Skip to content
Ecomwith

Google Search Campaigns: Keywords, Ad Groups, and Landing Pages

  • Beginner
  • 55 minutes

update: Added the Search language-matching change starting September 2026 and updated RSA asset-report reading and AI Max boundaries.

Give the first Search campaign one clear question: can people searching for this product move from a relevant ad to a matching page and make a purchase you can verify? Connect the keyword cluster, ad group, ad promise, and landing page before expanding the list.

Start with the product and the question

The example is a Shopify store selling a 20oz leak-proof commuter tumbler, with English-speaking buyers in Canada as its planned market. The product page has a price and a buy button. The owner still needs to check whether seal details, cup-holder dimensions, shipping, and returns answer the buyer’s question. Market choices and product or return promises here are classroom inputs that need real product and policy support before use.

Search structure case

Product
Leak-proof commuter tumbler
Capacity (oz)
20
Planned market and communication language
Canada / English
Structure question
One intent → one ad group → one page

The primary conversion is the main business action measured this round, such as a purchase. Value and currency record its amount and denomination; order deduplication keeps the same order from being counted twice. Verify those fields against the business record before spending. You can draft below while that work is unfinished, but keep launch on hold. Reconcile the definitions and windows used by Google Ads, GA4, and Shopify orders. Even 200 clicks will be hard to diagnose if brand, generic, and problem queries all share one ad and page.

Series start

Build on a verified business signal

Account and conversion QA asks whether the recorded business action is trustworthy. This lesson asks which searches one ad group and page can answer. You can complete the structure exercise independently; launch still requires verified measurement.

Define the five parts of the structure

A keyword is a trigger you submit; a search term is what the user actually searched. Meaning can connect the two, so review both rather than treating your keyword list as a list of actual traffic. The terms below describe campaign structure, matching, and copy.

Terms used in the build

Search campaign

A Search campaign is a Google Ads campaign triggered by user search queries. It helps you examine demand people are already expressing in search.

Where you see it You choose the Search campaign type in Google Ads, then set goals, budget, keywords, ads, and landing pages.If misread If the first Search campaign mixes brand, non-brand, competitor, and generic terms, the readout becomes hard to interpret.Example: A leak-proof commuter tumbler can use non-brand Search to capture clear demand such as "leak proof travel mug."
Ad group

An ad group organizes keywords and ads inside a Search campaign. Each ad group should address a set of closely related searches.

Where you see it You add keywords in an ad group and write responsive search ads matched to that keyword set.If misread When intents are mixed, one ad and one page must answer too many different searches, making the source of poor performance difficult to isolate.Example: "waterproof dog seat cover" and "dog hammock for car" may fit one group; "free dog seat cover" should be excluded or held.
Phrase match / Exact match

Phrase match can match searches that include the meaning of the keyword; Exact match steers closer to the same meaning or intent. They make the first query sample easier to inspect.

Where you see it You use quotation marks for Phrase match and brackets for Exact match when adding keywords.If misread If match scope is too broad, spend can flow into queries the page cannot answer.Example: Start with "waterproof travel mug" and [leak proof travel mug], then expand from search term evidence.
Responsive Search Ads / RSA

Responsive Search Ads combine multiple headlines and descriptions. Use them to state the product, problem, supporting evidence, and reason to click.

Where you see it You enter multiple headlines and descriptions when creating a Search ad.If misread Generic copy may create many combinations, but none of them necessarily answers the search intent.Example: "Leak-proof commuter mug," "fits car cup holders," and "30-day returns" are closer to purchase intent than "Premium Lifestyle Accessory."
Negative keyword

Negative keywords stop ads from showing on irrelevant searches. They are a first-campaign budget guardrail, not just a later cleanup.

Where you see it Negative keywords go at campaign, ad group, or account level; account-level negatives also reach Shopping and PMax inventory, so confirm the scope first.If misread Without starter negatives, free, tutorial, job, support, wholesale, or irrelevant-location queries can spend first.Example: When selling tumblers, first evaluate excluding free, DIY, repair, jobs, used, and other poor-fit searches.

Trace the goal through to the page

The five “layers” below are review questions, not five additional levels in the Google Ads account hierarchy. Follow the same buying question through the goal, intent, keywords, ad and page. A CTA is the call-to-action button you want the shopper to choose, such as Add to cart on the product page.

Give one campaign a primary job, such as non-brand product purchases. Keep its goal, market, budget, and bidding logic explainable. Many tiny campaigns can leave a small budget with thin samples everywhere. Read established brand demand separately. Use a suitable campaign or ad-group boundary; use separate campaigns when you need separate budgets, because ad groups share their campaign budget.

Search creation offers Sales, Leads, and Website traffic as objective choices. Choose the entry that fits this round’s purchase, lead, or visit task, then check the actual conversion goals. Use the labels and steps currently available in the account.

