Skip to main content
On the Work

Domain: Connection Strategy

The 4-week GTM plan the essay grounds — a working artifact pressure-tested live with DOMA.

A go-to-market plan for the Domain Beta launch and the 4-week run from ~400 to ~1,000 users. Self-contained working document, 2026-05-20. v6 — reopened to lead with on/in before the strategy machinery. Pressure-tested live with DOMA.

We conceptualize our visual field as a container and conceptualize what we see as being inside it. — George Lakoff & Mark Johnson, Metaphors We Live By, p. 30

What Domain is for

Domain is built for working on your work, not in your work.

Nobody is building for the second mode. The second mode is where most of the actual leverage in modern work lives. That's Domain's center of gravity.

The Lakoff epigraph isn't decorative. The "in" and "on" in working on your work, not in it are not stylistic prepositions — they are two of the small set of bodily-grounded orientational metaphors humans use to organize experience. "Working in" activates the container schema (you are inside the activity; the work fills your visual field). "Working on" inverts it (you are outside the container; the work has a shape, a boundary, a relationship to other work). These are different cognitive positions, not synonyms. The essay on this site develops the cognitive case at length; this document treats it as given and asks the operational question: if Domain serves the on-the-work mode, what does that mean for GTM?

The rest of the document answers that — and almost every conventional GTM tactic falls out of the picture once you take the on/in distinction seriously.

The two findings this strategy stands on

Two independent things turned out to be true under pressure-testing. Neither implies the other; the strategy needs both.

Finding A — Conviction about Domain is moment-bound. Users who experience Domain in an on-the-work moment develop real conviction — but that conviction doesn't survive the transition back into in-the-work execution. The user can't explain it to a peer later because they're no longer in the position from which they felt it. This decay is the failure mode behind the "I love it but can't get anyone to try it" pattern.

Finding B — The named GTM "channels" are stages of one pipeline, not competing options. Organic/WOM, content marketing, direct outreach, network partners, and waitlist look like five channels to choose between. For Domain specifically, they're the four stages of a single flow: Network → Waitlist → Activation → Referral. The waitlist is the network's exhaust; the activation is the rate-limiter; the referral is the compounding mechanic. Choosing among them is a category error.

Together: Finding A says every conversion mechanic must close before the user leaves the on-the-work moment. Finding B says budget concentrates on whichever pipeline stage is the bottleneck right now, not on a channel mix. These are the two load-bearing claims everything below derives from.

How this document is structured

Layer What it is What it answers
1. Connection strategy Diagnosis What is Domain for, where does conviction form, and how do the named channels actually relate to each other?
2. Tactical guardrail Guiding policy Given moment-bound conviction, what's the operational closing rule?
3. Budget allocation Coherent action — portfolio Given a sequential pipeline, where does a small team's attention concentrate?
4. Channel constraints Coherent action — tactical What does each channel do at its pipeline stage, and what must it not do given moment decay?

The bridges fan out from Layer 1's two findings and converge at Layer 4:


Layer 1 — Connection strategy: the diagnosis (expanded)

The "What Domain is for" section above established the frame. This section develops each finding in detail.

Finding A — Conviction about Domain is moment-bound

Conviction about Domain forms inside on-the-work moments, and does not transfer cleanly across the boundary back into in-the-work execution.

Specifically:

The ON_Discourse community pattern is this dynamic in production. The "Andrews" inside the tent love Doma in-room and can't transfer the conviction to peers. The moment shift is the failure mode, not the explanation.

If Finding A is wrong (conviction transfers fine), conventional referral mechanics would work and the strategy would look like everyone else's. Because it's right, the operational rules change. Finding A bridges to Layer 2 (the closing constraint).

Finding B — The named "channels" are stages of one pipeline

Organic/WOM, content marketing, direct outreach, network partners, and waitlist are not competing acquisition channels. They are sequential stages of a single pipeline:

Network → Waitlist → Activation → Referral

This is a claim about market structure for Domain specifically — not a generalization about all GTM. For Domain, the named "channels" are sequenced because the audience is already known (the network exists; the waitlist is what it produces), the conversion mechanic is high-touch (activation is the rate-limiter), and the referral mechanic is product-gated (the WOM moment has to fire at a specific point in the session).

