The brand-dependence story

Strip brand out and the picture changes hard. Brand's $88 CPA looks great, but brand is mostly capturing demand that already exists: the top brand search term "2 sons plumbing" alone spent $3,469 for 162 clicks. That is people who were going to call 2 Sons anyway. The brand campaign already shows for nearly every brand search, and it has an $800/day budget it does not use (it actually spends about $260/day), so most of that budget sits idle.

The non-brand campaign is the actual customer-acquisition engine, and it is running at a $292 reported CPA while showing for only about 1 in 6 relevant searches. That is what the account looks like with brand stripped out, on reported numbers. Growth has to come from here, and it is the most constrained part of the account.

Expected effects

• Reported conversions drop from 138 toward ~51 to 55. The triple-counted phone events collapse and booked jobs become the count. The top of the range allows for keeping the single de-duplicated web-lead form visible; the core number is the 51 booked jobs.
• Reported CPA reads honestly at ~$361 per booked job instead of the fictional $133.50. That $361 becomes the number we drive down, and it is finally a real target.
• Bidding optimizes to booked jobs, not button clicks. Maximize Conversions stops chasing raw calls (part of why non-brand pays $51/click) and starts pulling toward searches that produce actual jobs.
• The three-number call data unifies. One tracking number means call reports finally reconcile.

Verification plan (after the change goes live)

1. Day 0 (change day): screenshot the conversion-actions table showing exactly 2 primary (both Booked Job) and everything else secondary. Confirm the two retired form actions (Contact 2023, thank-you-with-slash) are archived. Log the change with date, who, and why, closing the shared-Gmail no-log gap.
2. Day 1 to 3: confirm reported "Conversions" is trending toward the ~51 to 55 range, not 138. Confirm the ServiceTitan Booked Job actions still show "recording conversions" status (imports healthy).
3. Day 3 to 7: reconcile Google Ads booked jobs against the ServiceTitan job list for the same window. They should match within normal import lag. If they diverge, check the offline-conversion upload and GCLID stitching before touching bids.
4. Day 7 to 14: watch that Maximize Conversions has not over-throttled on the smaller (but honest) conversion count. Non-brand may briefly pull back as the algorithm relearns; that is expected. Hold bids steady and let it settle.
5. Day 30: compare cost per booked job versus the $361 baseline. That is the first honest before/after we will have on this account, and the number we report to Natalie.

Non-brand keywords average Quality Score 4.2, with 14 of 22 scored keywords at 4 or below. Brand averages 8.3. Low Quality Score is why non-brand clicks cost about $51 each (versus $22 on brand), why the ads show for only about 1 in 6 relevant searches, and why single clicks on terms like "auburn plumbers" ($277 for 2 clicks) and "24 7 plumbers" ($187 for 1 click) are burning the budget.

A rebalance has two moves, and both are needed:

• Shift idle brand budget to non-brand. Brand uses only ~$260 of its $800/day. Non-brand's budget runs out daily, which costs it about a third of its chances to show. Moving brand's unused budget to non-brand is close to free.
• Lift Quality Score so the extra budget is not wasted at $51/click. Tighter match types (124 broad-match keywords are pulling competitor and out-of-scope traffic), better ad-to-keyword-to-landing-page relevance, and the non-brand landing page rebuild. Our proposed page is live as a mock at 2sons-lpmock.vercel.app, built to raise landing-page experience, one of the three Quality Score inputs.

Budget without Quality Score just buys more $51 clicks. Quality Score without budget still leaves the ads showing for only about 1 in 6 relevant searches. We do them together.

The test that cannot run

The "Landing Page Test" campaign is enabled but has spent $0, because it shares the same budget (ID ...220844) as the non-brand campaign. Two campaigns on one shared budget compete for the same dollars, and the established non-brand campaign consumes them. The LP test cannot get incremental spend without starving non-brand. The fix: give it its own budget and point it at the new mock, or pause it until the LP rebuild is ready. Right now it is a test that cannot run.

The July 20 PMAX pauses

Three Performance Max retargeting campaigns (Sewer & Drain, General Plumbing, Water Heater) spent $3,012 for 23 reported conversions before being paused. The change history shows three campaign edits on July 20 at 10:28am by p3servicesmarketingteam3@gmail.com (a shared Gmail account, via the web UI), lining up exactly with the three pauses.

Our read: the pause was probably the right call on the merits, but it was done blind. Retargeting is a weak fit for emergency plumbing; someone with a burst pipe calls the first credible result, they do not need to be followed around the web. PMAX retargeting also tends to claim credit for conversions that brand or direct would have captured anyway, which inflates its apparent 23 conversions. The problem is process, not outcome: no documentation, a shared internal login, no before/after measurement, so we would not know if it helped or hurt. When P3 gives the go-ahead, we would formalize the decision (keep it paused, redirect that ~$3,000/month toward non-brand) and note it, so the next person understands why.

What this does not touch (leave running)

Generic in-area plumbing terms that did not convert this window but are legitimate demand: "plumber near me," "plumbers near me," "plumbing in seattle," "plumbing company," "plumbing services near me." These are the core of the non-brand campaign. The fix for their efficiency is Quality Score and bidding, not negatives. We do not exclude them.

The graveyard

The ~48 paused campaigns are three generations of past-agency structures: manual-CPC relics with $1,000+/day budgets left attached (Water Heater Repair at $1,000/day, Sewer Pipe Lining at $1,500/day), an "H - {city} - {service}" grid (H - Seattle - Plumbing, H - Tacoma - HVAC, and so on), and WS-era 2023 campaigns. None are spending, so they are not urgent, but they make the account hard to read and carry oversized budgets that could be triggered by accident. Cleanup is a later, low-risk tidy.

Governance observations (factual, no blame)

• A shared Gmail login is making live changes. p3servicesmarketingteam3@gmail.com paused three PMAX campaigns on July 20 via the web UI. Shared logins make it impossible to know who changed what, and they are a security exposure. We recommend named user access with appropriate roles.
• Multiple hands, no change log. July edits came from jared.tuttle@ and farrukh.saleem@ (P3 domain) plus the shared Gmail. Changes are not documented anywhere we can see, so intent is lost.
• Three phone tracking numbers ((206) 710-2259, the "Joe's 2 Sons" number, and (206) 487-1757) fragment the call data. Consolidating to one number simplifies both tracking and the customer experience.
• Oversized budgets on paused relics (up to $1,500/day) sit ready to fire if a campaign is toggled on by accident. Worth zeroing out during cleanup.

None of this is a blame point for Natalie. It is the inherited state of the account, and it is fixable.