Five layers from goal to landing page

  • Campaign goal

    Job
    Answer one main question, such as non-brand purchase demand.
    A usable answer
    Goal, market, language, budget, and brand boundary are clear.
    Common break
    Brand, competitor, generic, and problem terms are blended.
  • Ad group intent

    Job
    Each group carries one close search intent.
    A usable answer
    One ad and one page can answer the group.
    Common break
    Using one ad group as a keyword warehouse.
  • Keywords and match type

    Job
    Start with high-intent, narrow, explainable keywords.
    A usable answer
    Start with Phrase / Exact and document held terms.
    Common break
    Expanding too broad on day one.
  • RSA promise

    Job
    Headlines and descriptions match the query intent.
    A usable answer
    Covers product, problem, supporting information, and reason to click.
    Common break
    One generic ad for every keyword.
  • Landing page and negatives

    Job
    The landing page supports the ad promise above the fold; negative keywords block low-intent queries.
    A usable answer
    Page, negatives, pause/expand rules are documented.
    Common break
    Specific ads send traffic to a generic homepage.

The worked example checks only goal and intent: 2/5. It means the goal and search intent are written, while keyword, promise, and page work remains. The checklist below is your own evidence and starts empty on first use or after clearing. The example ticks do not count as your completed work.

Checklist0 / 5 items
My five-layer structure checks
  • Brand, competitor, generic, and problem terms are blended.
  • Using one ad group as a keyword warehouse.
  • Expanding too broad on day one.
  • One generic ad for every keyword.
  • Specific ads send traffic to a generic homepage.

The count records your ticks. Each one still needs a field, page, or business record behind it before it can support a launch decision. Save these personal ticks in the page selections and structure group near the end.

Group by the buyer’s question

Start candidates from three places: the product type, a problem or use case, and the main value propositions already explained on the page. Keyword Planner helps discover candidates and forecasts; page fit decides what belongs. Terms can start together when one RSA and one page answer them. Split, hold, or support terms that need a different promise, page, or business condition.

Five common keyword clusters

  • Category purchase terms

    Example searches
    "waterproof dog seat cover"; "leak proof travel mug"
    Page
    Relevant product page or tight collection page
    Treatment and reason
    go; Can directly validate purchase intent.
  • Problem / use-case terms

    Example searches
    "keep car clean from dog hair"; "coffee mug for commute"
    Page
    Product page that explains the use case and solution
    Treatment and reason
    go; If the page can answer it, this can be a second intent group.
  • Brand terms

    Example searches
    "your brand travel mug"; "your brand dog seat cover"
    Page
    Brand page or most relevant product page
    Treatment and reason
    separate; Evaluate brand traffic separately so it does not hide non-brand performance.
  • Competitor terms

    Example searches
    "competitor alternative"; "brand x vs brand y"
    Page
    Needs a compliant, truthful comparison page
    Treatment and reason
    hold; In the first round, hold these terms unless the intent and compliance boundaries are already clear.
  • Low-intent / poor-fit terms

    Example searches
    free; DIY; repair; jobs; used
    Page
    Should not enter first purchase validation
    Treatment and reason
    hold; Use starter negatives or hold, so budget does not spend here first.
Exercise · pick one

Choose the demand to answer

Example queries
"waterproof dog seat cover"; "leak proof travel mug"
Ad-group boundary
Can become a first-round core ad group because the searcher already named the product type.
Matching start
Start with Phrase / Exact to protect the search-term sample before broad expansion.
Ad promise
RSA should state product type, key supporting detail, and shipping or return reason.
Landing page
Use the closest product page or tight collection page; the above-the-fold section needs the product, specifications, key information, and a buying path.
Negative candidates
free; used; DIY; repair; jobs
First readout
Read whether search terms still show buying intent, CTR is reasonable, and add_to_cart / purchase reconcile with Shopify.
Keep this held
Do not mix this group with broad content terms, competitor terms, or support searches.

Read the page requirement in the selected case before counting keywords. Informational searches may have business value, but give them a separate job and window when this round tests purchases. Competitor and comparison terms also need relevant comparison content and applicable policy checks.

Check four entry questions

Before a cluster enters round one

  • Purchase, comparison, or solution intent?

    Keep when
    Enter first keyword candidates.
    Hold when
    Hold or negate informational, free, tutorial, job, and support terms.
  • Is there a page that directly answers it?

    Keep when
    Map it to an ad group and landing page.
    Hold when
    If no page exists, do not buy clicks yet.
  • Is match scope explainable?

    Keep when
    Start with Phrase / Exact and watch search terms.
    Hold when
    Hold broad until conversion, negatives, and page evidence are stronger.
  • Does the term have enough business value?

    Keep when
    Connects to product, margin, inventory, and page evidence.
    Hold when
    If it only brings cheap visits without revenue logic, hold it.

