockouts	49	resources (STUDIO_MONTHLY)
Current occupancy	57.1% (28 of 49)	occupancy view
Studios left to 90% target	16 net (to 44 occupied)	44 − 28
Net move-ins / month required	rate to set (see scenarios)	gap ÷ months; ASAP = pick the pace
Gross move-ins / month required	net + ~1.4 churn	move-outs avg ~1.4/month, lumpy (see diagnosis)
Tour close rate	~80% (trusted)	Matador's direct report
Tours / month required	gross ÷ ~0.80	from the close rate
Qualified leads / month required	directional only	tours ÷ lead-to-tour; tours still undercounted

Pace scenarios (net gap 16; gross must also offset ~1.4 move-outs/month):

Net move-ins / month	Gross / month (+~1.4 churn)	Tours / month (÷0.80)	Months to 90%
2	~3.4	~4	~8
3	~4.4	~6	~5.5
4	~5.4	~7	~4

The targets are modest in absolute terms (mid-single-digit tours a month), as expected for a C-tier market, which shapes the experiment design: at this volume use decision rules and step-changes, not statistical significance. The immediate gap is that move-ins have fallen to ~0 while tours still run 3–6/month, so the first job is to restore move-in flow (the front door); the February and April churn spikes are a secondary thread to understand, not a steady leak to plug.

On the lead row. Lead-to-tour is only directional and will stay that way for the past. Matador's tour backfill is about as complete as it will get and cannot reach reliably back many months, so historical tour counts and the close rate are unrecoverable; tours ≥ move-ins is the only floor we trust. The ~80% close is Matador's anecdote, not from the DB. Funnel math above the tour line becomes trustworthy only going forward, if tours are logged reliably from here.

Diagnosis (2026-06-23, from production, validated)

Flow from inception, mirroring the analytics.occupancy_monthly definitions so it reconciles: a move-in is dated by COALESCE(start_time, billing_start_date), a move-out by COALESCE(cancel_at_period_end, cancelled_at), counting only MONTHLY reservations with a stripe_subscription_id and a recurring payment. Recomputing occupancy from this flow matches the view's occupied_count exactly every month, so the move-in and move-out counts are validated. Tours (type='TOUR') are unreliable historically: the backfill cannot reach reliably back many months, so treat move-ins as the tour floor (a tour is required for every lockout move-in, so true tours ≥ move-ins).

Month	Tours (rec.)	Move-ins	Move-outs	Net	Occupied
2025-10	0	14	0	+14	14
2025-11	0	2	0	+2	16
2025-12	0	4	0	+4	20
2026-01	3	1	1	0	20
2026-02	4	3	5	−2	18
2026-03	3	10	0	+10	28
2026-04	6	5	4	+1	29
2026-05	5	2	1	+1	30
2026-06*	4	0	2	−2	28
Total	25	41	13	+28

*June is partial (through the 23rd).

What it says:

Growth came in two event-driven batches, never a steady engine. The October launch put in 14, all from the pre-sale waitlist; March added 10, the February ad cohort landing on a March 1 start. Those two months are 24 of the 28 current members. Neither was organic, steady demand: one was the launch waitlist, the other a paid push. Between and after, with the waitlist spent and ads off, move-ins are thin to zero. There has never been a self-sustaining demand engine here, which is exactly what the lane call has to build.
The binding constraint now is the front door. Move-ins fell to 2 in May and 0 so far in June. The May-to-June slip (30 to 28) is zero new move-ins plus two departures, not a churn surge. Because a tour is required for every move-in, tours fell with move-ins; the recorded tour series is too undercounted to read as a separate signal.
Churn is real but secondary and lumpy. Thirteen move-outs over nine months is ~1.4/month, concentrated in two spikes (February 5, April 4); most months lose 0 to 2. There is no steady high-rate bleed.
Recorded tours are unreliable, historically and for good. Matador's backfill is about as complete as it will get and cannot reach reliably back many months. Since a tour precedes every move-in, true tours ≥ move-ins each month, so move-ins are the tour floor; we cannot recover how many more tours there were, so the historical close rate is unknowable. The ~80% close is Matador's anecdote, not measurable from the DB, and becomes measurable only going forward if tours are logged reliably from here.

Top of funnel: what changed in the ads (the move-in answer)

Ad spend and structure (Meta metaads.*, Google googleads.* warehouse) explain the move-in collapse, and the answer is counterintuitive: spend went up, not down, but moved off the channel that converts.

