19 months of account history, 37,442 search-term rows, every keyword, ad, audience, and conversion setting — pulled via the Google Ads API on Aug 15, 2026. Built to answer three questions: who the buyer is, why the account worked when it worked, and why it stopped.
$158k
total spend analyzed (2024→now)
$48
blended CPA on practitioner-intent terms
$639
July 2026 cost per conversion
710
people left in the YouTube remarketing pool
~$0
August spend — the account is stalled now
0.Three things that need attention this week
The account is effectively dark right now. August is pacing near zero ($161 in six days, junk clicks only). Cause, in evidence: the main campaign runs Maximize Conversions with a $200 target CPA still set, while true costs run $600+ — smart bidding throttles delivery it can't hit. Nothing serves, nothing books.
Smart bidding is optimizing toward garbage signals. The PRIMARY conversion actions include Google Business Profile local actions (website visits, "menu views", direction requests), "YouTube follow-on views", and "Clicks to call" — alongside the real booking action. Meanwhile the booking is counted twice (a GA4 purchase import AND a native thank-you-page action both enabled). When Dr. Dasari scaled spend and Google "just spent the money" on junk — this is the mechanism. The algorithm was told menu views are worth the same as bookings.
The famous YouTube retargeting pool is starved. "MGI YouTube Video Viewers 365days" holds 710 people. The site-remarketing lists (UC/CD 90-day) are CLOSED with zero members. Retargeting printed money when the channel fed tens of thousands of warm viewers into these lists; the early-2025 algorithm collapse cut the feed, the lists drained, and paid search has been buying cold strangers at 5–10× the cost ever since. This is the deep WHY behind rising CAC.
1.The timeline — three eras and a collapse
Bars are monthly spend; the line is account-reported cost per conversion. Shaded: Era 1 (2024), the golden window (Aug–Nov 2025), and the degradation (May–Jul 2026). The account was fully dark Sept 2024 → Jun 2025.
Era 1 (Jan–Apr 2024): exact-match practitioner keywords → per-condition /service/ pages. Conversion counting was looser then (page-view goals), so the $14 CPAs aren't comparable — but 6% click-to-conversion rates were real. Golden window (Aug–Nov 2025): $6–8k/mo, $218–380 per booking, phrase-match treatment terms → the fix-ibd-naturally page. The degradation (Dec 2025 → Jul 2026): three campaigns consolidated into one (Jan–Feb), PMAX added (Apr), spend scaled from $8k to $20k (May–Jun) — and conversion rate collapsed 0.92% → 0.28% while CPCs fell. Cheaper clicks, far worse buyers: the signature of loosened matching + polluted bidding signals, not an expensive auction. Dr. Dasari's instinct ("the same terms got 2–4× more expensive when I raised spend") was correct — the account bought progressively worse matches to absorb the budget.
2.What "a conversion" currently means
31 conversion actions exist; the enabled set below. Highlighted rows are PRIMARY — what smart bidding actually optimizes toward.