Broad is the default and widest keyword match type. It can match related searches using signals that include user activity, pages, assets, and other keywords in the group. This lesson starts with Phrase / Exact to make the first query sample easier to interpret. That is a bounded working choice. Neither quotes nor brackets lock queries to exact characters. Read actual search terms before expanding.

Choose an action in four situations

Here is a worked answer: the 20oz product-buyer terms fit “Split into one-intent ad groups,” with the RSA and page answering leak resistance, capacity, commuting, and shipping together. In the exercise, choose your own action. Changing the situation does not automatically answer it for you.

Exercise · pick oneCheckpoint

Which action keeps the sample interpretable?

① Choose the situation
② What would you do
When terms still show product-buying intent but use different angles, split by intent so one RSA and one page can answer each group.
If the page does not answer the problem behind the query, add clear above-the-fold information, use-case copy, and reviews before buying clicks.
Brand terms often convert better, so they must be separated from non-brand terms or they hide the real Search test quality.
Obvious low-intent terms should enter the starter negative list to protect week-one spend and search-term samples.
This shortcut looks convenient, but it prevents the ads, pages, and later review from producing a clear diagnosis.
Queries
"leak proof travel mug"; "20oz insulated tumbler"; [commuter coffee tumbler]
Problem
These terms all show buyer intent for the same product type, but they should not be mixed with gift, wholesale, or broad content terms.
Suggested structure line
Ad group: 20oz leak-proof commuter tumbler; keywords: Phrase / Exact; page: 20oz product page; negative candidates: free, DIY, repair, jobs, used.
Hold action
Do not casually add terms like "coffee accessories," "gift cup," or "bulk tumbler" into this group.

Selected exercise action

HoldUnanswered or action needs review
Choose an action, then read the case feedback. Low CPC, brand ROAS, or a longer list cannot replace page fit.

Work through the tumbler from query to page

Treat one RSA here as a set of promises to review together. Real accounts can contain multiple ads. State what the product is, which problem it solves, why the claim is credible, and why someone should click. Materials, genuine reviews, and verified shipping or return information can support the claim. Use stock, offers, and timing only when they are accurate.

Exercise · pick one

Read each writing step and its reason

Purpose
A keyword cluster is not every related term you can find. It is a small group of searches that one page can answer. Match type controls how far Google can expand from those terms. Phrase / Exact keeps round-one data easier to read. Broad is not forbidden; it is just harder for a beginner to diagnose in the first pass.
What to write
Keyword cluster: "leak proof travel mug", [20oz insulated tumbler], "commuter coffee tumbler". Match type: Phrase / Exact only in round one. Hold: best coffee accessories, gift mug, coffee blog ideas.
Why it fits
These terms ask for the same thing: a commute-friendly, leak-proof tumbler with clear capacity. One 20oz leak-proof product page can answer them. Generic and gift terms carry different intent, so adding them early makes the first search-term report hard to read.
Avoid
Do not add every Keyword Planner idea. The first Search campaign is not trying to cover the whole market; it is trying to produce one readable search sample.
Portable line
Write back to the sheet: focus on 20oz leak-proof commuter-tumbler queries, start with Phrase / Exact, and hold generic, gift, and content searches.

Give the ad promise visible page support

  • Waterproof dog car cover

    Ad wording
    Waterproof, dirt-resistant, easy to clean
    Supporting evidence
    Material, size, reviews, installation image
    Page content
    The product page shows waterproof performance and vehicle fit above the fold
  • Commute travel mug

    Ad wording
    Leak-proof, car-cup-holder fit, fast shipping
    Supporting evidence
    Size, seal design, shipping speed, returns policy
    Page content
    The product page shows capacity, seal details, and the shipping promise above the fold
  • Problem-solving query

    Ad wording
    Answer the problem before naming the product
    Supporting evidence
    Comparison image, use case, real reviews
    Page content
    Use-case section or collection page should not make users hunt for the answer

Single-product cold start

Ad group: Leak-proof commuter mug

Keywords: "leak proof travel mug"; "commuter coffee tumbler"; [leak proof coffee mug]

Promise: Ads promise leak prevention, commuter use, and heat retention, not gifts, wholesale, or coffee education.

Page evidence: Above the fold, show the lid and seal design, capacity, heat-retention claim, return policy, and real product images.

Held this round: Hold generic terms like "best coffee accessories" because the page cannot answer all intent behind them.

Multi-variant product

Ad group: Dog car seat cover

Keywords: "waterproof dog seat cover"; "dog hammock for car"; [dog car seat cover waterproof]

Promise: Ads focus on waterproof back-seat pet travel, not pet beds or cat carriers in the same group.

Page evidence: The page must support car fit, size, installation, waterproof material, and cleaning.

Held this round: Use "free dog seat cover" and "DIY dog hammock" as negative candidates so week-one sample stays clean.

