Google Ads Operating Playbook

Part 1

The doctrine: how Tom thinks about Google

Ten principles sit under every click in this playbook. When a situation isn't covered, decide with these.

Search is the highest-quality lead source in all of performance marketing. People are actively in market, typing what they want. Everything wider than search - Performance Max, YouTube, Display, Demand Gen - drops user quality. Smart Energy has tested them; nothing rivals search.
Keywords are the targeting. Keep control of them. Meta finds your customer from the creative; Google only does what your keyword structure tells it. Launch on exact match, expand deliberately from real search-term data, and never let Google default you into broad match.
Ad relevance is how you win. The ad and landing page must scream back exactly what the user searched. Someone typing "purifier" sees purifier copy, never filter copy. That is the entire reason ad groups exist.
Consolidate so the machine can learn. Never 20 campaigns with 20 ad groups: it spreads conversion data too thin. Split campaigns and ad groups only where there's a clear delineation (product, geo, hours, intent). Rough guide Google gave internally: ~30 conversions per campaign per month - treat it as folklore, not physics, and simply keep the structure as consolidated as possible.
Structure follows the sales team. In-hours and out-of-hours are separate campaigns because a lead called within minutes closes better. Control where the money goes; don't let Google chase cheap out-of-hours leads your team can't service.
Distrust every automated "help". AI Max, auto-generated ads and keywords, "Upgrade to broad match - Recommended", suggested budgets, auto-applied recommendations: all of it is Google spending your money more broadly. Experiment deliberately or not at all.
Volume and efficiency are an inverse trade. Lower target CPA = cheaper leads but fewer auctions entered. Decide each week which one the business needs, then move the matching lever.
Two levers, never confused. A campaign limited by budget needs budget. A campaign limited by target needs a looser target. Adding budget to a target-limited campaign does nothing.
Intent beats price-hunters. Rebate/discount searchers convert worse than product searchers. Spend followed lead quality out of the rebate campaign and into the battery campaigns - apply the same lens to every new keyword theme.
There is no consensus playbook - experiment per account. The community agrees on Meta, not on Google. When opinions conflict (bid strategies, match types, CBO vs ABO), run a 50/50 experiment in your own account and let it settle the argument.

Tom's one-liner

"Live by the Google Ads keywords, die by the Google Ads keywords." If you only do one thing every week, work the search term report.

Part 2

Researching a new category with Keyword Planner

Before any campaign exists, validate that traffic is there and map it into a structure. This is exactly how the filtration category was scoped.

1

Open Keyword Planner

Tools›Planning›Keyword Planner. Type one seed keyword for the category (here: home water filter) and set the location filter - all of Australia for sizing, your launch states later.

Keyword Planner navigation, seed keyword and filters annotated — **Where everything lives.** Tools sits in the far-left rail; Keyword Planner is under Planning. The filter bar controls geography, language, network and date range.

2

Sort by volume, read for themes

One seed returns 1,500+ ideas. Sort by average monthly searches so you see where the demand sits, then read the list looking for themes, not individual keywords: here nearly everything clustered around house/home + system, with a second vocabulary split between filtration and purification. That split becomes the ad group structure.

Keyword ideas list sorted by average monthly searches — **The size of the category in one screen.** Every variation of the seed with its monthly volume. Themes you find here become campaigns and ad groups.

3

Qualify intent before volume

The biggest head terms are often the worst buyers. water filter has 33,000 monthly searches, but that user may want a $20 jug. Deliberately skip bare head terms and require a qualifier - house, home, whole house, residential, system - because that user is in market for an installed system. Rule of thumb for a launch structure: work everything with roughly 1,000+ monthly searches into as consolidated a structure as possible.

Selecting keywords by theme in Keyword Planner — **Theming pass.** Tick the keywords that share a theme; each coherent theme with real volume becomes an ad group (or, if it needs its own budget, a campaign).

4

Read CPC and competition, loosely

Top-of-page bid (low and high range) gives an order-of-magnitude CPC estimate; competition "High" is normal on anything with volume. Take both with a grain of salt - campaign settings move real CPCs a lot - but use them to sketch funnel economics: CPC → cost per lead → cost per appointment → CAC.

Competition and top of page bid columns in Keyword Planner — **Economics preview.** Bid ranges on the right, competition in the middle. Directional only.

Launch gate

Do not build until three things exist: a landing page for the product, verified conversion tracking, and someone rostered to field the leads. Traffic without those burns money.