Action
Role
Category
Counting
Lookback
Contact Us (All Web Site Data)
secondary
CONTACT
one per click
30d
Email Submitted (All Web Site Data)
secondary
CONTACT
one per click
30d
Lupus Free Training (All Web Site Data)
secondary
GET_DIRECTIONS
one per click
30d
Call From Ad Extension
secondary
PHONE_CALL_LEAD
many per click
60d
Call Only Ads Hernia
secondary
PHONE_CALL_LEAD
one per click
30d
Call Ads
secondary
PHONE_CALL_LEAD
many per click
30d
Clicked Calendly Button
secondary
BOOK_APPOINTMENT
many per click
90d
www.mgiclinic.com - GA4 (web) purchase
secondary
PURCHASE
many per click
90d
YouTube channel subscriptions
secondary
ENGAGEMENT
one per click
30d
Form Submission
secondary
SUBMIT_LEAD_FORM
many per click
30d
Schedule A Discovery Call Button Click
secondary
CONTACT
many per click
30d
MGI Clinic (web) purchase
secondary
PURCHASE
many per click
90d
Website Click to Call
secondary
CONTACT
many per click
30d
Thank_You_Page_Transactions
secondary
PURCHASE
one per click
90d
Calls from ads
secondary
PHONE_CALL_LEAD
many per click
30d
Submit lead form (Page load thank-you-confirmation)
PRIMARY
SUBMIT_LEAD_FORM
one per click
90d
Discovery - Page View
secondary
PAGE_VIEW
one per click
90d
YouTube follow-on views
PRIMARY
UNKNOWN
many per click
30d
Android installs (all other apps)
secondary
DOWNLOAD
one per click
30d
Local actions - Website visits
PRIMARY
PAGE_VIEW
many per click
30d
Local actions - Other engagements
PRIMARY
ENGAGEMENT
many per click
30d
Local actions - Directions
PRIMARY
GET_DIRECTIONS
many per click
30d
Clicks to call
PRIMARY
CONTACT
many per click
30d
Local actions - Menu views
PRIMARY
PAGE_VIEW
many per click
30d
Fix: exactly one primary (the booking), everything else secondary; kill the GA4/native double-count; then import program registrations ($6,800) as offline conversions with a 10-day window so bidding learns toward buyers, not bookers.
3.Who the buyer is — intent economics
Every search term bucketed by intent. The gradient is unambiguous and confirms Dr. Dasari's field observations with 19 months of data.
Intent bucket
Spend
Conv
CPA
CVR
treatment-seek
$19,412
133.5
$145
1.23%
diet-supplement
$15,889
101.8
$156
1.20%
practitioner
$13,415
276.6
$48
4.61%
other
$9,796
72.5
$135
1.56%
condition-info
$3,113
16.5
$189
0.87%
drug-side-effect
$745
0.3
$2,234
0.14%
brand
$113
0.0
—
0.00%
trigger-onset
$37
0.0
—
0.00%
pediatric
$17
1.0
$17
14.29%
Full period 2024→now. Practitioner-intent converts at 4.6% — 3–9× every other bucket. Drug/side-effect terms: $2,234 per conversion. Trigger-onset and pediatric — the two highest-converting angles from clinical experience — have effectively never been run ($54 combined spend).
The buyer in one sentence: a US mobile user searching for a person — "crohn's disease specialist near me" (209 conversions, $10 each, the best term in account history) — not a diet, not a supplement, not information. "Near me" appears in 238 conversions: they think locally even though the program is remote. The winning page must bridge "specialist near me" → "the specialist who works with patients everywhere." A second, smaller buyer searches permanent escape — "how to get colitis to go away" (14 conv), "how to cure ibs permanently" (9) — language about ending the disease, never "managing" it. And they search for stories: "how i cured my crohn's disease" — first-person narrative demand that maps directly to the patient-turned-surgeon story.
4.The consumer's own words
Top converting terms (full period)
Conv
Spend
CPA
crohn's disease specialist near me
209.0
$2,179
$10
how to get colitis to go away
14.0
$260
$19
ulcerative colitis treatment
10.0
$990
$99
colitis treatment
9.0
$942
$105
how to cure ibs permanently
9.0
$210
$23
treatment for ulcerative colitis
7.7
$218
$28
how to control crohns
6.0
$169
$28
ibs specialist near me
5.3
$81
$15
crohn's disease treatment
5.0
$646
$129
colitis treatments
5.0
$278
$56
español medication for ulcerative colitis flare up
5.0
$96
$19
treatment of ulcerative colitis
5.0
$63
$13
gastroenterologist ulcerative colitis
4.0
$121
$30
ulcerative colitis specialist near me
3.3
$432
$130
crohn disease treatment medication
3.0
$65
$22
over the counter relief for ulcerative colitis
3.0
$29
$10
ulcerative colitis treatments
3.0
$370
$123
best ibs doctor in the world
3.0
$5
$2
irritable bowel syndrome
3.0
$46
$15
e n t y v i o
3.0
$283
$94
Top wasted spend (zero conversions)
Spend
Clicks
ibd specialist near me
$352
108
how to stop ulcerative colitis flare up
$274
71
holistic doctor for crohn's disease
$257
29
crohn's disease specialists
$237
26
crohn's remission without medication
$207
76
ulcerative colitis cure
$181
95
crohn's specialist
$174
49
how to cure ulcerative colitis
$157
90
healing ulcerative colitis
$155
86
best doctors for crohn's disease
$154
6
what is the latest treatment for ulcerative colitis
$152
93
crohn's disease treatments
$141
70
is there a cure for crohn's disease
$135
65
how to cure crohn's disease
$132
68
ulcerative colitis specialists
$129
40
holistic doctor ulcerative colitis
$126
33
Waste themes: acute "flare up" crisis searches (~$1.3k — an ER mindset, not a program buyer), "cure"-only researchers, competitor clinics (Mount Sinai), and one-click local leaks. This is the negative-keyword list, pre-written.
