Paid search

# Not live yet (gaps #2)
curl -X POST "https://api.sb/leaseline.app/campaigns?mode=try" \
  -H "Authorization: Bearer $STARTUP_BUILDER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"play":"paid-search"}'

 1import { $ } from 'startup-builder'
 2
 3return await $.Campaign.create({ play: 'paid-search' }, { mode: 'try' })   // dry run on the day-0 Startup, 2026-08-24 · Not live yet, gaps #2

npx startup-builder do Campaign.create --json '{"play":"paid-search"}' --try   # Not live yet (gaps #2)

{
  "name": "do",
  "arguments": {
    "code": "return await $.Campaign.create({ play: 'paid-search' }, { mode: 'try' })   // dry run on the day-0 Startup, 2026-08-24 · Not live yet, gaps #2",
    "mode": "try"
  }
}

 1{
 2  "type": "https://startupbuilder.dev/errors/refused",
 3  "title": "Refused",
 4  "status": 422,
 5  "detail": "paid-search can’t start yet: 2 checks fail and 1 warns.",
 6  "checks": [
 7    {
 8      "check": "use-when",
 9      "status": "pass",
10      "observed": "1,900 exact-intent searches a month across 14 terms",
11      "doc": "/plays/paid-search#check-use-when",
12      "fix": "Nothing to do: it counts searches for your exact-intent terms.",
13      "then": "Keyword.measure"
14    },
15  …

curl "https://api.sb/leaseline.app/events?where=name%20in%20(%22Integration.synced%22%2C%20%22Keyword.measured%22%2C%20%22Invoice.collected%22%2C%20%22Organization.announced%22%2C%20%22Offer.repriced%22%2C%20%22Experiment.matured%22)&limit=1" \
  -H "Authorization: Bearer $STARTUP_BUILDER_API_KEY"

# Not live yet (gaps #7)
curl -X POST "https://api.sb/leaseline.app/triggers/fire?mode=try" \
  -H "Authorization: Bearer $STARTUP_BUILDER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"like":"event_1402"}'

# Not live yet (gaps #7)
curl "https://api.sb/leaseline.app/decisions?where=on%20%3D%20%22batch_44%22" \
  -H "Authorization: Bearer $STARTUP_BUILDER_API_KEY"

# Not live yet (gaps #7)
curl "https://api.sb/leaseline.app/actions?where=on%20%3D%20%22batch_44%22" \
  -H "Authorization: Bearer $STARTUP_BUILDER_API_KEY"

 1import { $ } from 'startup-builder'
 2
 3const { events: [last] } = await $.Event.list({ where: 'name in ("Integration.synced", "Keyword.measured", "Invoice.collected", "Organization.announced", "Offer.repriced", "Experiment.matured")', limit: 1 })   // { play } compiled to this Play’s Trigger Events
 4const { subject } = await $.Trigger.fire({ like: last.$id }, { mode: 'try' })   // the same Event again, in try mode · Not live yet, gaps #7
 5const { decisions: [decision] } = await $.Decision.list({ where: `on = "${subject}"` })   // what it chose, and why · Not live yet, gaps #7
 6const { actions } = await $.Action.list({ where: `on = "${subject}"` })   // what it did, with its checks · Not live yet, gaps #7
 7return { decision, actions }

npx startup-builder do Event.list --json '{"where":"name in (\"Integration.synced\", \"Keyword.measured\", \"Invoice.collected\", \"Organization.announced\", \"Offer.repriced\", \"Experiment.matured\")","limit":1}'
npx startup-builder do Trigger.fire --json '{"like":"event_1402"}' --try   # Not live yet (gaps #7)
npx startup-builder do Decision.list --json '{"where":"on = \"batch_44\""}'   # Not live yet (gaps #7)
npx startup-builder do Action.list --json '{"where":"on = \"batch_44\""}'   # Not live yet (gaps #7)

