How to advertise on ChatGPT: the pre-launch plan.

Make these decisions in order

A campaign is not ready because copy exists. It is ready when the commercial and operational choices behind that copy agree.

1. 1

   Choose one lead offer

   Select a single service, customer situation, and next action. A focused opening test makes it possible to align the message, landing page, routing, economics, and capacity.

2. 2

   Draw the serviceable territory

   Use drive time, dispatch patterns, crew coverage, licensing, and ZIP-level margin, not the largest radius the company could theoretically travel.

3. 3

   Set the test budget

   Work backward from average completed-job value, gross contribution, historical qualification and close rates, available cash, and an explicit loss limit.

4. 4

   Define a qualified lead

   Write the conditions sales will apply to every inquiry, including service need, location, timing, authority, and any job-size threshold.

5. 5

   Approve a defensible promise

   Choose a message that connects the homeowner’s concern to a real differentiator without claiming availability, speed, financing, or expertise the operation cannot support.

6. 6

   Prepare the destination

   Build a page that continues the exact offer, establishes trust, answers the next objections, and makes calling or requesting service easy on a phone.

7. 7

   Lock the scorecard

   Agree what will be tracked, who owns the data, when reviews occur, and what evidence supports holding, changing, pausing, or expanding the test.

Decision 1: one offer beats a menu of five

A first campaign should not ask one ad to sell drain clearing, repiping, water heaters, sewer replacement, and emergency response. Those services have different urgency, economics, proof, staffing, and landing-page needs. Combining them may generate activity, but it makes the result hard to interpret: you will not know which promise earned attention or whether the operation could fulfill it.

Choose the service where customer need, job value, margin, capacity, and a clear next action overlap. For an HVAC company, that might be replacement consultations rather than every repair. For a roofer, it could be inspections for a defined damage scenario rather than a generic “all roofing” proposition. Confirm that the chosen service is actually available during the test window.

Write the offer as a sentence: “For this type of homeowner in this territory, we will provide this specific next step under these conditions.” If operations, sales, and marketing interpret that sentence differently, keep working. Focus now creates cleaner evidence later.

Decision 2: do the service-area math

A map pin and a mileage circle do not describe a profitable territory.

Start with actual drive time by time of day, technician or estimator starting points, tolls, after-hours limits, local licensing, and the cost of sending the right vehicle and crew. Then overlay current backlog. A ZIP code that is attractive for a planned replacement may be uneconomic for a small diagnostic visit; an area that works on Tuesday afternoon may fail for emergency coverage at night.

Review completed jobs by ZIP when records are available. Look at revenue and gross contribution, cancellations, return trips, travel burden, and close rate without pretending thin samples are conclusive. Separate “we have served it” from “we want more work there.” The campaign territory should reflect where the promoted offer earns margin and can be fulfilled now.

Document inclusions, exclusions, and edge cases for intake staff. If a lead falls just outside the area, decide whether it is declined, referred, or reviewed. Paid traffic should not force dispatchers to improvise territory policy.

Decision 3: size a test using your economics

There is no universal starter budget that is responsible for every trade, market, offer, or stage of platform availability.

Use your own average completed-job revenue and gross contribution for the selected service, then apply your historical path from inquiry to qualified opportunity to sold job. Work backward to the most you can afford to spend to acquire a customer while preserving the contribution the business requires. Do not substitute top-line invoice value for available margin.

An illustrative worksheet (not a benchmark) might begin with 20 inquiries simply to make the arithmetic easy. Plug in your own qualification rate, close rate, cancellation rate, contribution per completed job, and desired acquisition allowance. The output becomes a planning ceiling, not a forecast of what ChatGPT ads will deliver.

Finally, set a total test cap and pacing boundaries. The amount must be large enough that management will not panic after a few inquiries, yet small enough that losing the full test amount would not harm payroll or operations. If that balance is impossible, wait.

Decision 4: define “qualified” before leads arrive

A form submission is not automatically a sales opportunity. Write a definition sales can use consistently: the customer needs the promoted service, is inside the approved territory, has an actionable time frame, can authorize or meaningfully influence the work, and meets any legitimate job-size or property requirements.

Also choose disposition reasons. Useful options may include wrong service, outside area, duplicate, spam, tenant without authorization, timing mismatch, unreachable after the agreed follow-up sequence, estimate set, estimate completed, sold, and lost with reason. Keep the list short enough that staff will use it and precise enough that marketing can act on it.

This agreement prevents a common argument in which marketing celebrates lead count while sales dismisses the channel from memory. Each side should be able to inspect the same records and understand why an inquiry did or did not progress.

Decisions 5 and 6: connect the promise to the page

Choose an angle grounded in a homeowner’s decision: understand replacement options, schedule a damage assessment, restore hot water, or request help after a covered loss. Support it with facts the business can prove, such as relevant licensing, the process a customer can expect, service-area clarity, or real financing terms where properly disclosed.

Avoid empty superlatives and operational promises that collapse on a busy day. “Immediate response” is a liability when calls enter an unattended queue. “Free inspection” is misleading if important conditions are hidden. Approved copy should match scheduling, pricing, and field policy, not merely sound persuasive.

The landing page must repeat the service and next step, identify who the offer is for, explain the process, answer high-friction questions, display appropriate trust evidence, and provide a short form plus a tappable phone option when calls are accepted. Send conversions to a confirmation state and test every route. A generic homepage forces the visitor to reconstruct the offer and makes measurement unnecessarily messy.

Decision 7: agree how the test will be judged

Before launch, name the primary outcome and supporting indicators. The primary outcome might be qualified opportunities or sold work once enough time has passed; supporting indicators can include calls, completed forms, contact rate, appointments, estimates, and sales disposition. Reporting availability may evolve, so distinguish platform-reported events from CRM-confirmed business outcomes.

Record the review dates, minimum evidence needed for a decision, approved variables, budget ceiling, and stop conditions. Stop conditions can include broken routing, policy issues, inability to fulfill the offer, or spend beyond the approved risk limit, not disappointment after one unqualified call. Success criteria should be commercially meaningful and should never be presented as a guarantee.

Pre-launch readiness checklist

When every box has an owner, not just a checkmark, you are ready for launch QA. The next job is operational discipline: protect the test from broken tracking, erratic changes, slow follow-up, and decisions made on thin evidence.

• Commercial

  One offer, known job economics, written budget cap, and a truthful proposition approved by the business.

• Territory

  Included and excluded ZIP codes validated against drive time, margin, licensing, and current crew capacity.

• Conversion path

  Focused mobile page, working calls and forms, confirmation state, privacy handling, and correct routing.

• Measurement

  Event tracking, CRM source capture, qualification rules, disposition ownership, review dates, and decision thresholds.

• Operations

  Staffed response coverage, follow-up expectations, scheduling rules, and an escalation owner for failed handoffs.

• Governance

  Current platform access confirmed, required materials submitted, policy review complete, and uncertainty communicated internally.