Brands

Competitor and brand terms show up in the ideas list (Brita and friends). You can target them, but the flywheel starts with generic non-brand terms. Brand/competitor ad groups come later, with their own copy.

Part 3

Building a new campaign, click by click

The exact sequence from the filtration build. Fourteen decisions, each with the reason behind it. Total build time once keywords are mapped: under an hour.

1

Pick one conversion goal

New campaign → objective Leads → goal Contacts, and optimise to a single conversion action: Lead – Filtration. Down the track HubSpot adds appointment / sale / install actions for reporting, but the campaign optimises to one action only, so the signal stays clean.

Selecting the Contacts goal and Lead - Filtration conversion action — **One campaign, one conversion action.** Everything else that fires is reporting, not optimisation.

2

Campaign type: Search

Choosing the Search campaign type — **Search first, always.** Wider formats come later, if ever, and only as deliberate tests.

3

Website conversions only, and name it properly

Tick Website visits with the site URL; leave phone calls and every other reach option unticked - the funnel converts on the landing page. Then name the campaign to the account convention:

Search_Filtration_In-Hours_NSW-QLD - that is channel_product_hours_geo. Check the existing campaigns first so naming stays consistent; every report and script depends on it.

Website visits reach option and campaign naming convention — **The naming convention is not cosmetic.** In-hours vs out-of-hours and geo splits are how budget control works later.

4

Bidding: start on Clicks, no target

A brand-new campaign has no conversion history, so give Google the simplest job: Maximise clicks, no max CPC, no target CPA. The ladder from here: once conversions register, switch to conversion bidding; once you know what you're willing to pay, add a target CPA. (Full ladder in Part 10.)

Bidding dropdown set to Clicks — **Clicks → Conversions → Target CPA.** Never launch a cold campaign straight onto a target.

5

Networks: partners on, Display off

Search partner sites are acceptable at launch for extra volume. The Display Network is never opted into on a search campaign - it silently drains budget into low-intent placements.

Search partners enabled, Display network disabled — **The Display tick-box is the classic silent budget leak.** Check it stayed off every time you duplicate a campaign.

6

Locations: name each metro, set Presence

Enter each launch region by name - Sydney, Wollongong, Central Coast, Newcastle, Byron Bay, Tweed Heads, Gold Coast, Brisbane, Sunshine Coast - rather than blanket states, so spend concentrates where the sales team can service. Then open Location options and set Presence: people in or regularly in your locations.

Location list with reach figures and the Presence radio button — **Presence, never "Presence or interest".** "Interest" lets someone overseas researching a Sydney trip see your ad. This one radio button quietly wastes more money than almost any other setting.

7

Language English; skip EU political

English only. The EU political ads question is a No for this account.

8

Audiences: load signals, in Observation mode

Search targeting is keywords - but you can hand the algorithm audience data to learn from. Add every plausibly related segment as observation: Google then notices which audiences convert and bids more aggressively when a searcher belongs to one. You are not narrowing targeting; you're enriching it.

Who they are: Homeowners; tick all employment industries (industry trends surface in the data later).
Affinity: Do-It-Yourselfers, Family-Focused, Green Living Enthusiasts, Outdoor Enthusiasts, Health & Fitness Buffs, Sports Fans, Technophiles, Social Media Enthusiasts.
In-market (Home & Garden): Home Cleaning, Home Inspection, Landscape Design, Locksmiths, Painting, Plumbing, Home Improvement, Plumbing Fixtures - and the bullseye: Water Purifiers.

Audience segment added with Observation setting highlighted — **The Observation radio is the whole trick.** "Targeting" would narrow reach to these segments; Observation just feeds the bidder.

Browse tab with homeownership and industry segments — **Browse → Who they are.** Homeowners obviously; all industries because you can't predict which one over-indexes.

In-market Water Purifiers segment ticked — **The in-market segment that matches your product is gold.** Google already tagged these people as actively shopping - from stores visited, YouTube watched, searches made. When they search your keyword, Google doubles down.

9

Ad rotation and schedule

Ad rotation: optimise for best performing ads. Schedule: this is an in-hours campaign, so Monday–Friday, 08:00–17:00 - 8am start so a pipeline of leads is waiting when sales sits down, 17:00 close so nothing arrives unattended. Out-of-hours demand gets its own campaign when budget scales.

Ad schedule set to Mondays to Fridays 8am to 5pm — **The schedule is a sales-ops decision, not a media decision.** A filtration lead called within minutes converts; one called tomorrow doesn't.

