This is the deep-dive reference for Step 6 of optimization-weekly.md. The top-level routine has the 3-step harvest in checklist form. Open this doc when you need: the filter mechanics, the why, decision rules for borderline search terms, or guidance on which ad group a harvested term belongs in.
When to use this doc
- First time running the harvest on a new campaign or by a new operator
- When you forget the exact Search Terms filter combination
- When deciding whether a borderline search term is worth promoting
- Onboarding a new team member to the weekly rhythm
Why this procedure exists
The course’s teaches harvesting from Google’s keyword suggestions (the blue + button → “Keyword suggestions” pane). That works, but it has one structural weakness:
Google’s suggestions are predictions about what might be relevant, drawn from related searches across the platform. They’re guesses about future relevance.
The agency procedure pulls directly from search terms that have already converted in this account. Every term in the harvest list has demonstrated conversion intent in this client’s exact campaign. Zero false positives.
The harvest list size is also typically smaller than Google’s suggestion pool, which makes the weekly review faster and lower-risk.
If you only have time for one keyword-add procedure, do this one. The course’s blue-+-suggestions method is a useful supplement when you’ve exhausted the converted-search-terms pool, but it’s not the primary harvest.
When in the campaign lifecycle to start
Skip this step on weeks 1-3. The Search Terms report needs ≥2-4 weeks of data before the harvest is worth running. Below 2-4 weeks, the converted-terms pool is too thin: most converted terms are already covered by your initial keyword set, and you’re unlikely to find genuinely new ones.
After Week 4, run weekly. The pool refills as new converters surface.
Step 1: Filter the Search Terms report to “already converted, not yet added”
Procedure
- Open the campaign in Google Ads
- Navigate to Insights and Reports → Search Terms (or Keywords → Search Terms in the older interface)
- Set filters:
- Date range: All time
- Conversions > 0
- Added/Excluded: None
- Sort by Conversions (descending)
What each filter is doing
Conversions > 0, only show search terms that have actually produced leads. Drops 80-90% of the table immediately. The remaining list is “stuff that works.”
Added/Excluded: None, only show search terms that aren’t already in your keyword list and aren’t already negated. This is the key filter. Without it, you’d see every converting search term including the ones that triggered ads from existing keywords (e.g., a search term that matched your [plumber Sacramento] exact-match keyword would still show up as a “converted search term”, but it’s already a keyword, so promoting it would be redundant).
All time, don’t miss strong converters from 60+ days ago that haven’t been promoted yet. This is especially important for the first run of this routine on an established campaign: there’s often a backlog of converted terms that have never been promoted.
Sort by Conversions (descending), the strongest performers surface first, so even if the operator’s time-budget runs out mid-routine, the highest-leverage adds happen first.
What you’ll see
A short list (typically 5-30 rows) of search terms that have converted but aren’t yet captured as keywords or negatives. That’s your harvest queue.
A typical Week 5 campaign might show 8-15 candidates. By month 3-4, the pool shrinks to 0-2 per week as you’ve already promoted most strong converters. A shrinking pool is a sign the procedure is working, not a sign to stop running it.
Step 2: Review the list and save the good fits
For each search term in the harvest queue, ask three questions:
Question 1: Does this still match the client’s offering?
Sometimes a search term converted historically but the client has since narrowed scope. Examples:
| Historical converter | Current scope | Decision |
|---|---|---|
furnace install Folsom | Client now does HVAC repair only, no installs | Skip + add as negative |
commercial roofing Sacramento | Client moved to residential-only last month | Skip + add as negative |
emergency plumber Roseville | Still matches (residential plumbing in Roseville) | Save |
If the answer is “no, scope changed,” it’s not a harvest target, it’s a negate target. Add it to the next optimization-daily.md negation list instead.
Question 2: Is the term geographically still in scope?
Check service-area changes:
| Historical converter | Current scope | Decision |
|---|---|---|
plumber Lincoln | Lincoln was dropped from service area in March | Skip + add as Excluded Location and negative |
plumber Vallejo | Still serves Vallejo | Save |
Question 3: Is the search intent clean?
Some search terms convert by accident, high purchase intent but for the wrong service, or low intent but the user filled out a form anyway. Skim:
| Search term | Read | Decision |
|---|---|---|
emergency plumber 24 hours | High intent, clear commercial purpose | Save |
how to fix a leaky faucet | Educational; converted because user filled out the form by mistake or for a different reason | Skip (top-of-funnel) |
cheap plumber near me | High intent but low value (price-shopper) | Save with caveat (consider if the client’s offering is competitive on price) |
Roto Rooter Sacramento | Branded search for competitor; user converted but you don’t want to bid for competitor brand searches | Skip + add as competitor brand negative |
What “save” looks like
Copy the search terms you decided to save into a notepad or spreadsheet. Group them by which ad group they belong in (you’ll add them to the right ad group in Step 3).
A typical save list after a Week 5 review:
Ad group: HVAC Repair Sacramento
- emergency AC repair 24/7 Sacramento
- AC tune-up Roseville same day
Ad group: Furnace Service
- furnace pilot light not lighting
- furnace humming but not heating
Step 3: Add as exact match to the right ad group
Why exact match only for this harvest
The search term has already proven it converts in that exact form. Adding as exact match captures the proven traffic without inviting wider phrase-match expansion (which would broaden the keyword’s reach to terms that haven’t proven they convert).
This contrasts with how you build initial keyword lists at launch (where phrase + exact is standard), for a harvest, you’ve already filtered to proven converters, so exact match captures the win without dilution.
