2 · Top-line vs. target & impression share
All six use Maximize Conversion Value with no target ROAS, bidding on Won/consult value. ROAS is healthy (2.3–5.8×). But the defining number is impression share: every campaign sits near 10% IS, with the loss split roughly 40% to budget and 50% to rank. The ceiling is both money and quality/bids — there is 5–10× headroom before structure is even the constraint.
| Campaign | Spend | Conv value | ROAS | Value / conv | Search IS | IS lost — budget | IS lost — rank |
|---|
| New York (NYC metro) | $23,056 | $91,966 | 3.99× | $2,427 | 10.0% | 45% | 49% |
| Virginia | $20,160 | $97,725 | 4.85× | $2,327 | 10.0% | 36% | 56% |
| Maryland | $16,056 | $93,322 | 5.81× | $2,437 | 10.0% | 37% | 54% |
| New Jersey | $14,151 | $32,190 | 2.27× | $1,894 | 10.0% | 43% | 51% |
| Washington D.C. | $9,994 | $30,410 | 3.04× | $1,962 | 11.7% | 36% | 52% |
| Boca Raton, FL — EMDR | $7,742 | $28,800 | 3.72× | $3,600 | 10.0% | 33% | 59% |
Conversion volume (≈8–42 Won/consults per campaign over ~6 months) is thin for value bidding — especially FL (8) and DC (15.5). Pooling signal (portfolio bidding) or consolidating would help the algorithm learn.
3 · Search-terms report
Coverage & brand
- The search-terms report accounts for $46,970 of ~$91,364 spend (~51%) — ~49% hides under “other search terms”. Read it alongside that blind spot.
- Brand is just 2% of spend ($729) — this is almost pure non-brand acquisition, so blended numbers are not inflated by brand. No brand/non-brand split needed.
Match-type behavior (how queries matched)
| Match | Spend | CPA |
|---|
| Near-exact | $12,614 | $341 |
| AI / broad expansion | $10,533 | $975 |
| Broad | $8,966 | $664 |
| Near-phrase | $8,091 | $522 |
| Exact | $6,381 | $580 |
| Phrase | $386 | $193 |
Automated broad / AI-Max matching is the costliest per Won ($975) — it is drifting to lower-intent queries and needs tighter match types + negatives.
Telling pattern: the top zero-conversion search terms are
on-target high-intent queries — “couples therapy nyc” ($811), “emdr therapy nyc” ($553), “emdr therapy near me”. The queries are right; conversion fails
after the click. That points back to the
landing-page experience, not the keywords (some of this is also conversion lag + the 49% coverage gap). Genuine negatives to add are narrower: “online therapist”, bare “therapist near me” in FL.
4 · Keywords, match types & Quality Score
Average Quality Score 6.2 (median 7). The three sub-scores localize the problem precisely:
| QS component | Above avg | Average | Below avg | Read |
|---|
| Ad relevance | 91 | 35 | 7 | Strong — not the issue. |
| Landing-page experience | 36 | 76 | 21 | Weak — 21 keywords flagged. Matches the funnel finding. |
| Expected CTR | 20 | 62 | 51 | Weakest — drags ad rank, drives the ~50% IS lost to rank. |
Diagnosis: Expected CTR (too-broad terms, generic ads) and landing-page experience are the QS drags — ad copy relevance is fine. Fixing match types + destinations lifts both rank-lost IS and QS. Also review cross-ad-group overlap (the same query is eligible in multiple modality ad groups) and whether 6–8 ad groups per campaign over-fragments thin conversion data.
5 · Ads (RSAs) & assets
Strong- Ad strength EXCELLENT on all 40 ad groups. 15 headlines each (600 total), dynamic {LOCATION(City)} insertion.
- Account-level assets present: 24 sitelinks, 31 callouts, 4 structured snippets, plus call, price & promotion.
Gaps- Only one RSA per ad group — no second creative to test.
- No image assets and no lead-form assets anywhere — missed CTR / SERP-real-estate levers.
- 21 pinned headlines across the set — pinning discipline is OK (strength still Excellent) but keep it minimal.
- CBT ads reuse EMDR/trauma headlines — fix copy message-match (see funnel page).
6 · Bidding strategy
Maximize Conversion Value, no target ROAS, on all six. The goal (maximize Won value) is right, and ROAS is healthy — but value-maximize without a target spends to budget and lets CPA float. With realistic Won values loaded, consider a target ROAS floor on the strong campaigns (MD 5.8×, VA 4.9×) and watch the learning window on any change. For thin-data campaigns (FL, DC) consider portfolio bidding to pool signal. Conversion lag on Won (CRM import) is long — don’t over-read short windows.
7 · Audiences, segments & exclusions
Observation audiences are layered (in-market / affinity USER_INTEREST, remarketing USER_LIST, and combined audiences) — useful signal for Smart Bidding and future bid-adjustment reads. Confirm converter / existing-client exclusions and that any Customer Match lists are uploaded and matching. No paid-search remarketing exclusion of recent Won was verified — worth checking to avoid paying to re-acquire booked clients.
8 · Geo, device & schedule
Device
| Device | Spend | CPA |
|---|
| Mobile | $63,274 | $505 |
| Desktop | $27,373 | $820 |
| Tablet | $511 | — |
Mobile is 70% of spend and the more efficient device — so mobile landing-page speed/UX matters most.
Silent leak- FL-Boca uses “Presence OR Interest” geo targeting — it serves to people merely interested in Boca, not physically there. The other five correctly use “Presence.” Fix FL to Presence.
- No ad scheduling — running 24/7. A dayparting review (intake-line hours) is worth a look.
10 · Settings hygiene
| Setting | State | Verdict |
|---|
| Search Partners | Off | Good |
| Display Expansion | Off | Good — no Display leak |
| Geo targeting | Presence (FL: Presence OR Interest) | Fix FL |
| Ad rotation | Optimize | Good |
| Budget | Capped (~40% IS lost to budget) | Raise where ROAS supports it |
11 · Prioritized actions
| # | Action | Detail | Impact | Notes / learning |
|---|
| 1 | Fix the post-click layer | Point every ad group to a localized page (modality × state) with an on-page booking form; fix the 4 geo-mismatches, the dead virtual pages and the Arlington 404. | High | Highest ROI; lifts CVR + landing-page QS. No learning reset. |
| 2 | Set FL geo to “Presence” | Stop serving Boca ads to out-of-area “interested” users. | High | Immediate waste cut. Minor. |
| 3 | Tighten match types + add negatives | Rein in AI/broad expansion ($975 CPA); add “online therapist”, generic “therapist near me”; standardize the shared negative lists across all six (FL has none). | High | Cuts low-intent spend; improves Expected CTR / QS. |
| 4 | Lift Expected CTR | Modality-specific ad copy (stop CBT reusing EMDR headlines); add image & lead-form assets; second RSA per ad group. | Med | Better rank → recovers IS lost to rank. |
| 5 | Raise budgets where ROAS supports | ~40% IS lost to budget on 2–5× ROAS campaigns (MD, VA) — uncap to capture demand. | Med | Triggers a short learning window — stage it. |
| 6 | Add a target-ROAS floor / pool thin data | tROAS floor on strong campaigns; portfolio bidding for FL & DC to pool signal. | Med | Learning reset — stage after structure fixes. |
| 7 | Confirm audience exclusions | Exclude recent Won / existing clients from paid search; verify Customer Match. | Low | Avoids re-acquiring booked clients. |
Sequencing note: do the non-learning-reset fixes first (landing pages, geo, negatives, copy/assets), let them settle, then make bid-target/budget changes so you don’t stack multiple learning windows.