For the dog seat cover, also separate dog seat cover wholesale and dog car sickness: one may need wholesale terms, while the other asks a different question. Sharing the word “dog” is not enough to share a landing page. Keep pet beds, cat carriers, gifts, and generic content out of this first product-buyer group.

Landing-page experience also covers relevance to the ad and keyword, usefulness, easy navigation, and whether the visitor finds what the click led them to expect. Make specifications, promises, shipping, returns, and the purchase route easy to find. A complete-looking page still needs testing against real orders and conversion records; it does not establish that visitors will buy.

Draft the scenario, matching, network, and pinning choices

The following controls save practice inputs; they do not create or edit ads. Scenario and match action are independent: after changing the scenario, check whether the current match action still fits. Core choices, RSA, evidence, page selections, and working notes each have their own save, restore, clear, and JSON export. A save button handles only its group. Copying or downloading text reads the current work and does not save those groups.

Exercise · pick one

Build scenario

Example queries
"leak proof travel mug"; [20oz insulated tumbler]; "commuter coffee tumbler"
Planner use
Use product and use-case seed terms in Keyword Planner, then group candidates by whether one page can answer them. Planner output is a candidate and forecast, not an automatic add list.
Ad group
Ad group: 20oz leak-proof commuter tumbler. The ad and page answer only capacity, leak proofing, cup-holder fit, and returns.
Final URL
The matching product page needs capacity, seal details, price, shipping, returns, and a CTA above the fold. Do not send this high-intent click to a homepage.
Brand boundary
Read brand terms in a separate campaign or ad group, not in this non-brand buyer group.
Negative candidates
Negative candidates: free, DIY, repair, jobs, used. Write the scope before deciding whether to exclude.
Why the recommended action fits
Phrase or Exact keeps this round focused on one buying intent. They still match by meaning, not by a literal character lock.
Why other actions stay held
Broad is a valid feature. This cold-start record still needs clear conversion, query, negative, and page boundaries. Do not treat expansion as the default action yet.
Deliverable · recordNot saved

Search core choices and six checks

Start with item 1; your choices in the exercises above are carried over.

Match or hold action
RSA pinning
Networks decision

Not saved yet.

Kept in this browser only. Switching device or clearing browser data removes it; download a copy.

Current scenario and match action

NoteAction fits the selected scenario
Phrase / Exact fits the product-buyer and brand-defense cases; the page-gap case calls for Hold. This check reads your actual scenario and action selections.

Write and check the RSA assets

A real RSA needs at least 3 headlines and 2 descriptions, with up to 15 headlines and 4 descriptions. Headline and description limits are 30 and 90. This exercise retains the original weighted approximation: characters in its specified CJK/fullwidth ranges count as 2, with spaces and line breaks included. Over-limit drafts remain editable and saveable so you can correct them. A valid count does not establish policy approval, guaranteed pinned text, or final rendering.

The default copy is a worked example. 30-Day Returns, leak resistance, and cup-holder fit all need real support. Record the position and reason when pinning necessary legal or brand wording; pinning every position restricts combinations. If the account has multiple RSAs, review each one’s Final URL, policy state, and asset report.

Deliverable · recordNot saved

RSA headline and description draft

All 5 items are required. Start with item 1.

26 / 30

Use a supportable promise. Example: 20oz Leak Proof Travel Mug

20 / 30

Use a supportable promise. Example: Fits Car Cup Holders

14 / 30

Use a supportable promise. Example: 30-Day Returns

85 / 90

Use a supportable promise. Example: A sealed lid for commuting, with size, price, and shipping on the product page.

57 / 90

Use a supportable promise. Example: See leak-proof details, commuter use, and the return policy.

Not saved yet.

Kept in this browser only. Switching device or clearing browser data removes it; download a copy.

RSA length and required fields

NoteFive assets filled within exercise limits
This checks the text and its count only. Review Google Ads asset status, policy, and preview separately.

Read the five current inputs together below to find repetition, contradictions, or claims the page cannot support. The display address is a placeholder. This does not simulate every combination, truncation, device layout, or actual serving. Empty lines after clearing help expose missing assets.

Note · copyable

Current RSA combination check

Read current controls rather than guessing from old labels

Location has more than one layer: the target area and an advanced option such as Presence or Presence or Interest. The latter can include people outside the area who show interest in it. Targeting is signal-based, with no guarantee of perfect accuracy. Canada is the case plan, not proof of delivery coverage or a customer’s physical location. Check exclusions and the rationale against where you can actually serve.

Google’s language guidance describes a change starting in 2026-09: Search removes campaign-level language targeting and matches automatically using ad language. For PMax, that change applies to its Search Network portion; other channels have different rules. Check ad, landing-page, and service language, and record whether the account still shows a language control or uses automatic matching. See Google’s language change guidance.

