Paid search and SEM

TL;DR

Paid search is a compounding program of account structure, keyword strategy, ad copy testing, landing page match, and Quality Score discipline. A working account keeps cost per acquisition flat while scaling spend by 3 to 5x. The failure mode is turning on Performance Max, defaulting all match types to broad, and letting Google's automation run without the structural guardrails that keep spend on-intent.

The playbook, in one paragraph

Structure the account around actual buyer intent (brand, category, product, competitor, informational), build ad groups tight enough that ad copy matches search intent one-to-one, write three or more responsive search ads per ad group with pinned headlines where relevance demands it, protect Quality Score with landing page match and negative keyword hygiene, use Performance Max as a channel within a structured account rather than the whole account, and report on conversion metrics tied to actual revenue (not clicks, not impression share). Google's automation rewards accounts that give it clean data and punishes accounts that give it garbage.

Where this fits in the modern discovery layer

Paid search buys placement on the highest intent surface Google runs. A person searching "commercial roofers in Austin" is not researching. They are hiring. Paid search is the surface that pays back in weeks rather than months because the buyer has already decided to buy something.

Of the 19 surfaces in the Playbook, paid search interacts with three directly. SEO (paid and organic compete for the same intent, and Google Search Console data informs paid keyword expansion). LSO (Google Ads local campaigns and Google Business Profile are the same buyer moment expressed through two Google surfaces). And CWV (landing page speed is a Quality Score input, meaning slow landing pages cost more per click regardless of the ad).

The strategic point: paid search is the fastest lever, and the most expensive one to run wrong. A well structured account scales 3 to 5x with flat CPA. A poorly structured account scales spend at rising CPA until the budget stops making sense. The structural work in the first 60 days determines whether the account compounds or bleeds.

The five levers

1. Account structure that respects intent

Every account is structured around intent tiers: brand (defending on your own terms), category (broad head terms where you want to compete), product or service specific (the actual thing you sell), competitor (bidding on competitor names when the math works), and informational (research stage queries, usually deprioritized). Each tier gets its own campaign, its own budget, its own bid strategy, and its own reporting. Mixing intents inside one campaign lets Google spend where it is easiest, not where it produces the best return.

2. Match types that match the account's data volume

Broad match is Google's preferred setting because it maximizes reach and lets the algorithm learn. Broad match is also how accounts bleed budget on irrelevant searches when they lack the conversion volume for Google's algorithm to learn from. For accounts under 100 conversions per month, phrase match on head terms and exact match on high intent terms is safer. For accounts with strong conversion volume, broad match with tight negative keyword lists works. The right answer depends on the data, not on the default.

3. Quality Score as the north star

Every keyword has a Quality Score of 1 to 10 based on expected CTR, ad relevance, and landing page experience. Improving a keyword's Quality Score from 5 to 8 typically cuts CPC by 30 to 50 percent at the same position. Improving landing page match is the highest leverage Quality Score input for most accounts. Tight ad group structure, matching ad copy, and a landing page that answers the search query are the three moves that move Quality Score.

4. Performance Max in its lane

Performance Max is a powerful channel for ecommerce with a healthy product feed. It is a mediocre channel for B2B lead generation and a wasteful channel for accounts without conversion volume. When I use it, brand terms are excluded (via campaign level exclusion), asset groups are tight, audience signals are populated with real intent data, and results are measured against a proper attribution model rather than against Google's default reporting.

5. Attribution the CFO will believe

Google's data driven attribution model for accounts with the conversion volume to support it. Position based or time decay for smaller accounts. Google Ads conversion tracking reconciled monthly against GA4 and against the CRM or order management system. A conversion count that only lives in Google Ads and does not match the source of truth is a conversion count nobody trusts.

First 30 / 60 / 90 days

Days 1 to 30: audit and restructure

Full account audit. Campaign structure documented. Ad group sizes measured (ad groups over 30 keywords usually need to split). Match type distribution reviewed. Search terms report exported for the last 90 days to identify wasted spend and negative keyword opportunities. Quality Score distribution across the top 50 keywords documented.

Conversion tracking audit. Are conversions firing correctly. Are conversion values populated. Is Enhanced Conversions installed. Is the GA4 to Google Ads link working. Is the CRM feeding back offline conversions for lead gen accounts.

