Week 7 Tutorial: PPC Skeleton and Keyword Logic
Tutorial Build a PPC Skeleton
Acquisition
1 Orientation
⏱ 10 min
Brief; exemplar skeleton; filename rule.
Production
2 Structure & Themes
⏱ 35 min
Campaign to ad groups; themes by intent; initial negatives list.
Production
3 Keywords & RSA
⏱ 40 min
Build starter keyword sets with match types; write 6-8 headlines and 4 descriptions.
Investigation
4 Metrics & Safeguards
⏱ 25 min
Pick a conversion metric and micro-metrics; set a daily cap and two safeguards.
Assessment
5 Share & Upload
⏱ 10 min
Pair share; upload one PDF to Moodle.
Your tutorial session runs for two hours. Using the keyword research from your Week 6 SEO audit, you build a PPC skeleton: a campaign structure with keyword groups, match types, a negative keyword list, a starter RSA, and a set of efficiency safeguards. Your week has nine hours of structured learning time in total: one hour of lecture, two hours of tutorial shown above, and six hours of self-directed study covered in the independent study chapter.
AI-Assisted PPC Protocol
Activity 3 of today’s tutorial runs as a breakout room exercise where your group prompts an AI chat tool to draft your keyword expansion and RSA, then checks every line of what it produces before using any of it. Watch Google’s own responsive search ad tutorial before the session (linked in Activity 3) so you know what good ad copy looks like before judging the AI’s draft against it; any chat AI works identically for the prompting itself.
AI can support the Week 7 PPC task in two ways. First, it can generate headline and description variants for the RSA: give it the target keyword, the primary claim from your copybook, and the evidence element, and ask for eight headline options, then evaluate each for relevance, character count (30 characters maximum), and whether it stands alone as a claim. Second, it can suggest additional negative keywords: give it the campaign topic and your initial negative list, and ask it to identify fifteen further queries worth excluding.
Two restrictions apply. Keep bid recommendations, CPC estimates, and conversion rate predictions out of scope for AI, since these require live campaign data no AI model has access to. Base every efficiency metric estimate on publicly available benchmark data with a citation attached.
Tutorial Task
Work through these steps in sequence, across the activities below. Activities 2, 3, and 4 each have their own tables: fill them as you go, and by Activity 5 those three sets of tables are your finished skeleton, with no separate summary table to copy everything into again.
- Name your campaign and ad group, and import your Week 6 primary keyword plus two to three related long-tail terms into keyword themes grouped by intent.
- Assign match types: exact match for the primary keyword, phrase match for one to two related terms, broad match only where a negative list already protects it.
- Build your initial negative keyword list of at least ten terms using the anticipation method.
- Write a starter responsive search ad for your primary keyword: 6 to 8 headline variants (30 characters each) and 4 description variants (90 characters each), each one meaningful on its own.
- Pick your primary conversion metric and one micro-metric, then set a daily budget cap and two safeguards (additional negatives or a schedule restriction).
- State the CTR and conversion rate assumptions behind your CPA target, and label them [ASSUMPTION] with a testing method.
Every table below shows how Tidewood Guesthouse filled in the same row, directly alongside the blank cell where you fill in yours. Use Tidewood’s entry as your model, over the generic instruction “complete this cell”: read what Tidewood wrote, then write the equivalent for your own campaign in the same amount of detail.
In-Class Activities
Activity 1: Orientation (10 minutes)
Format: Whole class | Output: Three-question concept check result and skeleton scope note
Complete the three-question Moodle concept check individually and without notes.
Open the Week 7 Concept Check quiz in Moodle under the Week 7 activities panel.
Question 1. In the Google Ads auction, an advertiser with a maximum bid of MVR 6 and a Quality Score of 5 competes with an advertiser with a maximum bid of MVR 4 and a Quality Score of 9. Which advertiser achieves a higher Ad Rank?