10

Refuse the AI: no AI Max, no generated keywords, no generated ad

Do not enable AI Max ("pretty shitty" - direct quote). Skip keyword and asset generation, and delete every keyword Google pre-fills. You will paste your own researched, exact-match list.

Skipping Google's AI keyword and asset generation step — **Skip means skip.** Even after skipping, Google may still create an ad for you - step 4 of Part 4 deals with that ambush.

11

Enter keywords: systemised permutations, all exact match

Build the list as a grid, not a brainstorm: {filter, filtration, filter system, filtration system} × {house, home, whole house, whole home, residential}, water-prefixed, every one wrapped in square brackets for exact match. That produced 22 keywords, all with confirmed volume. Skip combinations that get too long - exact-match close variants will catch them anyway.

Exact match keyword list in square brackets — **[Square brackets] = exact match = control.** The legend under the box is the only Google-supplied text worth reading on this screen; the "Add more keywords" nudge below it is not.

12

Budget: custom, ignore the suggestions

Google will propose ~A$200–300/day. Set a custom budget at what the test actually needs: A$100/day for this launch. You scale from data, not from Google's appetite.

Custom budget of $100 replacing Google's suggested budgets — **Suggested budgets are an upsell, not an analysis.**

13

Review and publish

Fix anything in the Issues panel (a missing ad is fine - ads are built properly inside the ad groups next). An identity-verification dialog may interrupt; confirm and continue. Publish.

Campaign review screen with identity verification dialog — **Publishing without an ad is allowed** - the campaign sits pending until ads exist, which is exactly what you want while copy gets sign-off.

14

Get copy sign-off before it spends

Share screenshots of the ad copy with the team for approval while the campaign is pending. Nothing goes live unseen.

Part 4

Ad groups, keywords and ads

The relevance machine: one tightly themed ad group per search vocabulary, an ad that mirrors it, and sitelinks to match. This is the filtration/purification build, and it's the template for every future theme.

1

Name the ad group for its theme

Campaigns›Ad groups - the fresh campaign has one unnamed ad group. Rename it to its keyword theme (…_Filtration), matching account convention. "Generic" is the name used for a plain product-terms group in the battery campaigns; here the vocabulary split makes Filtration and Purification the right names.

Renaming the ad group to the filtration theme — **Ad group names describe the keyword theme.** Later groups follow the same pattern: _Generic, _Quote, _Tesla, _Brands…

2

Confirm every keyword has volume

Open the keywords list. New keywords show "Pending / Under review" - normal. The flag that matters later is Low search volume: a keyword nobody searches. All 22 filtration keywords had volume; that's plenty for A$100/day.

Keyword list showing exact match keywords under review — **22 exact-match keywords, all searched.** If one flags Low search volume, replace it - it can't serve.

3

Clone the ad group for the second vocabulary

Tick the ad group → Edit → Copy, then Edit → Paste into the same campaign. Rename the copy …_Purification. Then convert its keywords in bulk: select all → Edit → Change keyword text → Find and replace - filtration → purification, then again filter → purifier. All 22 purifier variants also had volume.

Copying the ad group via the Edit menu — **Copy, paste, rename.** Cheaper than rebuilding, and the settings carry over.

Find and replace panel changing filtration to purification — **Find-and-replace does 44 keyword edits in two passes.** The same tool exists for ads.

Why two ad groups

Filtration vs purification isn't pedantry: separate groups mean separate CPCs, conversion rates and cost-per-lead by vocabulary. That data feeds product naming, landing page language, even how sales talks about the product.

4

Hunt down the ad Google created behind your back

Even after removing the ad during the build, Google can auto-generate one - and because the account's history was all batteries, it generated a battery ad inside the filtration campaign. Open Ads in each new ad group and delete anything you didn't write, before the campaign leaves pending.

Deleting the automatically created battery ad — **The ambush, caught.** Same energy as Meta's Advantage+ auto-creative opt-ins: assume the platform tried, and check.

5

Create the ad: URL and display path

Final URL is the product landing page. The display path (two 15-character fields) should echo the keyword - /water-filter - purely cosmetic, purely for relevance. Google lists the ad group's keywords beside the headline fields: that list is your checklist.

Final URL, display path and keyword checklist in the ad builder — **The ad builder tells you what to cover.** Headlines that include these keywords are what ad relevance is scored on.

