Google Ads is no longer a platform won by keyword hacking. It is an algorithmic auction governed by machine learning bidding signals, conversion data density, and strict ad relevancy scores.
The Death of SKAG and the Rise of STAG
Years ago, PPC managers used Single Keyword Ad Groups (SKAG) to force match types. When Google redefined exact match to include "close variants", SKAG architectures exploded into unmanageable accounts with internal cannibalization.
The modern standard is Single-Theme Ad Groups (STAG). Under STAG, an ad group contains 3 to 6 tightly clustered keyword variants that share exact commercial intent. This concentrates impression data, gives the Smart Bidding algorithm clean signals, and allows Responsive Search Ads (RSAs) to achieve 90%+ ad relevance scores.
How a STAG Is Built in Practice
One test decides every grouping question. If two keywords can be served honestly by the same ad copy and the same landing page, they share a STAG. If either answer changes, the group splits.
- Group by buying intent: "Emergency plumber" and "24 hour plumber" describe one theme with one mindset, so they sit together even though the wording differs.
- Hold the keyword count at 3 to 6: Each STAG carries a small set of exact and phrase variants around a single service-and-location or product-and-use-case core.
- Write 2 to 3 RSAs per group: Build the full 15-headline allowance, pin only the headlines that must hold position (brand, core service, location), and leave the rest free for the algorithm to rotate.
- Name groups for diagnosis: A convention like STAG | Emergency Plumbing | Exact lets any manager read the account structure in seconds and spot drift early.
Ad strength should land at Good or Excellent before launch. Treat a Poor rating as a copy problem. Weak relevance drags Quality Score down, raises the CPC you pay for the same position, and quietly taxes every click the campaign buys from that point forward.
Telemetry Defense: Enhanced Conversions (SHA-256)
Third-party cookies are dying and browser tracking protections strip standard URL query parameters. Relying exclusively on standard client-side pixel fires leaves 25% to 35% of conversion events unaccounted for.
We implement Enhanced Conversions: when a customer submits a lead form or completes a checkout, their first-party data (hashed with SHA-256) is securely transmitted back to Google. This enables offline conversion imports and gives the bidding algorithm the true signal needed to optimize for actual revenue, not superficial form clicks.
Setup quality decides whether Enhanced Conversions helps or misleads. Capture email and phone at the form, hash the values in the browser or server-side through Google Tag Manager, and pass them with the conversion event on every submission. Check the Diagnostics tab weekly: a healthy implementation shows the feature active with a steady stream of matched records, while a broken one usually traces back to a form change that quietly dropped the data layer fields. Then close the loop with offline imports. Send CRM outcomes back to Google Ads within 24 to 48 hours of each status change, while the click identifiers are still fresh, and record qualified and closed stages as separate conversion actions with real values attached. The bidding system can only price a lead correctly when it can see what that lead became.
The 90-Day Scaling Playbook
- Days 1–30 (Calibration & Negative Defenses): Launch high-intent exact and phrase match STAGs under Maximize Clicks with manual ceiling bids. Build a universal negative keyword list to eliminate non-commercial query waste.
- Days 31–60 (Algorithmic Transition): Once 30+ clean conversions are logged, transition to Target CPA or Target ROAS bidding. Let Google's Smart Bidding target users with high propensity to convert.
- Days 61–90 (Scaling & Incremental Volume): Introduce targeted Performance Max campaigns with strict negative brand lists to capture high-converting inventory across YouTube, Search, and Gmail.
Three guardrails protect the whole 90 days. Change budgets in steps of 20% to 30% at a time, because larger jumps reset the learning period and scramble the data you just paid for. Judge every phase on a full conversion window, since B2B leads often convert days after the click that started them. And freeze structural edits during learning periods: one variable at a time, or you will never know which change moved the number.
Negative Keyword Architecture: The Cheapest Budget You Will Ever Recover
Audits of inherited accounts surface the same pattern again and again: a fifth to a third of spend landing on queries with no commercial intent at all. Job searches. DIY searches. Freebie hunters. STAG structure concentrates your budget; negatives decide where that budget is never allowed to go.