- The first advertiser, because their bid is higher
- The second advertiser, because their Quality Score produces a higher Ad Rank of 36 versus 30
- They achieve the same Ad Rank because both inputs are competitive
- The platform always favours the higher bidder regardless of Quality Score
Question 2. An exact match keyword [guesthouse Maldives] will trigger for which of the following searches?
- “cheap guesthouse Maldives with free breakfast”
- “Maldives guesthouse deals”
- “guesthouses Maldives” (plural)
- “best guesthouse Maldives house reef access”
Question 3. A PPC campaign has a click-through rate of 8% and a conversion rate of 0.5%. The cost per click is MVR 3. What is the cost per acquisition?
- MVR 3
- MVR 24
- MVR 600
- MVR 1,500
After the concept check, your tutor walks through the completed Tidewood Guesthouse skeleton, posted on the Week 7 Moodle page, so you see a finished decision record before starting your own. Keep it open beside you for the rest of the session: every table you fill in today appears there, filled in end to end. Confirm the upload filename rule now, before you build anything: W07_[YourName]_PPC.pdf.
Activity 2: Structure & Themes (35 minutes)
Format: Individual | Output: Campaign and ad group names, keyword themes grouped by intent, initial negatives list
Name your campaign and ad group, then group your Week 6 keyword list into themes by intent.
| Slot | Tidewood’s example | Your entry |
|---|---|---|
| Campaign name | Tidewood Direct Book 2026 | |
| Ad group name | House Reef Snorkelling Malé |
Now assign a match type to your primary keyword and two to three related terms, and justify each choice by reference to relevance, reach, or budget protection. Tidewood’s four keywords model one of each match type, including a broad match that only works because a negative list backs it:
| Keyword slot | Tidewood’s example (keyword, match type, and why) | Your entry (keyword, match type, and why) |
|---|---|---|
| Primary keyword | [best guesthouse Maldives house reef access], Exact: highest commercial intent, so it gets the tightest protection | |
| Related term 1 | “affordable guesthouse Maldives snorkelling”, Phrase: keeps the price-plus-snorkelling intent while allowing natural variation | |
| Related term 2 | “guesthouse vs resort Maldives price”, Phrase: captures comparison-stage searchers by name | |
| Related term 3 | guesthouse Maldives, Broad, backed by the negative list below: worth the reach once excluded terms protect it |
Build your initial negative keyword list using the anticipation method: think about the queries that would trigger your ad under broad or phrase match, but that come from audiences who will never convert because they are looking for something fundamentally different.
| Category | Tidewood’s example | Your entry |
|---|---|---|
| Free or low-cost alternatives | “free stay”, “hostel”, “backpacker”, “dorm” | |
| Employment-related searches | “jobs”, “internship”, “salary” | |
| Information-only searches | “what is a guesthouse”, “Wikipedia” | |
| OTA and review sites you are routing bookings around | “Airbnb”, “Booking.com”, “TripAdvisor” | |
| Competitor brand names (only once confirmed as competitors) | left blank for now: Tidewood has yet to confirm a direct competitor by name | |
| Irrelevant geographies, if your campaign is local | [NOT APPLICABLE]: Tidewood serves a single location |
List at least ten negative keywords total across these categories, each with a one-sentence reason it would attract irrelevant traffic.
Activity 3: Keywords & RSA, AI-Assisted Draft (40 minutes)
Format: Breakout rooms (3-4 students) | Output: AI-drafted keyword expansion and RSA, checked line by line, plus a 100-word reflection
Watch Google’s own tutorial, “Google Ads Tutorials: Create effective responsive search ads” (Google Ads Help, YouTube, 2 minutes), before your breakout room opens. It covers what makes a headline and description actually work, straight from the platform you’re writing for, so you know what “good” looks like before you judge the AI’s draft against it. The prompts and checks below are already spelled out step by step: copy them in as written, so no separate demo of the prompting itself is needed. Any chat AI (Gemini, ChatGPT, Copilot, Claude) works identically for this activity.