{
  "name": "do",
  "arguments": {
    "code": "const { events: [last] } = await $.Event.list({ where: 'name in (\"Integration.synced\", \"Keyword.measured\", \"Invoice.collected\", \"Organization.announced\", \"Offer.repriced\", \"Experiment.matured\")', limit: 1 })   // { play } compiled to this Play’s Trigger Events\nconst { subject } = await $.Trigger.fire({ like: last.$id }, { mode: 'try' })   // the same Event again, in try mode · Not live yet, gaps #7\nconst { decisions: [decision] } = await $.Decision.list({ where: `on = \"${subject}\"` })   // what it chose, and why · Not live yet, gaps #7\nconst { actions } = await $.Action.list({ where: `on = \"${subject}\"` })   // what it did, with its checks · Not live yet, gaps #7\nreturn { decision, actions }",
    "mode": "try"
  }
}

 1{
 2  "decision": {
 3    "$id": "https://api.sb/leaseline.app/decisions/decision_581",
 4    "$type": "Decision",
 5    "choice": "block-wasted-terms",
 6    "reason": "31 of 214 clicked terms match the negative rules in your file (jobs, free, template, login), and they cost $86 yesterday",
 7    "stakes": "low",
 8    "options": [
 9      "block-wasted-terms",
10      "queue-new-terms",
11      "reset-bids",
12      "take-the-next-step",
13      "defend-your-name",
14      "not-now"
15  …

action_2 adds the 31 negatives through the write proxy at once, because your rules cover them. action_3 queues 6 terms no rule covers for your OK in today’s batch; a batch waiting 48 hours halves the fenced arm’s budget.

≈ +$1,000 ARR a month at $1,100 spend

Worth

1 h 30 min week 1, then 20 min a week

Your time

$100 probe, then $1,000 a month

Spend

Day 9: first sale from a paid Visit

First signal

500+ exact-intent searches a month, a known margin

Use it when

Under 500 a month: editorial-seo · no known margin: set the cost (Offer.update)

Not yet when

Worth’s formula: about +13.6 incremental ICP Visits per $100 (your switchback, ■■□□ Early sign) × $1,100 a month = 150 Visits × 0.02 Visits to paid × $347 Pro ARR ≈ +$1,040. Under First signal: A day switchback at $1,100 a month tells a +10 lift per $100 in about 8 weeks; each scale step waits for a 28-day holdout read.

Start it this weekCopy Anchor Link

npx startup-builder add play paid-search   # 1 · writes plays/paid-search.mdx (Not live yet, gaps #5)

# 2 · dry run; below, its 422 on 2026-08-24
curl -X POST "https://api.sb/leaseline.app/campaigns?mode=try" \
  -H "Authorization: Bearer $STARTUP_BUILDER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"play":"paid-search"}'

# 3 · keep it: every change waits for your OK
curl -X POST https://api.sb/leaseline.app/campaigns \
  -H "Authorization: Bearer $STARTUP_BUILDER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"play":"paid-search"}'

# the probe’s $100 monthly Limit
curl -X POST https://api.sb/leaseline.app/limits \
  -H "Authorization: Bearer $STARTUP_BUILDER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"name":"paid-search-month","amount":100,"period":"month"}'

# off: every Campaign pauses through the proxy, and write scope is revoked
curl -X PATCH https://api.sb/leaseline.app/campaigns/paid-search \
  -H "Authorization: Bearer $STARTUP_BUILDER_API_KEY" \
  -H "Content-Type: application/json" \
  -d '{"endsAt":"now"}'

{
  "type": "https://startupbuilder.dev/errors/refused",
  "title": "Refused",
  "status": 422,
  "detail": "paid-search can’t start yet: 2 checks fail and 1 warns.",
  "checks": [
    {
      "check": "use-when",
      "status": "pass",
      "observed": "1,900 exact-intent searches a month across 14 terms",
      "doc": "/plays/paid-search#check-use-when",
      "fix": "Nothing to do: it counts searches for your exact-intent terms.",
      "then": "Keyword.measure"
    },
    {
      "check": "baseline",
      "status": "warn",
      "observed": "0 days of paid spend: no switchback yet",
      "doc": "/plays/paid-search#check-baseline",
      "fix": "n 0: the probe’s switchback becomes the baseline in 28 days.",
      "then": "Metric.query"
    },
    {
      "check": "billing-connected",
      "status": "fail",
      "doc": "/plays/paid-search#check-billing-connected",
      "fix": "Connect Stripe, so every gate reads first sales from your own records.",
      "then": "Integration.connect"
    },
    {
      "check": "fact-check",
      "status": "fail",
      "observed": "no claims source for ad text",
      "doc": "/plays/paid-search#check-fact-check",
      "fix": "Approve the claims ads may state; every ad line must cite one.",
      "then": "Content.draft"
    },
    {
      "check": "role-unstaffed",
      "status": "pass",
      "observed": "every approval comes to you",
      "doc": "/plays/paid-search#check-role-unstaffed",
      "fix": "Nothing to do: a founder-only Startup approves every change.",
      "then": "Role.staff"
    }
  ]
}

