Skip to content

Marketing System

This is the architectural map of Metrognome's marketing. It organizes everything we do into five layers, each grounded in an established framework, and it shows where each piece lives today, how mature it is, and what is still missing. We are building this system deliberately, one component at a time, rather than trying to produce a single perfect plan in one shot.

The cold-start fill playbook is one path through this system, not the system itself. It is the cold-start version of Layer 2 and Layer 3, instantiated through Layer 5. When you see a reference to "the playbook" below, that is what it means.

How to read this

Each layer answers one question and holds a small number of components. For every component the tables below record three things: what it is, where its content lives today (an existing doc, or "no home yet" when it is a gap), and how mature it is. Maturity uses these labels:

  • Trusted: tested with proven outcomes.
  • Complete: exists and is testable.
  • Partial: exists but incomplete or scattered.
  • Stub: a placeholder file describing what goes there.
  • Gap: nothing written yet.

Two modes

The single most useful distinction in the whole system is the difference between filling a new market and growing a full one. They are not the same job and they do not use the same metric.

  • Cold-start (the funnel). Filling the first building in a new market to roughly 90% occupancy, which is the threshold that unlocks refinancing. You cannot run a referral or community loop on a building that is mostly empty, so cold-start is a funnel: pull strangers in, get them to visit, close them. Salem and St. Louis are here.
  • Flywheel (the loop). Once a market is full, growth compounds through referral, community, and a waitlist that feeds the next building. This is the Reforge-style growth loop, where the output of one cohort becomes the input to the next. Portland is here.

Most of the system is shared across both modes. The places where they diverge are called out in Layer 2 and Layer 5.

Layer 1: Foundation

Question it answers: who we are, how we sound, who we serve, and through which channels this business can grow.

Two things ground this layer. First, the brand and voice model in brand.md: a message altitude ladder (Value, Trust, Identity, Belonging) crossed with multi-dimensional segments, spoken in one plain voice. Second, Brian Balfour's Four Fits: growth at scale needs four alignments, not just product-market fit, judged together. The two fits that matter most for our decisions are Product-Channel (a lockout is toured and high-consideration, so high-intent capture channels suit it and cold-interrupt channels do not) and Channel-Model (an LTV around $16,000, Portland-measured and not yet trusted in new markets, means we can tolerate high acquisition cost, so paid is structurally viable).

Component What it is Lives in Status
Brand and voice How we sound and present, conforming to the mg-voice skill 1-foundation/brand.md, mg-voice skill Partial
Market-Product fit Who we serve and what we sell against their need 1-foundation/icp.md, 1-foundation/positioning.md Partial
Product-Channel fit Which channels suit a toured, high-consideration product 1-foundation/README.md, 3-engine/channels.md Complete
Channel-Model fit Acquisition cost tolerance given LTV and pricing reference-ltv-portland-16000 (memory), 2-growth-model/goals-and-kpis.md Partial
Model-Market fit Whether market size supports the model docs/general/business-model.md, docs/general/strategy.md Complete

Detail and gaps: 1-foundation/

Layer 2: Growth Model

Question it answers: what is the one number that means we are winning, and what moves it?

Grounded in Reforge growth loops, Amplitude's North Star framework, and Sean Ellis. The North Star is the moment a customer gets the value. A small set of inputs (three to five levers we can actually move) drives it, and guardrails keep us from winning the North Star while breaking something else. The model takes a different shape in each mode: a funnel for cold-start, a loop for the flywheel, with a different North Star in each.

Component What it is Lives in Status
Cold-start North Star and inputs Move-ins per month on pace to fill the building 2-growth-model/README.md (defined), 2-growth-model/cold-start-fill-playbook.md, 2-growth-model/goals-and-kpis.md Complete
Flywheel North Star and inputs The compounding loop metric for full markets no home yet Gap
Guardrails What we will not sacrifice for growth 2-growth-model/guardrails.md Complete

Detail and gaps: 2-growth-model/

Layer 3: Engine

Question it answers: how does a stranger actually become a tenant, and through which committed channel?

Grounded in Balfour (products fit channels) and First Round (pick one lane and commit; the common failure is investing in too many channels and not funding any one enough). The engine is the motion from stranger to tenant, the single committed primary channel, and the creative that runs in it.

Component What it is Lives in Status
Acquisition motion The path: stranger, inquiry, visit, move-in, plus the on-ramp and the closer 3-engine/customer-journey.md, 3-engine/diagrams/ Complete
The committed lane The one primary channel we fund, with the others as secondary 3-engine/channels.md Complete
Channels and creative Per-channel strategy and the creative record 3-engine/channels.md, 3-engine/paid-meta.md (reconciled 2026-06-23), 3-engine/creative-log.md, 3-engine/mms2026.md (past campaign), marketing skills Partial

Detail and gaps: 3-engine/

Layer 4: Learning System

Question it answers: can we see what is working, and how do we run disciplined experiments instead of guesses?