Converting n-grams
Conv
Wasting n-grams
Wasted
crohn's disease
243
ulcerative colitis
$12,637
near me
239
crohn's disease
$5,326
specialist near
223
flare up
$1,360
specialist near me
223
colitis flare
$1,284
disease specialist
210
near me
$1,277
crohn's disease specialist
210
colitis flare up
$921
disease specialist near
209
treatment ulcerative
$898
ulcerative colitis
94
ulcerative colitis flare
$889
colitis treatment
22
treatment ulcerative colitis
$884
treatment ulcerative
21
colitis treatment
$832
treatment ulcerative colitis
20
cure crohn's
$666
go away
15
treatment crohn's
$633
Question queries (pain-point language)
Clicks
Conv
how to cure ibs permanently
293
9.0
how to get colitis to go away
241
14.0
how i cured my crohn's disease
177
2.5
how to control crohns
146
6.0
how to heal ulcerative colitis
111
1.0
how to cure ulcerative colitis permanently
97
1.0
what is the latest treatment for ulcerative colitis
93
0.0
how to cure ulcerative colitis
90
0.0
how i cured my ulcerative colitis
88
1.0
how to stop ulcerative colitis flare up
71
0.0
how to cure crohn's disease
68
0.0
is there a cure for crohn's disease
65
0.0
how to heal crohn's disease holistically
48
0.0
what is the best medicine for ulcerative colitis
45
1.0
Copy source material: control ("how to control crohns"), permanence ("go away", "permanently"), first-person proof ("how i cured my…").
5.Keywords — where the money went
Keyword
Match
QS
Conv
Spend
CPA
Campaign
crohn's disease specialist near me
Exact
—
118.0
$1,560
$13
Traditional_Search_Crohns_
ibs
Phrase
—
101.5
$2,826
$28
Traditional_Search_IBS_Lea
crohn's disease doctor
Exact
—
100.5
$695
$7
Traditional_Search_Crohns_
inflamed large intestine treatment
Phrase
3
41.0
$30,555
$745
Ash-UC Specialist-Campaign
treatment for intestinal inflammation
Phrase
—
38.8
$10,559
$272
Ash-Intestinal Inflamation
inflamed large intestine treatment
Phrase
—
35.5
$12,544
$353
Ash-Intestinal Inflamation
crohn's specialist
Phrase
3
30.9
$13,319
$431
Ash-UC Specialist-Campaign
treatments for colitis flare ups
Phrase
—
25.0
$1,064
$43
Traditional_Search_Ulcerat
ulcerative colitis treatment
Exact
—
19.0
$330
$17
Traditional_Search_Ulcerat
doctor for ulcerative colitis
Phrase
4
18.7
$6,895
$368
Ash-UC Specialist-Campaign
holistic treatments for colitis
Phrase
—
15.0
$802
$53
Traditional_Search_Ulcerat
colitis treatment
Exact
—
14.0
$403
$29
Traditional_Search_Ulcerat
how to get colitis to go away
Exact
—
14.0
$340
$24
Traditional_Search_Ulcerat
treatment for intestinal inflammation
Phrase
4
13.0
$5,170
$398
Ash-UC Specialist-Campaign
natural remedies for ulcerative colitis
Exact
5
12.3
$3,843
$312
Ash-UC Specialist-Campaign
crohn's disease doctor
Phrase
—
12.0
$1,189
$99
Ash-Intestinal Inflamation
The 2024 exact-match practitioner terms delivered cheap conversions. The current era's top spender — "inflamed large intestine treatment" phrase — burned $30,555 for 41 conversions at Quality Score 3 inside the consolidated UC campaign; the same keyword ran $353/conv (and "treatment for intestinal inflammation" $272/conv) inside the dedicated Intestinal Inflammation campaign before the March 2026 consolidation — the container degraded, not the keyword. Only 34 of 891 keywords have enough traffic for a QS reading; those that do average 3–5 (poor ad-and-page relevance → inflated CPCs).