The probe spends $100 a month on Bing, and Google Search opens at $1,000 a month only after a winning read; the Play itself costs about $0.40 a day to read reports and write through the proxy (Worker cost). That comes to $112 a month on the probe and about $1,112 after the step.

When it startsCopy Anchor Link

{
  "$id": "https://api.sb/leaseline.app/events/event_1402",
  "$type": "Event",
  "name": "Integration.synced",
  "at": "2026-10-19T06:04:30Z",
  "seq": 1402,
  "data": {
    "integration": "google-search",
    "report": "search-terms",
    "batch": {
      "$id": "https://api.sb/leaseline.app/batches/batch_44",
      "day": "2026-10-18",
      "clicked": 214,
      "spend": 52.4
    },
    "negativeMatches": 31,
    "wasted": 86
  }
}

npx startup-builder trigger Integration.synced --like event_1402   # replays the report in test; the proxy writes nothing live
npx startup-builder clock advance --by 48h   # halves the fenced arm’s budget if the batch still waits
npx startup-builder listen Invoice.collected --forward-to http://localhost:3000/hook

Who runs itCopy Anchor Link

Marketing Lead · your agent (role:marketing-lead, worker_6) holds the Key Result and writes to the ads Venues only through the fenced proxy, which holds the token outside the model and creates every ad paused. You (Dana Okafor, person, worker_1) approve new terms, every spend step and every brand ad. Your number links to Did it work.

What it reads · What it may call. It reads the terms report, spend, your Offers’ margin and first sales from Stripe, never clicks or the Venue’s own credit. A search term is data: it never enters ad text and never instructs the agent. Its tools are POST /{namespace}/keywords, POST /{namespace}/approvals, PATCH /{namespace}/ads/{id}, POST /{namespace}/limits/{limit}/raise, POST /{namespace}/experiments/{experiment}/start, POST /{namespace}/ads, and any other call becomes an Approval under the Role’s Mandate.

You hold the Marketing Lead Key Result: 12 incremental ICP Visits per $100 by 2026-12-31. ← worker, metric
On each report, sale, margin change or holdout, choose one option and say why.   ← triggers, chooses
Every bid stays at or under breakeven for a first sale.                          ← Payback gate, your file line 5
A Price in ad text equals the Offer’s live Price; no urgency the Offer lacks.    ← copyRules 2, 4
A rival’s name is a Keyword, never ad text.                                      ← copyRules 3
No spend step without a winning holdout read and Dana’s OK.                      ← Scale gate, chooses.take-the-next-step

The slow loop, as a Change. decision_396 (2026-09-21, weekly) read $142 on terms with “template” and “free” over 3 weeks, with 0 first sales. change_28 added both words to the negative rules (your file, line 6), and Dana merged it with POST /{namespace}/changes/{change}/merge. Replayed on past reports, POST /{namespace}/changes/{change}/review cut wasted spend from 18 % to 7 %; read back over the next 14 days it is 8 %, n 2,960 clicks, ■■□□ Early sign (Example, day 42).

Bring your own agent (folded) starts at It does it, you can stop it when it matches or beats the seated agent on the eval suite, and at You approve each one when it does not; it still writes only through the proxy.

npx startup-builder eval --scenario paid-search-56 --agent "claude -p" --seat marketing-lead   # seated: 8 % wasted, 0 Checks failed sent · yours: 11 % wasted → You approve each one

What it can do nextCopy Anchor Link

When it hands to youCopy Anchor Link

Every new term batch, spend step and brand ad comes to you: the Marketing Lead Role has no person, and a spend step is never the agent’s call. One Brief, as captured, then the Approval (brief, draft and checks are Not live yet, gaps #7).