6

Write 15 headlines in three buckets

You get 15 headlines of 30 characters, up to 3 ads per ad group - 45 headline slots. A new campaign is not about finding the perfect headline; it's about variety the algorithm can optimise across. Capitalise Each Word - it reads professional. The buckets:

Bucket 1 - brand (×1): Smart Energy Water Filters. One is enough; "energy" isn't obviously relevant to filtration, so don't over-invest.
Bucket 2 - keyword mirrors (×5–6): Water Filter Systems · Water Filtration Systems · Home Water Filter Systems · Home Water Filtration · Whole House Water Filters · Whole House Water Filtration. These drive ad relevance; watch the green ticks appear as you cover each keyword.
Bucket 3 - USPs and CTAs (the rest): $0 Upfront Cost & 0% Interest · No Payments For 90 Days · WaterMark Certified Systems · Certified WaterMark Installer · Local, Professional Installers · 99.9% Removal Of Contaminants · Get Your Free Quote Today · Don't Settle For Hard Water. USP headlines can serve across nearly every ad in the account; steal proven winners from Meta (the payment terms converted best there).

Headline fields covering the ad group's keywords — **Bucket 1 and 2 in place.** Keyword chips turn green as headlines cover them.

USP headlines including certification, finance and installer proof — **Bucket 3: reasons to choose you.** Certification, finance terms, installer proof, an outcome, and two CTAs. Ad strength climbs as variety grows - it's a diversity meter, not a quality meter.

On pinning

You can pin a headline to position 1, but Google penalises the control you take - unpinned ads earn more impression share. A decent compromise pattern: per ad group, one ad with pinned headlines, one without, and let the data speak.

7

Write 4 descriptions (90 characters)

One on payment terms (the proven converter), one on the free quote + certification, one on the outcome (99.9% removal of chlorine & contaminants), one on local qualified installation teams. The "Value too long" error is your editor: trim until it fits.

Description fields with a value too long warning — **Four descriptions, ≤90 characters each.** Repeating the strongest offer across headline and description is fine - only fragments serve at once.

8

Add business name and logo, save

Attach the business name and logo assets, then save. First draft done for the Filtration group.

9

Copy the ad to the sibling group and re-vocabulary it

Copy the ad → open the Purification ad group → paste. Landing page stays; every filter word must become purifier/purification, because the ad must scream what the user typed. Try the bulk tool first: select the ad → Edit → Change ads → Find and replace (Filtration → Purification, in Headlines). "Purification" is three characters longer than "Filtration", so some 30-character headlines will overflow - finish those by hand, dropping a plural or a word ("…Systems" → "…System"). Search the ad for filter and filtration until only the landing page URL matches.

Pasting the copied ad into the purification ad group — **Same ad, new vocabulary.** Paste into the sibling group, then transform.

Find and replace within ad headlines — **Bulk first…** Edit → Change ads → Find and replace, scoped to Headlines.

Manually editing headlines to purifier variants — **…manual where characters overflow.** Every headline now says Purifier or Purification.

10

Build sitelinks per ad group

Sitelinks are the extra links under the ad - they exist to grow your real estate on the results page. Two per ad group to start: Water Filter Systems and Whole Home Water Filters (purifier versions for the other group), each with two 35-character description lines - reuse the $0 upfront / 0% interest and WaterMark lines; only one sitelink serves at a time, so duplicate descriptions are fine. Duplicate URLs across sitelinks aren't: Google blocks them so an ad can't inflate itself with the same link twice, and only the better performer would serve anyway.

Sitelink creation panel with live ad preview — **Sitelink anatomy:** 25-character link text, two 35-character description lines, final URL. The preview on the right shows the extra shelf they add to the ad.

The sitelink-sharing trap

Sitelinks are shared assets. When you edit the copied sitelink for the Purification group, choose "This asset instance only" - otherwise your edit rewrites the Filtration group's sitelink too, and one vocabulary silently overwrites the other.

Sitelink edit applied to this asset instance only — **The radio that saves you:** apply to this instance, not to all associations.

Definition of done

Campaign settings ✓ · one conversion action ✓ · observation audiences ✓ · exact-match keywords with volume in both ad groups ✓ · vocabulary-matched ad per group ✓ · vocabulary-matched sitelinks per group ✓ · no auto-generated anything ✓ · copy signed off ✓. Now it can spend.

Part 5

Account structure doctrine

The Smart Energy account is deliberately simple, and it has worked for two years. Campaigns carve up budget, hours and geography; ad groups carve up message.