- Account-level universal list: Block jobs, careers, DIY, free, and informational modifiers once, at account level, so every new campaign inherits the defense from day one.
- Campaign-level themes: Add negatives for services you do not sell and locations you do not serve; a mature local account usually carries several hundred negatives by the end of month two.
- Weekly search-term review: Any query that has spent roughly 1.5 times your target cost per acquisition with zero conversions gets negated, or split into its own STAG if the intent behind it is real.
- Brand separation: Negate your own brand terms inside non-brand campaigns, so generic STAGs never pay premium CPCs for demand the brand campaign would have captured anyway.
This work feels unglamorous. It is also the highest-return hour in account management, because every blocked query is budget moved from a browser to a buyer.
Budget and Bidding Economics: How Much Is Enough?
Smart Bidding is a statistics engine, and statistics engines starve on thin data. As a working rule, a campaign needs around 30 conversions in 30 days before Target CPA holds steady, and roughly double that volume before Target ROAS behaves. Below those thresholds, consolidate: fewer campaigns, shared budgets, and themes merged until each bidding pool clears the bar.
Set the media budget from the unit economics backwards. If a closed customer is worth $4,000 in gross margin and one in five qualified leads closes, a lead is worth $800; a 25% marketing cost ratio then caps your target CPA near $200. Most advertisers never work the sum this honestly, which is why budgets get set by gut feel and judged by mood. Run your own numbers through our ROAS and profit calculator before setting a single bid.
Watch impression share on your core exact-match STAGs. Losing share to budget on a profitable theme means money is being left on the table. Losing share to rank at a healthy CPA usually means the bid ceiling, or the landing page behind it, needs attention first.
Five STAG Mistakes That Quietly Burn Budget
- Two intents, one group: Mixing research-stage and buy-now keywords in a single STAG forces one ad to serve two mindsets, and relevance scores punish the compromise.
- Broad match on day one: Broad match without conversion history sends the algorithm prospecting with your money; earn the data on exact and phrase first.
- A target CPA set by hope: Setting tCPA 40% below your historical cost per conversion starves delivery within days, then gets misdiagnosed as a market problem.
- Twelve campaigns, one small budget: Fragmented budgets mean no single campaign reaches the conversion volume Smart Bidding needs to stabilize.
- Weekly restructures: Rebuilding ad groups every week resets learning and destroys the comparison data that tells you what is actually improving.
How to Tell Whether the Architecture Is Working
Read the account in a fixed order. First, the search terms report: at least 80% of spend should sit on queries you would happily buy again. Second, conversion volume and cost trend over rolling 30 days, compared against the pre-restructure baseline rather than last week. Third, the Enhanced Conversions diagnostics, because bidding quality collapses silently when tracking breaks. Fourth, impression share lost to budget on your top three revenue STAGs.
Review structure monthly and copy quarterly. Accounts managed this way stop producing surprise invoices, and scaling becomes a decision you make from evidence. If your current setup needs that evidence base built first, our Google Ads management engagement starts with exactly this audit.
Google Ads STAG FAQs
How many keywords should a single STAG contain?
Three to six closely related variants is the working range for most accounts. The ceiling is set by intent rather than count: the moment a keyword needs different ad copy or a different landing page to convert well, it belongs in its own group.
How long does Smart Bidding take to stabilize?
Plan on two to four weeks after the switch, provided conversion volume holds near 30 per month. Every major budget or target change restarts a shorter learning cycle, which is why we batch changes instead of dribbling them out daily.
Should brand keywords live inside the STAG structure?
Brand terms belong in their own campaign with their own budget cap. Brand demand behaves differently, converts at a different rate, and will distort the bidding signals of your generic STAGs if the two are blended together.
What is the minimum budget to run this architecture?
Enough to buy roughly 10 to 15 clicks a day on your core themes. At a $12 average CPC that is about $150 a day; at $3 CPC it is closer to $40. Below that level, narrow the theme count until each surviving STAG can still gather meaningful data, and run paid search alongside Meta Ads rather than as your only pipeline.