In your breakout room, agree three roles before you start, then work through the four steps together. Rotate roles between Step 1 and Step 2 so nobody spends the whole activity watching.
- Driver: shares their screen, types the prompts, pastes the output into the table.
- Counter: counts the characters of every headline independently and calls out any over 30.
- Challenger: reads each headline aloud on its own and rules on the independent test.
Timings inside the 40 minutes: roughly 8 minutes on Step 1, 10 on Step 2, 15 on Step 3, and 7 on Step 4.
Paste a headline into any word processor and use its character count (Word and Google Docs both show characters with spaces under Word Count), or type it into a free browser character counter. Spaces and punctuation count toward the 30. Once your campaign is live, the Google Ads editor shows the count as you type.
Step 1: Expand your keyword set. Here is a complete, filled-in prompt for Tidewood Guesthouse, exactly as you would type it:
“I am building a PPC skeleton for Tidewood Guesthouse, a small guesthouse in the Maldives competing against nearby resorts on price and house reef access. My primary keyword is ‘best guesthouse Maldives house reef access’. Suggest five long-tail keyword variants, each with a recommended match type (broad, phrase, or exact) and a one-sentence reason for the match type choice.”
Now open your AI tool and write your own version of that same prompt, replacing every part with your own campaign’s details:
“I am building a PPC skeleton for [your campaign topic, plus a one-sentence description of what makes it distinct]. My primary keyword is ‘[your Week 6 primary keyword]’. Suggest five long-tail keyword variants, each with a recommended match type (broad, phrase, or exact) and a one-sentence reason for the match type choice.”
Read the five suggestions as a group. For each one: does the recommended match type make sense given reach versus relevance? Would a broad match here need extra negative keyword protection? Record all five below and rule on each one, since the ones you reject are as much a part of your decision record as the ones you keep.
| AI’s suggested keyword | Match type it recommended | Keep or reject, and why |
|---|---|---|
| “guesthouse Maldives with snorkelling from the beach” | Phrase | Keep: matches the house reef claim exactly and the phrasing is how guests actually describe it |
| 1 | ||
| 2 | ||
| 3 | ||
| 4 | ||
| 5 |
Carry every keyword you keep into your Activity 2 keyword table.
Step 2: Draft your RSA. Again, here is the complete Tidewood Guesthouse version first:
“My primary keyword is ‘best guesthouse Maldives house reef access’. My primary claim from my copybook is: ‘Tidewood Guesthouse gives you house reef snorkelling access, at a fraction of nearby resort prices.’ My evidence for this claim is: a private house reef accessible directly from the guesthouse, confirmed by the property’s own dive log and named in three guest reviews. Write 8 Google Ads responsive search ad headlines, each 30 characters or fewer, each meaningful on its own without needing another headline for context. Then write 4 descriptions, each 90 characters or fewer, expanding the claim with the evidence.”
Write your own version, filling in your own details:
“My primary keyword is ‘[your primary keyword]’. My primary claim from my copybook is: ‘[your Week 4 claim]’. My evidence for this claim is: [your evidence element]. Write 8 Google Ads responsive search ad headlines, each 30 characters or fewer, each meaningful on its own without needing another headline for context. Then write 4 descriptions, each 90 characters or fewer, expanding the claim with the evidence.”
Step 3: Check every headline against the independent test. Paste the AI’s output into the table below, then evaluate each line yourselves. Count each headline’s characters yourself; treat the AI’s own count as unverified.
Read each headline in isolation, without the others. Does it communicate something meaningful on its own? “Book Direct, Save More” passes. “With House Reef Access” fails because it depends on another headline for context. If the system pairs “With House Reef Access” with an unrelated headline, the ad reads as confused.