Smart Energy - Google Ads │ ├── Search_Battery_In-Hours_NSW Mon–Fri business hours · NSW metro │ ├── _Generic plain battery terms - most volume lives here │ ├── _Quote cost / price / quote intent → quote copy + LP │ ├── _Tesla Tesla terms → Tesla copy + Tesla landing page │ ├── _Brands other brand terms │ ├── _Alpha Alpha ESS terms │ └── _Best "best battery" style research intent ├── Search_Battery_In-Hours_QLD same shape, QLD keywords + QLD copy ├── Search_Battery_Out-Hours_NSW evenings/weekends · lower budget ├── Search_Battery_Out-Hours_QLD ├── Search_Rebate_All-Hours_AU federal rebate demand · winding down │ ├── _Generic ├── _NSW ├── _QLD ├── _VIC ├── _ACT ├── _TAS └── _SA │ state keywords + state copy · no WA ├── PMAX_Battery_All-Hours_NSW-QLD one generic Performance Max · looser, lower lead quality └── Search_Filtration_In-Hours_NSW-QLD new category test (Part 3–4 build) ├── _Filtration └── _Purification

Campaign list showing naming convention, budgets and limited flags — **The live account.** Naming makes the structure self-documenting; the Status column carries the two flags the whole weekly routine hangs off.

Why in-hours vs out-of-hours campaigns

They used to be one campaign - and Google drifted spend to out-of-hours because those leads are cheaper to buy. But response time collapsed and close rates with it. Splitting the hours into separate campaigns puts the budget allocation back in your hands: feed the hours your sales team can actually work. It also unlocks a further play when you want it: out-of-hours-specific landing pages, SMS and email ("we'll call you at 9am tomorrow"), because you know exactly when that user came through.

The rebate lesson

The AU-wide rebate campaign printed volume while the federal incentive ran hot - but rebate searchers are price-motivated, and sales confirmed the leads were materially weaker. As the incentive ramped down, spend moved into the standalone battery campaigns and lead quality improved "infinitely". Keep the pattern: offer-chaser demand gets its own contained campaign so it never pollutes the product campaigns' data - and gets defunded without ceremony.

Rebate campaign ad groups split by state — **One campaign, state ad groups.** NSW keywords with NSW copy, QLD with QLD, and so on - relevance again, this time geographic. No WA group: the business doesn't service it.

Intent ad groups: the _Quote example

Quote ad group with cost and price keywords — **Every keyword in _Quote carries buying intent** - cost, price, quote - and the ad copy answers in kind. Note the "Rarely shown (low Quality Score)" flag: the signal that a keyword's relevance chain (keyword → ad → landing page) needs work.

Structure is getting harder to keep tight

Google keeps broadening what a keyword matches, so thin slices like _Quote bleed into each other more each year. Direction of travel: split ad groups only on very clear delineations, keep the campaign count flat (especially while the category is soft), and expect Google to keep pushing toward broad/keywordless targeting. More keywords inside the existing shape: yes. More shape: no.

Where Performance Max fits

One generic PMax campaign runs alongside search: wider inventory (YouTube, Gmail, Display), cheaper, lower lead quality. It's a volume top-up, never the backbone. It's also the campaign type where offline conversion data matters most (see Part 13) - the algorithm needs help finding qualified users in unqualified inventory. YouTube and Demand Gen have been tested and retired; retest only with real creative assets and a promo worth broadcasting.

Part 6

The weekly operating rhythm

Google is a high-touch platform - the work happens inside the ads UI, weekly, in this order. No external tools required; one custom column set tells the story.

Day	Job	What you're deciding
Monday	Budgets & bid targets	Full Mon–Sun week just closed. Reconcile performance (through to HubSpot appointment/sale), check budget utilisation and limited-by flags, rework budgets and targets.
Tue–Wed	Keywords	Search term report triage: add relevant terms as exact keywords, exclude irrelevant ones as negatives. Weekly ideal, fortnightly minimum.
Thu–Fri	Ad-hoc deep dive	Rotate: headline CTR analysis, landing page conversion rates, PMax placements, sequence experiments.

The column set (build once, save as default)

Read left to right and it narrates the account: spend → visibility → engagement → outcome → constraint.

Column	Why it's there
Cost	Total spend - start of the story.
Impressions	How often you showed.
Clicks	How often the ad won attention.
Avg. CPC	Price of a visit.
CTR	Ad copy health (clicks ÷ impressions).
Conversions	Leads.
Cost / conv.	The number the business feels.
Conv. rate	Landing page + lead quality health.
Target CPA	What you asked for - compare against Cost/conv.
Search impr. share	Your share of eligible auctions.
Search lost IS (budget)	The scaling headroom metric (under Competitive metrics).