New Search campaigns include Search Partners by default; you can adjust the choice under the campaign’s Networks settings. Record why you keep or exclude partners and how you will review that choice. Do not assume standard Search gives site-by-site partner impression reports, or infer networks from the campaign name.

The campaign-level broad match setting described in the transition applies broad matching across the Search campaign and requires conversion-based Smart Bidding. This condition belongs to that campaign-level setting; it is not a requirement for every individual Broad keyword.

Broad keywords, the campaign-level broad match setting, and AI Max are different layers. Google describes an upgrade to AI Max starting in 2026-09 for campaigns using the campaign-level broad setting; this does not mean every Broad keyword already enables AI Max. Read its matching, text customization, Final URL expansion, brand, and URL controls. Expansion requires text customization. If a more relevant URL is selected, pinned assets are not used, so required wording needs a URL and pinning check. See AI Max guidance.

The RSA asset report has deprecated its Performance label. Read the available statistics, with full statistics dated from 2025-06-05. One impression can contain several assets, so their metrics cannot be added as independent contributions. CTR, CPC, CPA, and ROAS are directional rather than proof that one line caused a change. See asset reporting guidance.

Join the six checks to four evidence notes

Read the current admin field, then record its actual state and your reason. A tick is your record of that review, not an automated account verification; all six start empty. Record settings, reasons, dates, and evidence locations without account IDs, customer or visitor data, or credentials. The six ticks belong to the core choices group above; return there to save them. The four written notes below have their own Save button.

Checklist0 / 6 required
Six admin readback checks
  • Record seed terms, retained terms, held terms, and why one page can answer the group.
  • Record the region, Presence / Interest choice, ad and service languages, and the available language controls or automatic matching state. Do not infer them from a name.
  • In Campaign settings under Networks, record whether Search Partners are included and how that choice enters review.
  • Record whether brand and non-brand are separate and whether negatives sit at ad group, campaign, list, or account level.
  • The practice preview cannot replace Google Ads asset, policy, or Ad Preview status. Record those fields in the admin surface.
  • If AI Max is enabled, record search term matching, text customization, Final URL expansion, brand or URL controls, and whether pinning can still be respected.

Deliverable · recordNot saved

Four Search field readbacks

All 4 items are required. Start with item 1.

Example: Canada; Presence; English ad/page/service; current language matching state; Networks; non-brand purchase goal. Record fields and decisions, not account IDs, customer data, or visitor data.

Practice example: 20oz leak-proof commuter tumbler; Phrase / Exact; brand separate; free and DIY are ad-group negative candidates. Write the scope and held items. Do not treat a candidate as an approved negative.

Practice example: 3 headlines and 2 descriptions are within limits; Final URL matches the product page; policy and Ad Preview await admin confirmation. Do not treat this practice preview as evidence that an ad is approved or has served.

Example: AI Max off this round. Record matching, URL expansion, brand controls and pinning before later testing. AI Max is a Search optimization layer; verify the current account controls.

Not saved yet.

Kept in this browser only. Switching device or clearing browser data removes it; download a copy.

Current practice-record state

Hold

Practice readback or evidence is missing; keep the hold

Review scenario/action fit, pin-all or unchecked network assumptions, blank or over-limit assets, six ticks, and four written readbacks. Saving a record does not establish completeness or account acceptance.

After saving, use that group’s Restore to check its contents. Clear returns the group to its initial choices: RSA returns to the example copy, while personal ticks and notes become empty. The import covers the build choices, six checks, four readbacks, and five RSA assets. Other page choices and working notes use newly added grouped saving. The old timestamp stays in the old record; a native save time describes this save. Old text inputs were unbounded, while each native text field stores up to 2,000 characters. Keep a full copy of longer old work; a short import sample does not prove arbitrary-length preservation.

Know which admin fields to read

The English paths below identify the objects to find; they are not evidence of an account inspected in this lesson. Interfaces can move, and language controls follow the change described above. Find the equivalent object and record its actual fields and check date. Images can supplement the record, while fields, rationale, and business records carry the review.

Campaign settings / goal, bidding, and budget record

Google Ads > Campaigns > Campaign settings > goal, bidding, budget, networks, locations, languages, and start date.

Fields to record: campaign name; objective; conversion goal; bidding strategy; daily budget; network setting; location; language; start date

What to check with this record: Use these fields to confirm one campaign job and consistent budget, market, and communication-language assumptions.

Avoid: Do not keep changing goal, bidding, or location after launch; if you must, record it for change-history review.

Next step: If budget or bidding becomes unreadable, return to Budget and bidding basics.

Ad group / keyword intent record

Google Ads > Ad groups > Keywords; keep keyword source, match type, final URL, and held terms together.

Fields to record: ad group name; keyword; match type; intent label; landing page URL; held terms; starter negatives

What to check with this record: Shows that the ad group is not being used as a keyword warehouse and that one page can answer the shared intent.