Landing page audit. Which landing pages get the paid traffic. Do they match the search intent. Are they mobile fast. Do they convert against the account baseline.

Metric moving in month one: wasted spend identified and stopped. Typical accounts have 15 to 30 percent of spend going to irrelevant search terms or broken conversion setups. Stopping the bleeding funds the restructure.

Days 31 to 60: restructure and Quality Score

Account restructured into intent tiers. Brand campaign with brand only. Category campaigns for the head terms. Product or service specific campaigns for high intent transactional queries. Competitor campaigns where the math supports them. Each campaign with its own budget and its own bid strategy.

Ad groups tightened. Every ad group should be tight enough that a single responsive search ad matches every keyword in the group. This means ad groups of 5 to 15 keywords, not 50.

Responsive search ads rewritten. Three or more per ad group, testing headline and description permutations. Pinned headlines where regulatory or brand consistency requires. Ad strength minimum "Good" on every active ad.

Landing pages aligned. Where a campaign targets specific keywords, the landing page reflects those keywords in the H1, the first paragraph, and the CTA. Quality Score follows relevance.

Metric moving in month two: Quality Score average across the top 50 keywords. Target 7 or higher. CPC decline follows Quality Score improvement.

Days 61 to 90: scale and automate

Bid strategies shifted where appropriate. Target CPA for lead gen accounts with 30+ monthly conversions. Target ROAS for ecommerce accounts with 30+ monthly transactions. Manual CPC for small accounts that lack conversion volume for automation to work.

Performance Max launched where the account supports it. Product feed audited (Shopping accounts). Asset groups tightened. Brand exclusions in place. Reporting broken out separately so PMax performance is visible against other campaign types.

Reporting cadence established. Weekly metrics review covering spend, conversions, CPA, ROAS, wasted spend on search terms, Quality Score, impression share on priority terms. Monthly report to stakeholders with revenue attribution reconciled against the CRM or GA4.

Deliverable at day 90: a structured account, Quality Score at 7 or higher on the top keywords, wasted spend reduced by half, an automation strategy in place, and a clear roadmap for months four through twelve (feed optimization, audience layering, YouTube integration, offline conversion optimization).

Tools I use

Google Ads is the platform. Every serious paid search program lives here. I run accounts through the Google Ads interface directly rather than through third party tools because the automation, machine learning, and reporting live in the native platform.

Google Ads Editor for bulk changes. Restructuring a large account, adding hundreds of negative keywords, changing bid strategies across many campaigns at once. The web interface is not built for this.

GA4 for the behavior data behind the click. GA4 to Google Ads link populates conversion data, informs audience creation, and reconciles Google Ads reported conversions against session behavior.

Google Search Console as an input to paid keyword strategy. Which queries the site already ranks for organically, which have low CTR (potential paid opportunity), and which have high position but low click volume.

SEMrush and Ahrefs for competitive intelligence: which terms competitors bid on, their estimated spend, their ad copy patterns. Neither is perfectly accurate. Both are useful directional signal.

Google Tag Manager for conversion pixel management, event tracking, and Enhanced Conversions setup. GTM keeps the tracking layer out of engineering's deploy queue.

Google Sheets for the search term report analysis, negative keyword lists, and monthly reporting. Google Ads reporting is fine for a first look. Sheets is where the actual analysis happens.

The CRM (HubSpot, Salesforce) for offline conversion imports on lead gen accounts. Feeding closed-won data back to Google Ads teaches the algorithm to bid for revenue, not just for form fills.

What kills the program

1. Broad match without conversion volume

The account is under 30 conversions per month. Every keyword is on broad match because Google recommended it. Half the spend is going to irrelevant searches Google matched to the keywords. The account bleeds until match types get tightened or negatives get built. Broad match is a tool for accounts with the data to feed it, not a default for every account.

2. Performance Max as the whole account

The founder or agency turned on PMax and paused all other campaigns because PMax "does everything." PMax cannibalizes brand searches, ships spend to Display where intent is weakest, and hides which asset groups are driving results. PMax as a channel inside a structured account works. PMax as the whole account rarely does.