Custom columns including search lost impression share to budget — **The story in one row:** NSW in-hours spends its budget, converts at target, and still loses 21.31% of eligible impressions to budget - that's the growth lever, quantified.

Part 7

Monday: budgets and bid targets

Two flags, two levers. Everything else on Monday is arithmetic.

Campaign list on a 7-day window with limited by budget banner — **Work off Last 7 days.** Google literally banners the constraint ("…currently limited by budget despite generating 5 conversions today"). The Status column shows which flag each campaign carries.

Limited by budget → the uncap formula

"Search lost IS (budget)" is the percentage of eligible auctions you missed because the budget ran out. Not gospel, but a very usable indicator. When a campaign performs (cost/lead profitable, HubSpot confirms downstream conversion), you want it at 0%. The loop:

Set the date range to Last 7 days.
Read lost IS to budget - say 32.19% on a $650/day budget.
Increase the budget by that proportion: 650 × 1.32 ≈ $859/day.
Re-check next Monday. Still above 0% and performance holding? Repeat. Performance plateaus or the flag clears? Stop.

Lost IS is a moving target: add 300 keywords, or a rebate announcement spikes demand, and the same budget is suddenly more limited. That's why it's checked every week, not set once. If total monthly budget is fixed, this becomes reallocation: pull from the campaign with the worst cost-per-sale, feed the one that's limited and converting.

Limited by target → the other lever

The PMax campaign has a $250/day budget but spent only ~$1,600 over the week - because its $80 target CPA was unrealistic against an actual ~$103 cost per conversion. Adding budget there changes nothing; the target is the constraint. Loosen the target to enter more auctions, tighten it to prioritise efficiency - knowing the trade is real in both directions.

Monday decision tree - "we want more volume"

Is the campaign flagged Limited by budget?
Yes → raise the budget by the lost-IS proportion. Monitor weekly until 0% or performance plateaus.
No → next question.

Is it flagged Limited by target (or spending under budget with actual CPA above target)?
Yes → relax the target CPA (e.g. $80 → $100+). Expect volume up, efficiency down. If the gap is small, nudge budget and target together.
No → the constraint is demand or coverage: add keywords from the search term report (Part 8) - and note added keywords will re-tighten the budget.

…or "we need efficiency"

Pull the target CPA down in steps (e.g. $80 → $70) and accept the volume cost. Decide as a team, weekly, which side of the trade the business is on - that call drives everything else.

Bidding change - live from 17 Aug 2026

Google is making targets binding: if actual CPA runs above your target, the campaign will throttle spend until it can hit the target, instead of overspending past it. Unrealistic targets now cost you volume silently. Review every target CPA against its actual (the two columns sit side by side in the column set) and keep targets honest.

Part 8

Tuesday–Wednesday: the search term report

The single most important recurring job in the account. Keywords are what you guessed people would type; search terms are what they actually typed. Close the gap in both directions, every week.

Search terms report with broad match recommendation card — **Insights and reports → Search terms**, scoped to one campaign. Ignore the "Add broad match keywords" card with its tempting +13.4% - that's Google selling broadness, the opposite of this workflow.

1

Filter to the untriaged

Set Added/Excluded: None so you only see terms you haven't already dealt with. Most rows will be close variants of your exact keywords - you're already showing for them; adding the good ones as keywords in their own right increases your eligibility to show.

Search terms filtered to added/excluded none — **Only the undecided.** Sort by impressions to see reach; **sort by cost when time-poor** - triage where the money actually went.

2

Add the relevant as keywords - into the right ad group

Tick relevant terms → Add as keyword. Route each to the tightly themed ad group (a cost term to _Quote, a Tesla term to _Tesla, plain terms to _Generic) - that's what keeps the ad copy relevant to the new keyword.

Add as keyword panel with ad group selector — **The ad group selector is the important field.** A keyword in the wrong group gets the wrong ad.

Change match types menu converting broad keywords to exact — **The broad-match trap, disarmed.** Terms added from this report default to broad match. Either wrap them in [brackets] as you add, or afterwards: filter keywords to Match type = Broad → select all → Edit → Change match types → Exact. Never the "Upgrade to broad match" option, whatever the Recommended tag says.