If you want the term to also run as phrase match later, add the phrase variant as a separate decision in a future weekly cycle, after the exact match has its own track record.
Procedure
For each search term in your save list:
- Identify the ad group whose theme matches the term’s intent. This is the most important judgment call in the procedure. The intent should drive placement, not word overlap. Examples:
emergency AC repair 24/7 Sacramento→ Ad group “HVAC Emergency Sacramento” or “Emergency AC Repair” (intent: emergency)AC tune-up Roseville same day→ Ad group “HVAC Maintenance Roseville” (intent: scheduled service, not emergency)furnace pilot light not lighting→ Ad group “Furnace Service” (intent: specific furnace issue)
-
Open that ad group → Keywords → + button
-
Add the term in exact-match notation (
[brackets]):
[emergency AC repair 24/7 Sacramento]
[AC tune-up Roseville same day]
- Save
Optional batch tool
If you’re adding 10+ keywords across multiple ad groups, paste them into keywordmatchtypetool.com to auto-format. For 1-3 keywords, just type them inline.
What if the term doesn’t fit any existing ad group?
Two scenarios:
Scenario A: the term is genuinely a new theme. Example: a residential-plumbing campaign harvests tankless water heater installation and you don’t have a tankless ad group yet. Don’t add it inline as a one-off. Make a note and create a proper “Tankless Water Heaters” ad group in the next monthly review (per optimization-monthly.md Variant A’s ad-copy section). One-off keywords in the wrong ad group hurt ad relevance.
Scenario B: the term is a sub-theme of an existing ad group. Example: the campaign has a “Furnace Service” ad group and you harvest furnace replacement quote Roseville. The term is more specific than the ad group’s general theme, but it still belongs there because the ad copy and landing page apply. Add it.
Use judgment. When in doubt, prefer adding to the closest existing ad group; if the term genuinely needs new copy, defer to the monthly review.
Frequency and pool dynamics
Run the harvest every week after Week 4. The pool size will follow this rough trajectory:
| Campaign age | Typical harvest size per week |
|---|---|
| Weeks 1-3 | (skip, pool too thin) |
| Weeks 4-8 | 5-15 candidates |
| Months 3-4 | 2-5 candidates |
| Months 5+ | 0-2 candidates |
A shrinking pool is good, it means you’ve captured most of the strong converters. Don’t stop running the procedure; keep checking weekly. Even at 0 candidates per week, the 5-minute filter check is worth it because:
- Markets shift. New search behaviors emerge. The pool can spike unexpectedly.
- Seasonal services see new converters every season change.
- If a competitor pulls back, traffic redistributes and new converters appear.
Logging
After the harvest, log changes in clients/<client>/optimization-log.md. Format:
2026-05-06 weekly | harvest pool: 6 candidates (filtered All time + Conv>0 + Added/Excluded:None)
2026-05-06 weekly | added [emergency AC repair 24/7 Sacramento] (exact) to ad group "HVAC Emergency Sacramento"
2026-05-06 weekly | added [AC tune-up Roseville same day] (exact) to ad group "HVAC Maintenance Roseville"
2026-05-06 weekly | skipped "Roto Rooter Sacramento" (competitor brand) | added as campaign-level negative phrase "Roto Rooter"
2026-05-06 weekly | deferred "tankless water heater installation" to monthly review (new ad group needed)
A complementary method: Google’s keyword suggestions
If you’ve finished this harvest and have time, Google’s keyword suggestions tool (the blue + button → “Keyword suggestions” pane on each ad group) is a useful supplement:
- The agency harvest captures proven converters (highest signal)
- Google’s suggestions capture predicted-relevant terms (broader, noisier)
Best run monthly, not weekly, alongside the monthly review. It exposes terms that haven’t converted yet in your account but have signal from Google’s broader data. Worth a 15-minute pass once a month after you’ve done the agency harvest.
Common questions
”What if a term has 1 conversion vs 10 conversions?”
Both qualify for harvest. The Search Terms filter Conversions > 0 is binary. A 1-conversion term might be thin signal, that’s fine; promoting it costs nothing if you set it as exact match (it’ll only trigger on that exact form, which has already converted). If it converts again post-promotion, you’ve captured a new keyword. If it doesn’t, it sits as a 1-conversion exact-match keyword that’s not hurting anything.
The more aggressive interpretation: a term with 1 conversion in 6 months might be a fluke. Use judgment, if the term is borderline on intent (Question 3 above) and only has 1 conversion, you can skip it.
”What about high-cost converters?”
A search term that converted once but cost $300 to do so is in your harvest queue (Conversions > 0 is true). Should you add it? Yes, but expect Bucket 2 of the pause cascade to revisit it if the cost/conv stays high after promotion. Promote it, let it run for 2-3 weeks, then the pause-keyword-cascade.md will catch it if it’s structurally over-budget.
”Can I add terms as phrase match instead of exact?”
For this harvest specifically: no, exact only. The harvest is built on the proven-form principle. If the exact match converts well over time, you can add a phrase-match variant in a separate weekly cycle as a deliberate expansion test.
”What if the search term contains a brand name (mine, not a competitor’s)?”
If the client searches their own brand and converts (e.g., Smith Plumbing Sacramento for Smith Plumbing), they’re a returning customer or someone Googled the brand after seeing an offline ad. Branded search is high-intent but free traffic territory, the client’s organic listing should rank for it. Adding it as a paid keyword can be wasteful unless you’re defending against competitors bidding on your brand. Default: skip own-brand terms.
Back to: optimization-weekly.md