Month	Google "Salem" search	Meta Cherry City	Meta dominant objective
Feb	$551	$606	link-click Salem promos
Mar	$0	$0	—
Apr	$0	$1,009	Leads + Sales
May	$27	$2,679	Traffic (new) + Leads
Jun	$72	$2,172	Traffic + Leads

High-intent search delivered in February, then went dark, then only trickled. Daily delivery data (verified 2026-06-25) shows three phases for the Google "Salem" search campaign: February ran hard (5th–22nd, $551, ~29,000 impressions, 507 clicks); March and April were completely dark (zero spend, zero impressions, no reach at all); May and June came back but barely (54 then 201 impressions for the whole month, $27 then $72), the bid-strategy throttle later found in the account. So the earlier "switched off" framing was an inference from near-zero spend: in March/April that was right in effect (no reach), but May/June it was active and merely starved, not off. February's search push (plus the Meta Salem link-click promos) is the likely source of March's batch of 10 move-ins. The link is tighter than timing alone: nine of the ten March move-ins were created (signed) in February (Feb 5–27) on a March 1 start, so the batch is the February cohort landing on a first-of-month billing date, squarely inside the February spend window. With delivery near-zero from late February on, the pipeline had nothing upstream.
Why it throttled: bid strategy, not budget (verified 2026-06-25). The campaign was bid-limited, not budget-limited, spending only ~$6.75 of a $27.50/day budget (~25%). The cause was conversion-based smart bidding (Maximize Conversions, with a $16 Target CPA) on a campaign that records only ~1-5 conversions/month, far below the ~15-30 smart bidding needs, and $16 sat well below the actual cost per conversion (Feb ~$110, June ~$144, blended ~$69). With an impossible target and almost no conversion signal, the algorithm bid down to near-zero and stopped winning auctions. Fix in progress: switched to Maximize Clicks with a ~$3-4 max-CPC cap to buy volume first; budget held at $27.50 until it actually spends it. The open question for the coming week is whether daily spend now climbs (bidding was the limit) or stalls below budget (Salem search demand is the limit). Conversion-based bidding (Target CPA, Maximize Conversions, new-customer value bidding) is premature until conversion volume builds.
Meta spend ramped onto the wrong objective. ~$1k April → ~$2.7k May → ~$2.2k June. On May 6 the objective-aligned April campaigns (Monthly on OUTCOME_LEADS, Hourly on OUTCOME_SALES) were paused for "Cherry City – Unified Traffic" (OUTCOME_TRAFFIC), which became the single biggest line. A traffic objective buys the cheapest clicks and landing-page views, not leads or buyers. A smaller OUTCOME_LEADS campaign returned May 20.
Net: spend roughly quadrupled while move-ins fell to zero, because it shifted from high-intent search to low-intent Meta traffic.
Caveats: these are platform-reported conversions, not move-ins traced to source (the attribution gap); move-ins lag and June is partial. Recorded tours are too undercounted to use as a counter-signal (a tour precedes every move-in, so tours fell with move-ins); the clean fact is that move-ins collapsed as high-intent search went dark.

This, not churn, is the move-in answer. The February/April churn spikes remain a separate, secondary thread for a Matador conversation.

Lane call (this market)

The Four Fits point one way for a toured, high-consideration product, and the ad history confirms it empirically: high-intent search is the engine, and it was working in February before it was turned off. Commit to it; do not spread the small envelope across channels.

Primary (funded first): Google high-intent search. Make the "Cherry City - Salem Search" campaign actually deliver as the committed primary (fix the bid strategy and fund it, not just toggle it on; it was active but starved May–June). It matches a high-consideration, toured product (bands actively searching for a place to practice), and February's search spend, when it last delivered, preceded the only real move-in batch we have. Fund this to the top of the $2–3k envelope.
Secondary: Meta, fixed to a leads/conversions objective. Keep Meta, but move budget off "Cherry City – Unified Traffic" (OUTCOME_TRAFFIC) onto a leads or conversions objective. Traffic buys clicks we cannot convert; leads at least optimize toward an inquiry. Cap it below search.
On-ramp (the first visit): tour is the proven path; free hours (comp mechanism) is the experiment to run against it (Layer 4).
Closer: Matador (community manager, on-site).
Second act (parallel, not the engine): community and scene-building from day one; referral give-50/get-50, untested and worth a real test, not reliance.

It is reallocation, not new money. Meta alone is already running ~$2.2k/month, roughly the whole envelope, on a traffic objective. The call is to redirect the existing spend: search primary, Meta-leads secondary, kill the traffic line. No budget increase is required to make it.

First thing to watch after the switch: whether restoring high-intent search lifts move-ins within the lag window. Move-ins are the reliable signal; tours are readable only if Matador logs them reliably from here.

First experiments (Layer 4)

Two experiments, sequenced. At Salem's volume we use decision rules and step-changes, not statistical significance (see the learning system). Log both in the experiment log there.

Experiment 1 — Restore high-intent search; move Meta off traffic

The binding-constraint experiment, and a correction we hold with high prior given the February precedent.

Hypothesis: restoring the Google "Salem" high-intent search campaign as primary, and moving Meta off OUTCOME_TRAFFIC onto a leads/conversions objective, lifts qualified inquiries, tours, and move-ins. The only real move-in batch we have followed the February search spend.
Lever: channel and objective reallocation within the existing $2–3k/month envelope (no new money).
Input it targets: first visits booked (and qualified leads upstream).
ICE: Impact high (sits on the binding constraint), Confidence high (February precedent), Ease high (reallocation) → run first.
Metric and threshold: lockout move-ins/month (the reliable signal); tours/month too if logged reliably from here. Step-change to watch, allowing search-to-signing lag (~4–8 weeks): move-ins back to ~3+/month.
Decision rule: if move-ins step up inside the window, keep funding search as primary. If search is restored and they do not, the constraint is downstream (close rate or lag), not the channel, and the next experiment moves there.
Guardrails: junk-lead rate; CAC (generous, LTV is high).