3

Exclude the irrelevant as negatives

Tick irrelevant terms → Add as negative keyword. Negatives default to exact - Google protects you from negative-broad wiping out your reach - and they stop budget bleeding immediately. Typical kills: brands you don't sell or service, other-product bleed, DIY/cheap intent.

Adding irrelevant search terms as negative keywords — **No decision is also fine.** You don't have to rule on every term - just never leave a money-burning irrelevant one alive.

4

Before excluding a big spender, check what it converted

Segment → Conversions → conversion action splits each term's conversions by action. With CRM stages piping in, a term that looks irrelevant but shows appointments or sales stays alive. Data before dogma.

Segmenting search terms by conversion action — **The full-funnel view, inside Google Ads.** This is most of what third-party attribution tools would tell you, free, provided the CRM integration is live.

Coverage doctrine

The account deliberately skips the biggest generic heads (solar battery, home battery) in favour of deeper-intent phrases (solar battery and panel package, solar battery installation). Close variants still catch the heads. Revisit that call when volume is needed - adding the heads as exact keywords buys eligibility, at the cost of intent purity.

Part 9

Late week: deep dives

One rotating analysis per week, not all of them every week.

Headline CTR analysis

Open an ad → View asset details → filter Asset type: Headline. For nearly a decade the only headline data was impression volume; Google now shows performance per headline, and CTR is the metric that matters - a headline's whole job is earning the click. There's no bulk view: it's ad by ad, manual, and worth it. Feed what you learn back into Bucket 3 headlines everywhere (and into Meta creative).

Asset details showing CTR per headline — **Which message earns the click.** "$0 Upfront Cost & 0% Interest" vs "Get A Free Quote Now" vs "5,000+ Battery Installations" - the CTR column settles the argument. This kind of table is also the perfect thing to export to an AI tool for analysis; just never hand it the keys to make account changes.

Landing page conversion

Conversion rate by landing page (Insights → Landing pages). Vocabulary-matched pages (purifier page for purifier terms) are the next level of the relevance machine once volume justifies them.

PMax placements

Where PMax actually served. Wide inventory means junk placements creep in; prune what you can, and watch lead quality with sales.

Part 10

The bidding ladder and experiments

Bid strategy is a spectrum - clicks, conversions, conversion value - and campaigns climb it as data accrues.

Rung	Strategy	When
1	Maximise clicks	Launch. No conversion history; the job is traffic and data.
2	Maximise conversions (no target)	Conversions are registering; let the bidder chase them unconstrained while you learn the natural CPA.
3	Maximise conversions + target CPA	You know what a lead is worth. The target becomes the weekly efficiency lever (Part 7).
4	Value-based bidding (tROAS / values on stages)	Only with CRM stages piping values back, and only where volume supports it. Rarely for search; genuinely useful for PMax.

Value bidding can go backwards

Accounts have gone "backwards in a very big way" jumping to value bidding: sales-stage events are far rarer than leads (you will not see 30+ sales per campaign per month), so the algorithm starves. If you try it, hybrid it - optimise for all lead steps weighted by value - and test it in one geo against tCPA in the other before believing it.

Experiments: settle arguments with a 50/50

Campaigns›Experiments splits one campaign's traffic randomly into control and treatment and reads the result to statistical significance. It beats geo tests (QLD vs NSW carries noise a random split doesn't). Test one variable at a time: bid strategy is the canonical example - NSW in-hours, 50% tCPA vs 50% value bidding, one month.

Experiments page with create experiment cards — **The right way to try anything Google is selling.** Note the AI Max experiment card front and centre - if you ever test AI Max, this contained 50/50 is the only sane way to do it.

Part 11

Bing (Microsoft Ads)

Bing copied Google's UI nearly screen for screen - everything in this playbook transfers. The differences are all in your favour.

Sequence: validate on Google first (traffic exists, CPC economic, leads convert), then extend to Bing "probably quite quickly". Bing is the cheap second market for proven search demand.
Same UI, same weekly routine: campaigns → ad groups → keywords → search term report. Budgets Monday, keywords midweek.
More manual, more targeted: Microsoft isn't pushing the broad-match/AI agenda nearly as hard. Exact-match discipline still works beautifully there.
Leaner structure: volume is a fraction of Google's, so consolidate harder - Smart Energy runs one battery campaign and one rebate campaign with merged geo targeting, so conversion data isn't spread thin.
Do not use the Google import tool. It exists and it's one click - but it ties the accounts together so Google changes pipe through to Bing. Different structure, different budgets, different decisions: rebuild manually.