[
  {
    "$id": "https://api.sb/leaseline.app/tasks/task_427",
    "$type": "Task",
    "assignee": "worker_1",
    "due": "2026-09-30T17:00:00Z",
    "on": "experiment_6",
    "brief": {
      "what": "the Bing probe’s 28-day switchback",
      "read": "+15.1 incremental ICP Visits per $100 [+4.2, +26.0], breakeven 8",
      "chose": "take-the-next-step",
      "why": "above breakeven on a winning read",
      "step": "open Google Search at $1,000 a month",
      "never": [
        "a raise without your OK",
        "the fenced arm before 30 first sales"
      ]
    }
  },
  {
    "$id": "https://api.sb/leaseline.app/approvals/approval_96",
    "$type": "Approval",
    "verdict": null,
    "deadline": "2026-10-21T06:00:00Z",
    "draft": "6 new terms, paused: “renewal notice software for landlords”, “owner statement software”…",
    "route": "queue-new-terms",
    "checks": [
      {
        "check": "fact-check",
        "status": "pass"
      },
      {
        "check": "scale-gate",
        "status": "pass"
      }
    ]
  }
]

You answer with POST /{namespace}/approvals/{approval}/approve or POST /{namespace}/approvals/{approval}/refuse, or POST /{namespace}/tasks/{task}/claim (Take this one) and POST /{namespace}/tasks/{task}/return with a reason, each a one-call window. A term batch unanswered for 48 hours halves the fenced arm’s budget, and at 96 hours the arm pauses (negative-backlog); an unanswered spend step simply never happens.

On time: you, 7 of 8 batches answered within 48 h, median 9 h, n 8, and 1 of 1 spend steps within its day, ■■□□ Early sign beside Other Startups’ 88 % on time. Your queue: counted by kind: in week 1, the account plan and 3 batches; now about 2 batches a week, and a spend step at each holdout read.

The fixed line: POST /{namespace}/deals/{deal}/assign sets owner to the receiving person, the agent is never a Deal’s owner, and Meetings land on the owner’s calendar under the owner’s name.

When it won’t sendCopy Anchor Link

The captured refusal, then the Stop that fired as a Campaign.paused reason (Not live yet, gaps #2); the rewrite that went live is the Sep 8 ad in Example.

{
  "type": "https://startupbuilder.dev/errors/refused",
  "title": "Refused",
  "status": 422,
  "detail": "The new ad for “lease renewal software” was not created: one headline states a result the claims source lacks.",
  "checks": [
    {
      "check": "fact-check",
      "status": "fail",
      "doc": "/plays/paid-search#check-fact-check",
      "fix": "Cut the claim, or name the record that backs it.",
      "then": "Content.draft"
    },
    {
      "check": "scale-gate",
      "status": "pass"
    }
  ]
}

[
  {
    "reason": {
      "limit": "conversion-feed",
      "observed": "0 conversions sent in 81 h",
      "bound": "72 h (3 × the 24 h gap)",
      "window": "Oct 2 to Oct 5"
    },
    "doc": "/plays/paid-search#stop-conversion-feed",
    "task": "task_433"
  }
]

Did it workCopy Anchor Link

Days were assigned at random, ads on or off (a switchback, rule o: 28 days each half); both halves are read from that assignment. Each figure’s call is POST /{namespace}/metrics/query by node and by arm; the parent field and every node above the leaf are Not live yet (gaps #4).

{
  "$id": "decisions/decision_581",
  "$type": "Decision",
  "choice": "block-wasted-terms",
  "reason": "31 of 214 clicked terms match the negative rules in your file (jobs, free, template, login), and they cost $86 yesterday",
  "stakes": "low",
  "options": [
    "block-wasted-terms",
    "queue-new-terms",
    "reset-bids",
    "take-the-next-step",
    "defend-your-name",
    "not-now"
  ],
  "by": "worker_6",
  "on": "batch_44",
  "expected": "wasted spend under 10 % of the next 7 days",
  "reopensWhen": "wasted share over 10 % in the next Integration.synced terms report"
}

The leaf counts ICP Visits on ads-on days minus ads-off days, for each $100 spent, over the switchback’s 56 days; the Google Search step on day 35 keeps the same day assignment, and its own 28-day holdout reads on Nov 22.