Cross-era caveat: the /service/ pages' low CPAs partly reflect 2024's looser conversion counting. Still, fix-ibd-naturally absorbed $100k at $395/conv in the modern era — that's the benchmark our new pages have to beat. URL variants (http/www/bare) fragment the data — standardize final URLs. Correction (Aug 15): the /service/… URLs are host-scoped redirect ALIASES of the real pages /crohns-disease/ and /ulcerative-colitis/ — the rows are the same content split across two URL forms; combine them when reading this table (the “$2,719/conv duplicate” was remarketing traffic hitting the canonical page directly, not a separate bad page).
Country
Spend
Conv
CPA
US
$135,668
1093.0
$124
UK
$12,424
24.0
$518
CA
$8,363
13.0
$643
UAE
$1,929
7.0
$276
2528
$623
1.0
$623
2682
$359
2.0
$180
DE
$324
0.0
—
AU
$18
0.0
—
UAE: $276/conv on just $1.9k spent — underfunded relative to its close-rate reputation. UK/CA convert 4–5× worse than US.
Remarketing list
Type
Search
Display
Status
AdWords optimized list
Logical
0
0
OPEN
All Users
Remarketing
0
0
CLOSED
All visitors (AdWords)
Rule Based
0
0
CLOSED
Similar to All Users
Similar
0
0
CLOSED
Similar to AdWords optimized list
Similar
0
0
OPEN
Purchasers of www.mgiclinic.com - GA4
Remarketing
24
16
OPEN
MGI YouTube Video Viewers 365days
Rule Based
710
0
OPEN
MGI YouTube Users Revised
Rule Based
710
0
OPEN
YouTube Video Viewers *Specific Videos
Rule Based
0
0
OPEN
UC - Remarketing 90 Days
Rule Based
0
0
CLOSED
Remarketing Audience - 90 Days
Rule Based
0
0
CLOSED
CD - Remarketing - 90 Days
Rule Based
0
0
CLOSED
The flywheel graveyard: YouTube pools at 710, site lists CLOSED at zero. Rebuilding these is a content problem (feed the channel), not a bidding problem.
7.The plan — in priority order
Fix conversion goals first (day one, before touching bids). One primary action: the booking. Demote local actions, YouTube follow-on views, and click-to-call to secondary; remove the GA4/native double count. Every other change is built on this.
Un-throttle August. With clean goals, remove the $200 tCPA from Maximize Conversions (his own tested learning: hard caps starve the account) and restore delivery; watch cost per booking, not cost per blended conversion.
Wire the real feedback loop. GCLID capture at booking → offline import of registrations ($6,800, ~10-day window). This is the only way bidding ever learns "optimize for the program, not the call."
Rebuild the practitioner core. Dedicated exact/phrase campaigns per condition on the proven winners ("[condition] specialist near me", "doctor for [condition]") → per-condition pages (the /service/-page structure won before; our new LPs take that slot).
Ship the negative-keyword program. Flare/crisis terms, drug names, competitor clinics, research-paper language — the waste table above is the list.
Contain PMAX until clean conversions exist; re-judge it only on booking-CPA + close rate afterward.
Launch the two never-run proven angles. Pediatric and trigger-onset (antibiotics/Accutane framing in ads) — highest close rates in clinical experience, zero historical ad spend. Needs the new LPs.