Part 12

Meta, and how it differs

Running both well means holding two opposite mental models at once.

	Google	Meta
Targeting	Keywords you choose and prune	The creative is the targeting; go broad and let the algorithm find buyers
Lead quality	Highest anywhere - active intent	Lower per lead; wins on volume and cost
Workload	High-touch, in-platform, weekly optimisation	"Insanely" low-touch - decent creative in a broad ad set does the work
Learning signal	~30 conversions/campaign/month (soft, black-box guidance)	~50/week per ad set to exit learning (explicit)
Structure	Consolidated but deliberate: hours, geo, intent themes	Minimal: few campaigns, broad ad sets - fragmenting data is the same sin on both
Platform tricks to resist	AI Max, auto-ads, broad-match defaults, suggested budgets	Advantage+ auto-creative opt-ins - same ambush, different logo
Community consensus	None - accounts differ; run experiments	Broad consensus (creative-led, broad targeting); CBO vs ABO still account-specific - both win somewhere
What transfers over	Winning messages. Meta's proven converters ($0 upfront & 0% interest, no payments for 90 days) became Google USP headlines; Google's headline-CTR winners feed Meta creative back.

Part 13

CRM data and Google reps

HubSpot → Google offline conversions

Two distinct payoffs once stages sync:

Reporting. Appointment and sale conversions appear against campaigns, keywords and individual search terms inside the ads UI - the segment-by-conversion view in Part 8 becomes a full-funnel report. This alone changes exclusion decisions.
Optimisation (later, carefully). Bid toward users likelier to reach deep stages - the value-bidding rung of Part 10, most useful for PMax. Keep optimising search to lead until volume genuinely supports more.

Keep optimisation on lead; let appointment/sale flow in for reporting from day one.

Handling the Google account manager

Genuinely useful for competitor insights, category growth/shrink data, product updates. But account managers are compensated on product adoption: every call will push the newest product (AI Max, broad match, PMax expansions) - usually meaning wasted spend and loose guardrails. House rule: take the insights, log the pitches, adopt nothing on the call. Anything worth trying goes through a 50/50 experiment on your terms. (Same posture as Meta reps.)

AI tools on the account

Connecting Claude or similar via MCP for analysis - headline CTR tables, search-term summaries, report generation - is a great fit. Handing an AI the reins to change targeting is not: too much context lives behind every keyword decision. Analysis yes, account control no.

Part 14

Checklists

New campaign launch

Landing page live, conversion tracking verified, lead-handling rostered
Keyword Planner research done; themes mapped; ≥1,000-searches terms structured
One conversion action selected (lead for this product)
Search type · website visits only · naming convention applied
Maximise clicks, no target CPA
Search partners on · Display off
Metros by name · Presence only · English
Observation audiences loaded (homeowners, industries, affinity set, in-market match)
Schedule matches sales hours (in-hours: Mon–Fri 8:00–17:00)
AI Max off · generated keywords cleared · exact-match list pasted
Custom budget set (ignore suggestions)
Ad groups named by theme; keywords volume-checked
Auto-created ads deleted from every ad group
Per ad group: 15 headlines (brand / keyword-mirror / USP), 4 descriptions, display path, logo
Sitelinks ×2 per ad group, vocabulary-matched, instance-only edits
Copy signed off before spend

Weekly

Mon - 7-day view: reconcile performance to HubSpot appointment/sale
Mon - limited-by-budget campaigns: uncap by lost-IS % (if performing)
Mon - limited-by-target campaigns: review target vs actual CPA; volume/efficiency call made as a team
Tue/Wed - search term report per campaign: filter Added/Excluded = None (sort by cost if short on time)
Tue/Wed - relevant terms → exact keywords, routed to the themed ad group; broad additions converted to exact
Tue/Wed - irrelevant terms → negative keywords (check conversion segments on big spenders first)
Thu/Fri - one deep dive: headline CTR / landing pages / PMax placements / experiment review
Anytime - recommendations and rep suggestions logged, nothing auto-applied

Never

AI Max on a live campaign (contained experiment only, if ever)
Broad match keywords left unconverted
"Presence or interest" location setting
Display Network on a search campaign
Auto-applied recommendations
Publishing ads Google generated
New campaigns/ad groups without a clear delineation to justify the split
Budget added to a target-limited campaign (fix the target)
Google→Bing import tool (accounts become chained)
Adopting a rep pitch on the call