How it moves it: ads showed to people searching your exact-intent terms at bids at or under breakeven, and 412 wasted terms were blocked the next morning (search-icp-visits-per-100 by term).

Guardrails: treated against held back: cost per first sale, $295 against a $347 margin, and Visits to paid, 0.35 % against 0.10 %.

{
  "$id": "https://api.sb/leaseline.app/campaigns/paid-search",
  "$type": "Campaign",
  "play": "play:paid-search@1",
  "path": [
    "search-icp-visits-per-100",
    "icp-reached",
    "new-arr",
    "net-new-arr",
    "arr"
  ],
  "experiment": {
    "metric": "search-icp-visits-per-100",
    "unit": "per-100-usd",
    "heldBack": 28,
    "decision": null
  },
  "readout": {
    "value": 13.6,
    "n": 56,
    "maturing": 0,
    "interval": [
      4.1,
      23.1
    ],
    "readability": "B",
    "basis": "alternating-weeks",
    "pooled": false
  },
  "ledger": {
    "spend": 1180,
    "workerCost": 22,
    "hours": 5.4,
    "pipelineArr": 2400,
    "wonArr": 1388,
    "projectedArr": 1040,
    "paybackDays": 310
  }
}

Outside benchmark: the catalog holds none for this Play yet, so Other Startups’ results stand in: +9.0 incremental ICP Visits per $100, ■■■□ Likely, n 47 Startups.

Before it · After it. Before it: editorial-seo and aeo find the exact-intent terms worth a bid. After it: every form, chat or email from a paid Visit starts instant-response, and retargeting hears the Visitors who left without one.

ExampleCopy Anchor Link

Leaseline ran it from 2026-08-24 to 2026-10-19 (56 days): the Bing probe from day 0, Google Search from day 35 at $1,000 a month. $1,180 spent on 28 ads-on days; 412 negatives blocked and 44 new terms approved. Read every captured thread at ?threads=all; three follow the tables.

Search: “lease renewal software”, Sep 2, 10:14 · the ad: Lease renewals, sent 60 days out · Per-owner notices and reports · Pro from $29 a month.¹

The click · landed on the renewals Page; a trial started at 10:22 and paid on Pro on Sep 16. Stripe recorded the first sale, and the conversion send reached the Venue at 10:24 the next day.²

- Cut late renewals by 40%
+ Renewal notices sent 60 days out

The auction report, 2026-10-06 · a rival began bidding on “leaseline” · the agent drafted one brand ad group, paused, and the Approval went to Dana with a deadline of Oct 7 at 06:00 (approval_91).³ Dana approved it at 21:40 on Oct 6, inside the deadline, and the ad group went live through the proxy at 21:41.

¹ The ad’s Price equals Pro’s live Price; no rival name in ad text. ² Every gate counts the sale from Stripe, never the Venue’s own credit. ³ A brand ad group is yours to approve, on every rung.

Your settingsCopy Anchor Link

---
limit: 1000       # dollars a month; it was 100 for the probe
holdout: switchback-days
---
Bids stay at or under breakeven for a first sale.
Block terms with jobs, free, template or login the next morning.
No spend step without a holdout read and my OK.

$ npx startup-builder show plays/paid-search
limit $1,000 a month · day switchback · breakeven bids · 4 negative rules · spend steps to you     your file  plays/paid-search.mdx:2,3,5,6,7 (Ask 2026-08-24; line 2 by your OK on day 35, line 6 by change_28, 2026-09-21)
family ads · worker role:marketing-lead, owns never · chooses 6 (2 rule, 4 read) · approval every-send · checks fact-check, scale-gate · stops 12 · placements google-search, bing, through the write proxy    catalog  play:paid-search@1 (Not live yet, gaps #5)

Every key from the catalogCopy Anchor Link

Compiled, not written: edit the Play record paid-search and its Leaseline run, then regenerate. Every step runs from a sentence or a line of code. In your words: the CRM words you know, and where each lives here.