Grounded in Sean Ellis (high-tempo, cross-functional experimentation). This is the layer that turns marketing from stabs in the dark into a structured practice. It has two halves: measurement (can we observe the truth) and experiments (how we test and decide).

Component What it is Lives in Status
Measurement and attribution The data we can trust, the known gaps, and the live scorecard 4-learning-system/scorecard.md (settled, weekly), 4-learning-system/runs/, 2-growth-model/goals-and-kpis.md Partial
Experiments The weekly loop, prioritization, decision rules, and a running log 4-learning-system/README.md (loop + log), 5-execution/cold-start-instances/salem.md (first experiments) Partial

Detail and gaps: 4-learning-system/

Layer 5: Execution

Question it answers: who does what, on what cadence, and how do we apply all of the above to a specific building?

Grounded in the multifamily lease-up discipline (the direct analog to filling a new building to stabilized occupancy by a financing deadline) and the Amazon PR/FAQ form for the alignment-facing documents. This layer is where the system meets a real market.

Component What it is Lives in Status
Cadence and roles The recurring sync (eventually a skill) and who owns what 5-execution/README.md, 4-learning-system/runs/, ../log.md Complete
Market instances Per-building application: target, velocity back-calc, current phase 5-execution/cold-start-instances/_template.md, 5-execution/cold-start-instances/salem.md Partial

Detail and gaps: 5-execution/

Immediate priority: Salem

The fastest path to action in Salem runs through four components, in order:

  1. Populate the Salem instance (Layer 5). Done: 5-execution/cold-start-instances/salem.md. Cherry City is a stalled lease-up in recovery, not a fresh cold-start: open ~8 months, 57% (28 of 49). The brief is a net gap of 16 move-ins to 90%, ASAP, on a $2–3k/month envelope. The diagnosis (validated against the occupancy view) shows the binding constraint is the front door: growth came in two batches (October launch, March) and move-ins have since fallen to ~0, while churn is real but secondary and lumpy (~1.4/month). The move-in collapse is explained by the ads: the Google "Salem" high-intent search campaign was switched off after February while Meta spend ramped onto a low-intent traffic objective (spend quadrupled, move-ins went to zero). The February/April churn spikes remain a secondary Matador conversation.
  2. Make the lane call (Layer 1 and Layer 3). Done in salem.md: Google high-intent search is the funded primary (turn the "Salem" campaign back on); Meta is secondary and must move off OUTCOME_TRAFFIC to a leads/conversions objective. It is a reallocation of the existing ~$2.2k/month, not new budget.
  3. Define the first experiment (Layer 4). Done in salem.md: Experiment 1 is the lane call itself (restore high-intent search, move Meta off traffic), judged on move-ins; Experiment 2 is the on-ramp (tour vs free hours), queued behind it. Method and log live in 4-learning-system/.
  4. Stand up a minimal scorecard (Layer 4). Defined in salem.md: spend, move-ins, and occupancy are trustworthy now; lead and tour rows stay directional because tour history is lost (the backfill is about as complete as it will get) and channel attribution, though now live (PR 852 + 853, 2026-06-25), only closes going forward. The close rate is measurable only going forward.

Filling those four also instantiates four of the system's components as a byproduct, which is the point of building it this way.

Observed drift (2026-06-23 audit)

Reading the existing docs against this structure surfaced four mismatches. They are recorded in the layer files too; collected here for visibility.

  • The instance template contradicted the playbook on lanes. The template listed community and referral as the primary lane and paid as mere ignition, the opposite of the playbook's settled thesis. Fixed: the template's lane section now matches the playbook (high-intent capture and targeted paid as the committed engine, community and referral as the parallel second act).
  • 3-engine/paid-meta.md reconciled (2026-06-23). It framed Meta as primary and the Traffic objective as current strategy; both now contradicted by the committed lane and the Salem diagnosis. Added a reconciliation callout, corrected the channel role and current campaign state, and flagged the contested "professional, not scrappy" lesson. Deep mechanics (attribution, audiences) left intact.
  • The LTV figure was inconsistent. goals-and-kpis.md carried ~$12,000 while the measured number is ~$16,000. Fixed: updated to ~$16,000 and flagged as Portland-only, not yet trusted in new markets.
  • channels.md is referenced but does not exist. Three docs point to a forthcoming channels.md. That is the Layer 3 committed-lane and channel-strategy gap; the docs already know it is missing.

Research foundation

Status of the skeleton

The architecture is settled and the layers are substantially fleshed out. Layer 1 (Foundation) has brand.md, icp.md, positioning.md, and glossary.md reconciled to the brand model, plus the four fits stated as a set. Layer 2 pins the cold-start North Star and inputs. Layer 3 records the committed lane (channels.md) with paid-meta.md reconciled to it. Layer 4 has the measurement, experiment loop, and a running log. Layer 5 carries the cadence and roles and a fully worked Salem instance.

The main remaining gaps are deliberately deferred, not overlooked: the flywheel North Star (for full markets, not the cold-start need), brand and voice (awaiting a focused work session), and the lease-up phase timeline. The brand/voice session and revisiting the playbook against this structure are the natural next phase.