Experiment 2 — On-ramp: tour vs free hours

Hypothesis: offering free hours (via the comp mechanism) as an alternative first visit increases first-visit volume and/or visit-to-move-in versus tour-only.
Lever: the on-ramp offer.
Input it targets: first visits booked; visit-to-move-in.
ICE: Impact medium-high, Confidence medium (untested), Ease medium (needs comp setup and on-voice copy) → run after Experiment 1 has restored readable volume.
Metric and threshold: first visits and move-ins by on-ramp type; directional step-change.
Decision rule: keep whichever on-ramp produces more move-ins per dollar; they may co-exist.
Note: do not launch simultaneously with Experiment 1 if it would confound the read; sequence it once search volume is back.

Offer / hook

Free hours via the comp mechanism (the MMS free-practice mechanism), framed as "free hours," tested against "free practice." Founding-member / recovery rate: to define. Keep all copy on-voice (plain; see 1-foundation/brand.md).

Local partners

To identify: Salem-area music shops, gear retailers, venues, all-ages spaces, schools and music programs, and any active scene nodes. (Gap.)

Readiness checklist

[ ] Instrumentation live — partial: Matador has backfilled some DM-booked tours but the backfill is incomplete; recorded tour count still understates real volume. Going-forward capture is live as of 2026-06-25 (PR 852 + 853): a required CM lead-source dropdown on staff tours, an ad-spend/lead sync cron, and the cold_start_scorecard view (funnel plus ad columns). Ad history is backfilled and the cron keeps it current; the one ongoing manual dependency is Matador's lead-source entry on new tours. Tour-to-channel still fills going forward only, not for past tours.
[x] Scorecard settled and standing (weekly; canonical in 4-learning-system/scorecard.md, baseline logged in runs/)
[x] First experiments defined (Experiment 1 ad reallocation; Experiment 2 on-ramp)
[x] Move-in collapse diagnosed — caused by the ad shift (high-intent search switched off, Meta moved to a traffic objective); fix is the lane call / Experiment 1
[~] Lane call executing (Aaron, who runs the accounts): Meta OUTCOME_TRAFFIC campaign turned off [done]; Google search delivery being fixed [in progress] — found bid-limited (spending ~$6.75 of $27.50/day) under a conversion-based bid strategy that starved it on too-few conversions (~1-5/mo); switched to Maximize Clicks with a ~$3-4 max-CPC cap on 2026-06-25. Hold budget until it spends what it has; watch whether spend climbs, then move-ins over the ~4-8 week lag.
[ ] Founding-member / recovery offer defined
[ ] Free-hours launch planned
[ ] Local partnerships identified
[ ] Tour and follow-up process staffed (Matador; confirm DM inquiries are worked, not left sitting)
[ ] Tour logging reliable from here (historical backfill is about as complete as it will get; close rate measurable only going forward)
[ ] Churn spikes understood — Feb (5) and Apr (4); secondary, lumpy (~1.4/month avg). In motion: Matador is filling cancellation notes on all cancelled Cherry City reservations (enabled by #845); pull and characterize the 13 move-outs once done.

Scorecard

Settled and standing, read weekly. The canonical rows, sources, queries, and the middle-gap reality live in 4-learning-system/scorecard.md; the baseline (row-zero) snapshot is runs/2026-06-25-salem-scorecard.md.

Baseline at a glance (June): occupancy 57% (28/49); move-ins 0; Meta spend ~$2,172 (partial-month at the 2026-06-25 snapshot; full June ~$2,343 per the now-live view) vs Google "Salem" search ~$72 (the inverse of the lane call, mid-correction); Meta Instant Form leads 33 (63 since February) but untraceable to move-ins.

Google gridding (resolved 2026-06-25): the Google search campaign was originally named "Salem", which matched neither "Cherry City" nor "MG10", so its spend sat unallocated and the card read google_spend = 0. It was renamed to "Cherry City - Salem Search" and its 42 historical rows relabeled, so the card now shows Google on Cherry City's rows (Feb $551, May $27, June $72). This confirms the lane-call story directly from the card: February's $551 Google search push, with Meta near zero, preceded March's batch of 10 move-ins. Keep new campaigns named with the location name or mgId so they grid automatically.

The middle's limit is linkage, not volume. Lead volume is captured (Meta lead action back to February; on-site inquiries since March), but tour-to-channel is dark (1 of ~25 recorded tours attributed) and move-in-to-channel is near-empty (3 MONTHLY attributions ever), so a channel's leads cannot be traced through to a signed lockout. Spend, move-ins, and occupancy are trustworthy today; the middle is directional until that linkage closes. The fix is live as of 2026-06-25 (PR 852, CM lead-source on staff tours, last-fallback) and closes the linkage going forward, not for past tours, so expect the verdict to firm up over the coming weeks of new CM-entered tours, not retroactively.