What do you count as a conversion?
One to three actions that each mean a real person made contact, such as a validated form submission, a phone call of at least sixty seconds, or a booked appointment, counted once per lead and never counted on a button click.
Google buys more of whatever you count, so this choice sets the ceiling for the whole account. Booked and closed outcomes are sent back from your CRM, which is what teaches bidding to look for customers and not for submissions.
- Primary actions mean a human made contact; everything else is Secondary
- Calls count past a duration floor or when AI-qualified, never every ring
- CRM outcomes come back to Google keyed on the click ID
Playbook conversion tracking method, built on ON-SITE Rentals with lead capture on every page
Where do you let it serve?
Only to people who are in your service area or regularly there, because Google defaults to people who merely showed interest in your area, and that setting sends part of the budget to people who cannot hire you.
The playbook records practitioners finding the default wrong in about 70 percent of the local accounts they audited. Under automated bidding, location bid adjustments are ignored, so geography is controlled by how the campaigns are structured instead.
- Presence-only targeting on every campaign
- A metro campaign and a wider regional campaign that excludes the metro
- A monthly check of where clicks actually came from
Playbook targeting and settings method; the 70 percent figure is the practitioner audit finding it cites
How do you build the account?
By intent and budget, not by keyword: a core service campaign, an urgent-need campaign with its own budget, and a brand campaign, with six to twelve tightly themed ad groups and no broad match at launch.
Each ad group holds five to fifteen keywords around one intent, and the expansion groups are built but paused until the account has enough conversions to learn from. The build goes in by script or API, so every launch is repeatable and reviewable.
- Brand terms kept in their own campaign and excluded everywhere else
- Negative keywords set at account, list and campaign level
- Search partners, display expansion and auto-apply all off
Playbook structure and bidding method, applied to the ON-SITE Rentals build
How do you choose the keywords?
We start from the words your customers use, not the words your trade uses, then test each term for intent, for what a lead is worth to you and for how crowded it is, so you only bid where the math works.
Keyword Planner is read for direction, not for forecasts. In the Ahrefs study of 72,635 keywords, it over-reported the volume of 91 percent of them, so the playbook treats volume as a rank and not as a budget.
- Maximum bid set from what a lead is worth and your landing page conversion rate
- Competitors found free through Auction Insights and the Ads Transparency Center
- Research kept to half a day, not a week
Playbook keyword research method; the 91 percent figure is from the Ahrefs study it cites
How do you write the ads?
Two responsive ads per ad group at launch, written from different angles, in sentence case, with only claims that are true and also on the landing page, so the ad and the page say the same thing.
Small local accounts cannot run a statistically valid test on a dozen unpinned combinations, so the method pins service and city to the first headline, judges on conversions per impression and waits for enough data before declaring a winner.
- No phone number in ad text; calls run through a call asset
- No competitor names in ad text, ever
- Ads sent to the matching service page, not the homepage
Playbook creative and quality score method
What about the landing page?
Every ad lands on the page for that exact service, with the headline echoing the search, a tappable phone number and a short form above the fold, and a site that loads fast on a phone.
The page keeps its normal header and footer, because squeeze pages with no way around are penalized. Only real trust signals go on it: reviews, license, insurance and years serving your area.
- Message match between the ad, the keyword and the headline
- Mobile load under two and a half seconds
- Lead capture on every page, not only the landing page
ON-SITE Rentals: 37 routes, a quote captured from every page, cut over with zero downtime
How do you handle bidding?
The account starts on Maximize Clicks with a capped bid and stays there until it has thirty conversions in thirty days, then moves to Maximize Conversions and later a target cost per lead, one change at a time.
Moving earlier hands Google a model with too little to learn from. After the switch, budgets move by no more than 20 percent in a step and targets by no more than 10 to 15 percent, and never both in the same week.
- Week one: daily search term review, nothing else touched
- Month two: experiment only after twenty to thirty conversions
- Month three: target cost per lead set just above actual, then stepped down
Playbook launch protocol, a 90 day window written down before the account spends
What do you refuse to do?
We do not start with Performance Max, chase an ad strength score, use bid adjustments as a steering wheel or buy click-fraud tools for a small account, because the playbook records why each one costs more than it returns.
In the Adalysis study the playbook cites, Search beat Performance Max on 84 percent of shared queries across more than 3,300 campaigns. Performance Max enters only after the account has qualified conversions and a CRM feedback loop.
- No lead-form assets, which draw lower-quality and spam leads
- No Quality Score as a goal, because Google calls it a diagnostic
- No annual spy-tool subscription for one client
Playbook section "what we deliberately do not do"; the 84 percent figure is from the Adalysis study it cites
What do you check before a dollar is spent?
Every build is run against a 64 point red-team checklist covering ownership, tracking, keywords, structure, settings, ads, landing pages, competitors and reporting, and a single "no" stops the launch.
Each item is a documented way a local account fails while the dashboard still looks fine. The last three questions are asked after the checklist, so the person launching the account has to say out loud what could still be wrong.
- A test click from a fresh IP proves tracking before launch
- Billing in your exact legal name so verification does not fail
- A "not sure" counts as a "no"
Playbook red-team checklist, run against the ON-SITE Rentals build