If Finding B is wrong (the channels really are independent), the right strategic move is portfolio diversification across them. Because it's right, the right move is concentrating on whichever pipeline stage is the actual bottleneck. Finding B bridges to Layer 3 (budget allocation).

A and B are independent

Finding A would still be true if the pipeline structure were different (five disconnected channels with moment-bound conviction in each). Finding B would still be true if conviction transferred fine (a sequential pipeline of regular conversions). Neither implies the other. The strategy needs both.

The reason the doc still works as one strategy: A and B converge at Layer 4. Per channel, the question is both "what's this channel's job in the pipeline" (Finding B) and "what must it not do that would let conviction decay across a moment boundary" (Finding A).


The bridge: Finding A → Layer 2

Domain converts in on-the-work moments, and that conviction decays at moment-boundary — so closing must happen inside the moment, before the user transitions out.

The bridge is the word "so" — and it carries Finding A's full logical weight. If A is wrong, the guardrail is unnecessary. If A is right and the guardrail isn't enforced, the team converts users who'd already convinced themselves and loses everyone else.


Layer 2 — Tactical guardrail: the closing constraint

The rule

Every conversion mechanic either closes inside the on-the-work moment, or it doesn't close at all.

This is a single test applied to every tactic in the four-week sprint. It is not a preference. It is the operational expression of the diagnosis.

What "close inside the moment" means concretely

What the guardrail rules out

Under deadline pressure, teams drift into all of these. They all violate the constraint:

The guardrail's test is independent of the channel name. It applies to any new tactic the team considers.


The bridge: Finding B → Layer 3

The pipeline is sequential, so the 4-week question isn't "which channel" — it's "which pipeline stage is the bottleneck right now," and the budget concentrates there.

Finding B (the pipeline structure) bridges directly to Layer 3. The closing constraint from Layer 2 doesn't determine where the budget goes — Finding B does. The closing constraint determines how the work in that budget bucket gets executed once the stage is identified.

This is why Layer 2 and Layer 3 are distinct. Layer 2 tells you how mechanics must work. Layer 3 tells you which mechanics to invest in. They're independent moves derived from independent findings.


Layer 3 — Budget allocation: 70 / 20 / 10 against the pipeline

The 4-week question isn't "which channel to focus on." It's "which stage of the pipeline is the bottleneck right now." Diagnosis is Week 1's primary work; the rest of the budget responds to what Week 1 reveals.

The split

What the 10% is honest about

In the pressure test, DOMA confirmed: the agentic re-surfacing mechanic (a user's agent surfacing a relevant Toby essay at the right moment) is not currently in production. The 10% bucket protects an asset whose payoff is months out. It earns its 10% because the corpus has compounding value the team isn't currently capturing — not because it converts users this month.

The waitlist kill switch

The 50% activation assumption is untested. The doc commits to a protocol:

Diagnosis without measurement is just hope.


The bridge: A + B converge → Layer 4

Per channel, the question is both "what's this channel's job in the pipeline" (Finding B) and "what must it not do that would let conviction decay across a moment boundary" (Finding A).

This is where the two findings converge. Layer 4 is the only place in the doc where both A and B must be answered at once, per channel. The thing that keeps Layer 4 distinct from Layer 3: Layer 3 says how much attention each stage gets; Layer 4 says, per channel, what the channel does at its stage AND what it must not do given moment decay.


Layer 4 — Channel-specific recommendations and constraints

For each of the five named acquisition channels, here's what it does and what it must not do.

Waitlist activation

Organic / WOM (the viral coefficient)

Direct / high-touch outreach

Content marketing

Network partners


The four-week sequence

Week 1 — Diagnosis week

Week 2

Week 3

Week 4

Launch is "done" when:


The strategy/product collapse (named explicitly)

Three of the four 70%-bucket mechanics are product asks, not marketing asks:

  1. Session-2 mechanic — Domain re-triggering an on-the-work moment for returning users.
  2. Post-session WOM mechanic — firing inside the just-ended moment.
  3. Two-sequence waitlist segmentation — requires signup metadata to segment.