Avoid: Do not add every Keyword Planner idea; terms the page cannot answer should be held or listed as negative candidates.

Next step: If queries drift, move next to Search terms, negatives, and match types.

RSA / assets / final URL promise record

Google Ads > Ads > Responsive search ad; record headline, description, path, asset status, policy status, and final URL.

Fields to record: headline set; description set; path fields; final URL; asset status; policy status; above-the-fold content

What to check with this record: Confirms that the query, ad promise, and above-the-fold page content tell the same story.

Avoid: Do not claim offers, comparisons, inventory, or returns the page cannot prove; do not fill generic headlines just for Ad strength.

Next step: If clicks are good but post-click behavior is weak, return to CRO page role or landing-page fit.

Ad Preview / change history check

Tools > Ad Preview and Diagnosis; Campaigns > Change history; record preview query, location, device, eligible / disapproved / limited status, and launch changes.

Fields to record: preview query; location; device; ad status; policy status; change date; changed by; launch note

What to check with this record: Use this check to confirm that the ad has diagnosable serving conditions and a launch change record, rather than existing only as a row in a sheet.

Avoid: Do not validate ads by Googling them yourself; personalization and auctions can mislead, and it can create unnecessary impressions.

Next step: If status is disapproved, limited, or not showing, fix serving status before optimization.

Week-one readout / keep, pause, expand record

Google Ads > Search terms, Keywords, Ads, and Landing pages; reconcile with GA4 events, Shopify Orders, and change history.

Fields to record: search term sample; CTR; CPC; add_to_cart; purchase; order id sample; negative action; keep / pause / expand decision

What to check with this record: Confirms that week-one decisions do not come from one ROAS number; decisions come from query sample, post-click behavior, and order evidence.

Avoid: Do not rebuild the structure on day two; wait for the minimum sample and record why to keep, pause, expand, or fix the page.

Next step: Once actual query samples exist, review search terms, negatives, and match types; the next series lesson first covers budget and bidding.

UTM Link Builder

Use the intended landing-page URL and consistent source, medium, and campaign names, then keep the generated link with the handoff. This distinguishes traffic sources; it does not verify Purchase deduplication or reconcile conversion attribution.

Name this campaign’s links consistently

Set the negative-keyword scope before excluding

Free, DIY, repair, jobs, used, support, login, and tutorial are candidates for this purchase test. Check their relationship to the business, page, and current objective before excluding them. Search negatives have broad, phrase, and exact types, but their rules differ from positive matching and do not automatically cover every close variant. More negatives are not automatically safer: overblocking can remove useful demand.

Four negative scopes

  • Ad group

    Use for
    Terms that only pollute one intent group, such as repair, DIY, or support inside one group.
    Protect against
    Do not block terms that may still be valuable in another ad group.
  • Campaign

    Use for
    Brand separation, market or business boundary, and low-intent searches clearly excluded from this first campaign.
    Protect against
    Do not block broad roots before Search terms data shows the problem.
  • Negative keyword list

    Use for
    Reusable competitor, support, used, jobs, wrong-product, or similar stable exclusions.
    Protect against
    Do not put temporary edge-case tests into a long-lived shared list.
  • Account-level

    Use for
    Risk terms the whole account should not buy, such as illegal, free, jobs, repair, or clearly unsupported categories.
    Protect against
    Do not harm Shopping, PMax, brand, or future high-value demand by mistake.

Account-level negatives affect relevant Search and Shopping ad inventory, including those portions of PMax. Do not treat that as one exclusion rule for every PMax channel. Check the scope carefully for brand terms and valuable terms a future test may need.

Find the break before expanding automation

For the week-one readout, CTR is clicks divided by impressions; CPC is spend divided by clicks. If Purchase is the only sale outcome counted here, CVR compares those purchases with the corresponding clicks, while CPA allocates ad cost to each counted purchase. State the conversion action and reporting scope first. A cart addition is an earlier action and must not be blended with purchases as the same sale result.

The 50–100-click window is a classroom prompt to review, not a Google threshold, a significance test, or a guarantee of sufficient data. Order value, conversion lag, query mix, and budget still matter. If intent remains relevant after 50–100 clicks but conversion is weak, inspect price, shipping, trust, page, and checkout before widening matching.