3. Ignoring the search terms report

The search terms report is the single most valuable diagnostic in Google Ads. Weekly review pulls out negative keyword opportunities, new expansion terms, and match type problems. Accounts that ignore the search terms report are accounts that quietly pay for irrelevant clicks month after month.

4. Landing pages that do not match ad copy

The ad says "commercial roofing repair in Austin." The landing page is the homepage, which mentions residential roofing, siding, and gutters, and asks the visitor to scroll to find the service they searched for. Quality Score drops. CPC rises. Conversion rate falls. Matching landing pages to search queries is the highest leverage move in most accounts.

5. Optimizing for clicks instead of conversions

The account is optimized for CTR because CTR is the easiest metric to move. Conversions are flat or declining. Google will happily send you clicks that never convert if that is what the bid strategy asks for. Optimization goals should be conversion metrics, and conversion metrics should be tied to actual revenue.

6. Never reconciling against the CRM

Google Ads reports 200 conversions. The CRM shows 87 closed-won. The account is being optimized to a phantom conversion number that includes duplicates, junk leads, and bad data. Monthly reconciliation between Google Ads, GA4, and the CRM prevents this.

KPIs that matter

Cost per acquisition (CPA) or return on ad spend (ROAS). The primary metric. CPA for lead gen, ROAS for ecommerce. Trending flat or better as spend scales.

Conversion rate by campaign. Diagnostic for which campaigns are worth scaling. Under 1 percent for a lead gen campaign or under 2 percent for an ecommerce campaign is a signal to fix the landing page or tighten the targeting.

Quality Score distribution. Percentage of active keywords at Quality Score 7 or higher. Target 60 percent or higher. Under 40 percent means the account structure or landing pages need work.

Impression share on priority terms. Are we winning our share of auctions on the terms that matter. Under 60 percent on brand terms is a warning sign. Under 30 percent on category head terms may indicate budget constraint or bid strategy misalignment.

Wasted spend percentage. Spend on search terms that did not convert and are not relevant. Target under 10 percent. Above 20 percent means the negative keyword list is stale or match types are too broad.

Assisted conversions. How often paid search touched a conversion that closed on another channel. Especially important for B2B where paid search often assists rather than closes.

Revenue reconciliation delta. Difference between Google Ads reported revenue and CRM or ecommerce reported revenue. Should be within 5 to 10 percent. Larger gaps mean tracking is broken somewhere.

FAQ

Should we run Performance Max?

For ecommerce with a healthy product feed and conversion history, yes, with brand terms excluded and a proper account structure surrounding it. For lead gen with a small conversion volume, no. Performance Max needs data to work. Below 30 conversions per month, it wastes budget hitting the wrong intent.

How is Quality Score calculated and why does it matter?

Quality Score is Google's 1 to 10 rating of expected click-through rate, ad relevance, and landing page experience for each keyword. Higher Quality Score means lower cost per click and better ad position at the same bid. Improving Quality Score from 5 to 8 typically cuts CPC by 30 to 50 percent.

How much should we spend on Google Ads to start?

Minimum meaningful budget is enough traffic to generate 30 conversions per month at your current conversion rate. Below that, the algorithm cannot optimize and reporting is noise. For most B2B this means $3K to $10K per month. For ecommerce with cheaper CPCs, $2K to $5K.

Should we bid on our brand terms?

Yes, if competitors bid on them. Otherwise a competitor is buying visits from users searching your name. If nobody bids on your brand, brand campaigns are a small tax on traffic you would have gotten free. Test pausing brand for two weeks and measure organic pickup.

What is the right attribution model?

Google's data driven attribution model for accounts with sufficient conversion volume. Position based or time decay for accounts with lower volume. Never last click except as a diagnostic. Last click drastically undervalues top of funnel touches and shifts budget toward the wrong campaigns.

Why is our cost per lead going up?

Usually one of four reasons. Competitor entry into your terms, seasonal auction pressure, Quality Score decline from ad or landing page changes, or match type drift on broad match keywords bringing in irrelevant traffic. Diagnostic is a search terms report and a Quality Score audit.

Related reading

• The full writing archive
• Conversion rate optimization playbook
• Content marketing operations playbook
• Email lifecycle marketing playbook
• All case studies and playbooks