Capture brand. Zero brand terms are bid today (16 clicks all period). Pennies to own, and the only way paid search harvests the YouTube flywheel when it restarts.
Scale UAE deliberately — $276/conv on tiny spend + the strongest close-rate geo per clinical experience. Watch call quality by geo through the offline loop.
Restart the audience flywheel. Reopen site lists, refresh the YouTube-linked lists, and feed the channel (scripts + shot lists already in motion) — retargeting can't work on a 710-person pool.
8.Guidance × data — cross-examining the brief
Every operational claim from Dr. Dasari's brief and the kickoff call, tested against the 19-month dataset. Where the account's own history agrees, where it refines the guidance, and where the account is violating instructions it was given.
9
confirmed by the data
2
refined — guidance right, details sharper now
2
untested — the account never ran them
3
action required — account contradicts guidance
Confirmed
“Intent-focused search terms do the best — specialist, doctor for Crohn's / UC / intestinal inflammation.” — Brief §2
Emphatically true: practitioner-intent runs $48/conversion at 4.61% CVR — 3–9× every other intent, and the best term in account history is “crohn's disease specialist near me” (209 conversions at $10). Two refinements the data adds: the load-bearing words are specialist/doctor, not holistic (“holistic doctor for crohn's disease” alone wasted $257 at zero conversions recently) — and in the last 90 days practitioner terms spent $2,589 with ZERO conversions. The thesis is right; the current execution of it is broken.
Confirmed
“Diet and supplement focused search terms close at a very very low rate.” — Brief §2
$156/conversion at 1.20% CVR — 3× worse than practitioner intent at the booking step, before the close-rate penalty he observes on calls even begins. Notably, the current campaign was violating this guidance: “best food for ulcerative colitis” burned $3,200 for 10.8 conversions.
Confirmed
“Side effect queries close at a very very low rate — steroids, mesalamine, biologics, drug names.” — Brief §2
The data is even harsher than the guidance: $2,234 per conversion, 0.14% CVR across 765 drug/side-effect terms. Hard negative-keyword territory.
Refined
“P-max just sucks dry the spend, bad close rates.” — Brief §3
The one claim the booking-layer data appears to contradict — and shouldn't be trusted on: Apr–Jul 2026, IBD-PMAX shows $85/conversion vs $573 for the search campaign. But PMAX's “conversions” are the polluted blend (local actions, engagement events), its search terms are invisible to audit, and the close rate — his actual complaint — is invisible to Google. Verdict: the doctrine stands until the offline registration loop exists to judge PMAX on real buyers. Contain it; re-try it only against registration data. Also noted: it was launched in April against his standing instruction.
Confirmed
“Retargeting people who interacted with my YouTube videos — very high close rates, very low lead costs.” — Brief §3
The guidance is right and currently impossible: the infrastructure exists but “MGI YouTube Video Viewers 365days” holds 710 people, and both site-remarketing lists are CLOSED at zero members. Retargeting worked when the channel fed the pool; the 2025 algorithm collapse cut the feed. Rebuilding this is a content problem (feed the channel), not a bidding problem.
Confirmed
“Eliminating high-CAC search terms resulted in higher CAC across the board — we did better leaving them in.” — Brief §3
The timeline agrees: the consolidation from three campaigns to one (Jan–Feb 2026) plus term cleanup directly preceded the conversion-rate collapse from 0.92% to 0.28%. Mechanism: with polluted conversion signals, pruning removes the good anchors and Maximize Conversions re-spends the budget on cheaper, worse matches.
Confirmed
“We have not done great once limiting the CAC upper limit ($200 for example).” — Brief §3
Correct — and the experiment was never reverted. The $200 target CPA is STILL SET on the main campaign and is the proximate cause of the August blackout: smart bidding starves delivery it can't price. Removing it is step 2 of the plan (after the conversion goals are cleaned, step 1).
Confirmed
“The sweet spot seems to be around $200/call, which brings cost per registration to ~$800. Best timeframe: Sept–Nov 2025.” — Brief §3
The golden window is real, with one date refinement: August 2025 was actually the single best month ($218/booking), with Sep–Nov at $243–380. Degradation began with the December budget jump and went terminal in May–Jun 2026 ($495 → $639).