Week-one observations and evidence

  • Day 0

    Read first
    Confirm conversion tracking is accepted, the campaign is eligible, Campaign settings are clean, and Ad Preview plus Change history preserve the launch record.
    Hold this action
    Do not change the core structure because the interface prompts you; record the launch state first.
    Keep
    Conversion status, Ad Preview query, change timestamp, and launch judgment.
  • Day 1-2

    Read first
    Check whether impressions and clicks start. If there is no spend, review bid strategy, budget, location, keyword coverage, and ad status first.
    Hold this action
    Do not rewrite every RSA or break apart all ad groups because day one or two has no conversions.
    Keep
    Whether spend started, whether settings block delivery, and whether the query sample is still too small.
  • Day 3-7

    Read first
    Read Search terms daily, add obvious low-intent negatives, and check post-click behavior plus order samples.
    Hold this action
    Do not change match type, budget, RSA, and landing page together; change one explainable variable at a time.
    Keep
    Search terms, negative action, CTR, add_to_cart, purchase, and Shopify order id sample.
  • 50-100 clicks

    Read first
    Use CTR, search terms, CVR, post-click behavior, and order samples to diagnose structure, page fit, or conversion trust.
    Hold this action
    Do not pause the campaign based only on one or two days of ROAS; first confirm the sample size and locate the break.
    Keep
    The keep, pause, expand, or page-repair decision sentence.

Next action for Broad / AI Max

  • Hold

    Current condition
    Conversion tracking is not accepted, Search terms sample is thin, page fit is unclear, or negative scope is not documented.
    Next action
    Keep a Phrase / Exact draft while completing the missing checks; do not treat broad or AI Max as a cold-start repair button.
  • Check later

    Current condition
    You have 50-100 clicks, high-intent Search terms look stable, but conversions are still low or post-click breakpoints are visible.
    Next action
    Diagnose page, price, shipping, trust, and checkout first; if needed, test one controlled expansion variable.
  • Route to advanced

    Current condition
    Conversion signal is stable, Smart Bidding is readable, and Search terms plus negative lists are auditable.
    Next action
    Route broad / AI Max to budget, bidding, or advanced optimization lessons instead of scaling it during the initial setup.

Assumptions matter in automated bidding too. Xing and colleagues’ auction study models public impression values and private advertiser budget and ROI constraints, asking how truthful reporting can serve participants’ interests and how participation can leave them no worse off. These are the incentive-compatibility and individual-rationality questions. Its experiments use synthetic data. It supports making goals and constraints explicit, not a claim about Google’s current matching rules, a store’s profit, or an optimal click window.

Run a half-hour pre-launch review

Review the objective, budget, market, communication language, and landing page for the same campaign together. This 30-minute agenda allocates review time; it does not promise every gap can be resolved in half an hour.

Launch review meeting

  1. Choose one job

    Choose non-brand purchase, problem solving, brand defense, or a comparison job. Confirm its objective and primary conversion; state the current pressure, required signal, and held work.

    How to confirm

    Allocate 5 minutes; give unresolved work an owner and a review condition.
  2. Connect cluster, RSA, and page

    Read the cluster and Phrase / Exact start, then run the tumbler example. Check the ad group, promise, specifications, page evidence, and Final URL together.

    How to confirm

    Allocate 10 minutes; give unresolved work an owner and a review condition.
  3. Check settings, negatives, and preview

    Review market and current language behavior, Networks, negative scope, policy, Ad Preview, and Change history. Hold Broad / AI Max when measurement or page evidence is unfinished.

    How to confirm

    Allocate 10 minutes; give unresolved work an owner and a review condition.
  4. Write the review line and owner

    Write the week-one window, reviewer, and keep, pause, expand, or page-repair conditions. Record the first evidence, this week’s action, held changes, and the next route.

    How to confirm

    Allocate 5 minutes; give unresolved work an owner and a review condition.

When to continue or hold

Stop hereGo
All keywords are stuffed into one ad group
Each ad group serves one search intent
Evidence
Intent grouping and landing page are documented
Brand, non-brand, competitor, and generic terms are blended
Brand terms are read separately; competitor/generic terms have boundaries
Evidence
Campaign goal and brand boundary
Specific ad, generic page
RSA promise matches the landing page above the fold
Evidence
Headline, supporting information, CTA, and landing-page URL
No starter negatives or pause rules
Low-intent terms are blocked and keep/pause/expand rules are clear
Evidence
Starter negatives and first-review rules
What you finished

No checkpoint finished in this lesson yet. Start with “Go to the exercise”.

Your work

  • Record · Draft the scenario, matching, network, and pinning choicesNot saved
  • Record · Write and check the RSA assetsNot saved
  • Note · Write and check the RSA assetsNot copied
  • Record · Join the six checks to four evidence notesNot saved
  • Record · Run a half-hour pre-launch reviewNot saved
  • Record · Run a half-hour pre-launch reviewNot saved
  • Note · Run a half-hour pre-launch reviewNot copied

Kept in this browser only. Switching device or clearing browser data removes it; download a copy.

Choose a brief and write four working notes

A brief keeps the cluster, ad group, promise, page, negative scope, and review window in one explanation. It is a writing reference; selecting one does not fill your personal evidence. Write the four notes yourself, including an owner and review condition where useful.