The strategy can name this; it can't paper over it. The pace of the 4-week target depends on these mechanics shipping. This is not a finite-game framing ("hit 1k or fail"); it's a diagnostic about what GTM can and can't do on its own. Product is being developed in parallel; the strategy describes what's possible at each level of product readiness.


What this strategy commits to

Three commitments, one per bridge:

  1. Close inside the moment. Every conversion mechanic either closes before the user leaves the on-the-work moment, or it doesn't close at all. (Finding A → Layer 2.)
  2. Concentrate on the bottleneck pipeline stage, not the channel mix. A small team's attention follows whichever pipeline stage is rate-limiting. The 70/20/10 split is the default allocation; the actual allocation shifts based on Week 1 diagnosis. (Finding B → Layer 3.)
  3. Per channel, name both the job and the prohibition. What each channel does at its pipeline stage AND what it must not do given moment decay. The prohibitions are as important as the recommendations. (A + B → Layer 4.)

Open questions for the team

  1. What's the current waitlist signup rate per week? If signup inflow has plateaued, the pipeline reframe says activation work alone caps near 250 net new — the network has to actively seed during the launch window.
  2. Does Domain have signup source data on the 500? The two-sequence segmentation depends on it.
  3. Can the session-2 product mechanic ship in 3 weeks? Hard product question.
  4. What does the post-session WOM moment look like, concretely?
  5. Who owns the flagship essay? Toby or Andrew? Co-authored?
  6. Which applied AI product is the canonical contrast partner?
  7. Is the public corpus actually agent-accessible right now?

Appendix: v1 → v2 → v3 → v4 → v5 → v6

v1 → v2. Reframed around the original GTM question. Restored the 10% bucket to "AI-native (the agentic bet)." Demoted the contrast move. Added session-2 product mechanic. Sharpened audience framing to pre-aware-not-pre-converted. Named the strategy/product collapse.

v2 → v3. Reframed on-the-work / in-the-work from a channel filter to a closing constraint after DOMA exposed that on/in was doing post-hoc narration on channel selection. Added the pipeline reframe (network → waitlist → activation → referral as sequential, not competing). Added kill-switch tiers and two-sequence segmentation. Cast big bets as moment generators paired with closing mechanics.

v3 → v4. v3 still led with "close inside the moment" as the thesis, conflating Layer 1 (diagnosis) with Layer 2 (guiding policy). v4 separated the four layers per Matt's hierarchy — connection strategy → tactical guardrail → budget allocation → channel constraints — and wrote the bridges between them explicitly using Rumelt's kernel of strategy (diagnosis → guiding policy → coherent action) as the spine. Dropped the finite-game framing; product evolution is the substrate. Recast Layer 4 as per-channel constraints, not just recommendations.

v4 → v5. v4 still pretended Layer 1 had a single diagnosis (moment-bound conviction) and let the pipeline reframe sneak into Layer 3 without an explicit bridge. DOMA caught the gap: "the pipeline isn't a budgeting technique, it's a diagnosis about market structure." v5 named two independent findings at Layer 1 — Finding A (moment-bound conviction) and Finding B (sequential pipeline structure) — and let the bridges fan out honestly: A → Layer 2, B → Layer 3, A + B → Layer 4. Each commitment in the closing section now maps to one bridge.

v5 → v6 (this rewrite). v5 was structurally correct but the sequence misled the reader. The on/in distinction — the foundation everything stands on — was buried three sections deep, inside Layer 1, after the thesis paragraph and the layer table. The reader got the strategy machinery before they were given the world the machinery operates in. DOMA's note: "the reader gets the definition but never feels why it matters before the machinery starts." v6 reorders: the Lakoff visual-field quote opens the doc, "What Domain is for" comes immediately after as the foundation, then the two findings, then the layer table. The logic, layers, and bridges are unchanged. The reader now understands the world before being shown the strategy for operating in it.