Confirmed
“Very poor luck running YouTube ads — targeting individuals who watched certain videos didn't work.” — Brief §4
Nothing contradicts it: the cold-YouTube campaigns spent $275 lifetime with nothing meaningful converted, then were removed. Thin evidence either way — but no reason to re-test cold video while the search core and remarketing pool are unfixed.
Untested
“People whose UC / Crohn's / IBD followed vaccination, antibiotics, or Accutane convert at a very high rate.” — Brief §5
The account has never tested this: 229 matching queries drifted in organically, $37 total spend, zero targeting or dedicated pages ever built. Clinical experience says it's the highest-converting angle that exists; the ads account can neither confirm nor deny. This is the biggest untapped test in the plan — with the agreed compliance framing (antibiotics/Accutane in ad copy, patient-voice specifics on the page).
Untested
“Pediatric patients — kids as young as 4 — great great results.” — Brief §5
Same status: $17 lifetime spend, 1 conversion on 7 clicks, no campaign has ever existed. A dedicated parent-targeted build is greenfield.
Refined
“Geographic regions with highest close rates: UK, Can, UAE, NZ, AUS, US.” — Brief §6
At the booking layer the data reorders the list: US $124/conversion on $136k (the engine), UAE $276 on a token $1.9k (underfunded — scale-test it), UK $518 and CA $643 (4–5× worse per booking), AU/NZ effectively no data. His list may still be right about close rates on the phone — but UK/CA can only be justified once offline registration data proves their close rates compensate for 4–5× acquisition costs. Sequence: US core → UAE test → UK/CA pending evidence.
Action required
“We had higher return requests when we advertised refund guarantees in the initial copy.” — Brief §7
Current live copy is clean — but two still-enabled ads inside the paused 2024 UC campaign read “Real Recovery Guaranteed.” Dormant landmine: if that campaign is ever reactivated, guarantee copy goes live again. Scrub before any restructure touches old campaigns.
Confirmed
“The same exact search terms, one month ago, were half or a quarter of the price. Every time I increase spend it goes up — that shouldn't happen.” — Kickoff call
Confirmed, with the mechanism identified: CPCs actually FELL during the scale-up while conversion rate collapsed — the auction didn't get more expensive, the account absorbed budget by buying progressively worse matches, priced against polluted conversion signals. His instinct that “we're screwing something up” was exactly right.
Action required
“Optimize for purchases of the full product ($6,800), not just the discovery calls ($50).” — Brief §1
The account implements none of this: primary conversions are polluted, the booking is double-counted, no GCLID is captured at booking, and no offline registration import exists. This is the single biggest gap between guidance and account reality — and it's the first two steps of the plan plus the Calendly/Kajabi integration currently waiting on access.
Action required
“You can't say cure, can't say fix, can't say heal, can't say reverse.” — Kickoff call
His own rule, violated by his own live ads: “cure” appears 18 times in running RSA copy, “reverse/reversing” 6×, “fix” once — and 17 ad URLs point at /fix-ibd-naturally/, a banned word baked into the slug. Fully consistent with the CPM-penalty pattern he described on Meta. Rewrite the RSA copy set, and give the replacement landing pages a compliant slug.
The headline of the cross-examination: Dr. Dasari's field judgment survives contact with his own data almost perfectly — 9 of 16 claims confirmed outright, and the two “refinements” sharpen rather than overturn him. The failures are all in implementation: the account never operationalized his #1 instruction (optimize for registrations), never tested his two highest-conviction angles (pediatric, trigger-onset), left his rejected tCPA experiment running, and runs ad copy his own compliance rules ban. The strategy was never wrong. The account just stopped listening to it.
Method notes: data via Google Ads API v21, GAQL, pulled Aug 15, 2026. "Conversions" = account-reported primary actions of each era; cross-era CPAs are not directly comparable (counting definitions changed). PMAX search terms are not exposed by the API. Change history covers the trailing 30 days only (24 changes, all by the incumbent manager).