Exercise · pick one

Search campaign brief reference

Situation
The shopper already searches for the product type, such as leak proof travel mug or 20oz insulated tumbler. The first round should capture clear product demand without mixing in branded or problem-oriented queries.
Cluster
"leak proof travel mug", [20oz insulated tumbler], "commuter coffee tumbler"
Ad group
Ad group: 20oz leak-proof commuter tumbler; focus only on product-purchase intent.
RSA promise
RSA promise: 20oz, leak-proof, car-cup-holder fit, 30-day returns; avoid generic premium lifestyle copy.
Final URL
Final URL: 20oz leak-proof commuter tumbler product page with specs, leak-resistance details, price, shipping, and CTA above the fold.
Negative seed
Negative seed: free, DIY, repair, jobs, used; keep brand terms in a separate record, not in the non-brand group.
Week-one window
Week-one window: wait for a readable search-term sample, CTR, CPC, add_to_cart, purchase, or order evidence before keep, pause, or expand decisions.
Write-back
Write back: launch only 20oz product-buyer terms; page and RSA answer the same demand; week-one edits come from queries, post-click behavior, and order evidence.
Deliverable · recordNot saved

Page selections and personal structure checks

Start with item 1; your choices in the exercises above are carried over.

Not saved yet.

Kept in this browser only. Switching device or clearing browser data removes it; download a copy.

Deliverable · recordNot saved

My Search structure working notes

Start with item 1.

Example: Canada English market, non-brand purchase validation, small daily budget first.

Example: leak-proof commuter mug; "leak proof travel mug", [leak proof coffee mug]; Phrase/Exact.

Example: headlines promise leak proofing and cup-holder fit; page shows seal, size, and shipping.

Example: negate free, DIY, repair, jobs; pause query drift; expand close terms after conversion evidence.

Not saved yet.

Kept in this browser only. Switching device or clearing browser data removes it; download a copy.

The note below carries current choices with their context. Its final feedback index contains the full explanations for all four cases, unfiltered. Use the entry for your actual selected case and action. Grouped JSON backs up input values; this text adds readable guidance and handoff context.

Note · copyable

Search structure handoff

Common questions

What does the 20oz keyword-to-ad-group worked example help me decide?
It helps you decide whether one keyword cluster belongs in one ad group, whether one RSA and one Final URL can answer it, and which free, DIY, repair, jobs, used, or similar terms should start as negative candidates.
Which keywords should the first Google Search campaign buy?
Start with high-intent non-brand product terms or problem-solving terms that the page can answer and the business can value. Do not blend brand, competitor, generic, and informational terms into the first readout.
How should Google Ads ad groups be separated?
Separate them by search intent and page fit. Terms that one RSA and one landing page can answer may start together. Terms that need a different promise or page should be split, held, or supported with page work first.
Why do beginners often start with Phrase / Exact?
Phrase / Exact protect first-week readability so you can judge query, page, and conversion evidence. They are not always safe, and broad is not always wrong; they simply keep the first sample easier to explain.
How should I write RSA copy?
Write RSA copy around product, problem, proof, and click reason rather than filling headline slots. If the copy promises 20oz, leak-proof, cup-holder fit, or returns, the Final URL first screen must show matching proof.
How do I choose Final URL and starter negative keywords?
Choose the product or collection page that can keep the promise for that intent, not the homepage by default. Starter negatives usually come from low-intent or wrong-intent terms such as free, DIY, repair, jobs, used, support, login, and tutorial, then you choose ad group, campaign, list, or account-level scope.
When can I consider broad or AI Max?
Consider broad or AI Max only after conversion tracking is accepted, Search terms have a readable sample, page fit is clear, negative scope is auditable, and roughly 50-100 clicks still leave an explainable expansion case. Route that decision to budget, bidding, or advanced optimization work. The 50–100-click window is an example review prompt; it does not by itself establish sufficient evidence or readiness to expand.
What should I watch in week one, and what should wait?
On Day 0, check tracking, eligible state, Ad Preview, and Change history. On Day 1-2, check whether impressions and clicks begin. On Day 3-7, read Search terms and post-click behavior daily. Do not change match type, budget, RSA, and landing page together, and do not kill the campaign from one or two days of ROAS.

What to learn next

Take the same cluster, page, primary conversion, and hold conditions into budget and bidding. If queries drift from purchase intent, carry samples into search-term governance. If queries fit but post-click behavior is weak, repair the page or check tracking. Change one explainable variable at a time.

Use the search-term review lesson when actual queries are available; the next lesson covers the budget and bidding boundary for this same draft.

Edition notes

  • Author Ranfeng Wei
  • Published
  • Updated
  • Reviewed
  • Next review

This update: Reorganized the keyword, ad-group, RSA, page, and week-one workflow while retaining drafts, feedback, and exports; clarified current Search language behavior, asset reporting, and AI Max boundaries. Distinguish five review questions from account levels and explain CTA and week-one metric definitions.