The first two rows show Tidewood’s own checking process: one headline that passed unchanged, one that failed and got rewritten. Your rows start below them.
| RSA element | AI’s draft | Character count | Passes independent test? | Your fix, if needed |
|---|---|---|---|---|
| Tidewood example | “Tidewood Guesthouse Maldives” | 28 | Yes | (none needed) |
| Tidewood example | “With House Reef Access” | 22 | No: fragment | “Snorkel Steps From Your Room” (28) |
| Headline 1 | ||||
| Headline 2 | ||||
| Headline 3 | ||||
| Headline 4 | ||||
| Headline 5 | ||||
| Headline 6 | ||||
| Headline 7, optional | ||||
| Headline 8, optional | ||||
| Description 1 | ||||
| Description 2 | ||||
| Description 3 | ||||
| Description 4 |
Step 4: Reflection (100 words), and log your prompts. As a room, agree on and write 100 words directly below your table answering: which of the AI’s headlines needed no change, which needed a rewrite and why, and what this tells you about verifying AI output before it goes anywhere near a live campaign.
Five of the eight headlines passed unchanged. The two strongest, “Tidewood Guesthouse Maldives” and “Book Direct, Save More”, each made a complete claim on their own. Three failed. “With House Reef Access” and “Steps From The Water” were both fragments: they read as the second half of a sentence and would confuse anyone who saw them paired with an unrelated headline. One more, at 34 characters, simply exceeded the limit, and the AI had reported it as 30. The pattern is that the model writes fluent copy and reports its own constraints unreliably. On a live campaign, the fragment would have run until somebody noticed.
Copy both prompts you used, exactly as typed, into the prompt log below. Your Week 7 submission requires at least two logged prompts, and these are them: capture them now while you still have the exact wording.
| Prompt | Prompt text as typed | What the AI returned | What you decided to keep, change, or reject |
|---|---|---|---|
| 1 (keyword expansion) | |||
| 2 (RSA draft) |
No campaign topic yet? Borrow Tidewood Guesthouse for the session and swap in your own campaign during independent study.
The AI returned eight near-identical headlines? Add one line to your prompt: “Each headline must make a different claim: one keyword match, one differentiator, one call to action, one trust signal.” Then run it again, and log both attempts.
Nothing failed the independent test? Say so in your reflection and explain how you tested it. Then check character counts again: an over-length headline is the more common failure, and it is easy to miss.
Activity 4: Metrics & Safeguards (25 minutes)
Format: Individual | Output: Conversion metric, micro-metric, daily budget cap, and two safeguards
Pick your primary conversion metric (one booking enquiry, one sale, one download) and one micro-metric that gives an early efficiency signal before conversions accumulate (click-through rate on your primary keyword, or landing page bounce rate).
For the CPC estimate, use Ubersuggest, the same free tool you used for Week 6 keyword research: it reports an estimated paid CPC alongside search volume for any keyword. Cite the tool and the date you checked. Keep this number out of scope for AI, since no model has access to live auction data.
| Calculation element | Tidewood’s example | Your entry |
|---|---|---|
| Primary conversion metric and its value | Booking enquiry, valued at roughly MVR 800 (typical nightly rate × minimum stay, adjusted for margin) | |
| Micro-metric to monitor daily | Click-through rate on the exact-match keyword: an early signal, since it moves before enough bookings accumulate to trust CPA | |
| Estimated CPC (from Week 6 keyword research or platform benchmark) | MVR 15-25, from keyword competition level [ASSUMPTION: test after two weeks of real spend] | |
| Maximum CPA you would accept for the campaign to break even | MVR 800, matched to the booking enquiry value above | |
| Daily budget cap | Enough for roughly 50 clicks a week at the estimated CPC | |
| Safeguard 1 (additional negative keywords or schedule restriction) | Add negatives from the search query report every week for the first month | |
| Safeguard 2 (additional negative keywords or schedule restriction) | Pause ads outside reservations team hours, so no enquiry goes unanswered |
A daily cap limits how much a Quality Score problem can cost you before you notice it. Pair the cap with a schedule restriction (running ads only during the hours your team can respond to enquiries) or a fresh negative keyword pass, so the safeguard catches waste even between weekly reviews.