August 3, 2026 · Kept for the record, updated September 18 · Account 730-631-9113 · Search-term window: Jul 27 to Aug 3 (since the first cleanup went live)
For the record. This round was approved and loaded on August 4. On the August 3 call, negative keyword work was delegated to us outright, so the rounds since (ten in total, list at 969 terms as of September 11, reviewed daily) have run without an approval step. Nothing on this page needs action. The live view is the Status & Plan page.
What the first round bought you
−30%
cost per conversion in week one: $30.34 before the Jul 27 cleanup, $21.22 after
$510
of the week’s $755 in search spend still went to searches in the buckets below
54
proposed phrase-match negatives, same playbook you approved on Jul 26
The first negative build closed most of the waste. Watching every search that spent money since, the remaining leaks fall into four clean buckets. The single biggest click of the week was $30 for "t d bank."
Funded outcomes, tracked
15
funded outcomes landed in the July upload batches (loans and new deposit accounts); 14 matched back to a specific ad click
$871
July cost per funded outcome, recovering fast: $510 in April, then the June outage drove it to $1,331; July clawed back to $871, and April proves where this account can get back to
1–3 days
between upload batches; the offline loop is alive and steady
The outcome loop works: funded and approved outcomes flow back from the application system and match to ad clicks. A timing note that works in July’s favor: a click takes one to three weeks to become a funded loan, so July’s 14 is still maturing: loans from late-July clicks are still closing and will raise that count, and none of the post-cleanup clicks have had time to fund yet. The first clean before-and-after read arrives mid-August.
The full application funnel
Month
Completed applications
Approved
Funded
Approval rate
Funded rate
April
127
25
19.7
20%
15%
May
98
13
10.1
13%
10%
June (outage)
84
8
8.2
10%
10%
July (as of Aug 3)
49
17
14.0
35%
29%
July (matured, by funding date)
54
21
18
39%
33%
August (by funding date)
67
20
16
30%
24%
September 1-17 (still maturing)
36
7
5
19%
14%
Rows added September 18. August and September count outcomes by the date they were approved or funded, as uploaded by Alpharank; September understates until its uploads land, which takes two to three weeks.
Completed is a submitted loan application; Approved and Funded are the outcomes that follow. July’s story in one line: half the applications of June, but the approval rate more than tripled. Part of that is mix, 29 of July’s 49 applications came from brand searches, and brand applicants approve at much higher rates. The negatives on this page push non-brand’s mix in the same direction, fewer applications that were never going to pass underwriting.
The August focus: more completed and approved loans. Grow completed applications back toward April’s pace (about 127 a month) while holding July’s quality. April’s volume at July’s approval rate is roughly twice the funded loans at the same budget, which is the whole game.
Where funded outcomes actually come from
Month
Brand funded
Cost each
Non-brand funded
Cost each
Blended
April
14.1
$64
5.5
$1,651
$510
May
7.0
$178
3.1
$2,963
$1,037
June (outage)
4.2
$175
4.0
$2,542
$1,331
July (still maturing)
12.0
$141
2.0
$5,246
$871
August (by funding date)
14.5
$42
1.5
$5,265
$531
September 1-17 (still maturing)
5.0
$73
0
–
$1,390
Historically, about 85 percent of funded loans with a visible search term come from brand searches (“commonwealth one,” “cofcu”). Brand capture is cheap and steady, the floor of this account. The growth lever is non-brand: driving its cost per funded from July’s $5,246 back to April’s $1,651, then toward $1,000. That is exactly what the negatives on this page, the August budget discipline, and eventually bidding to funded outcomes are built to do. The goal we suggest managing to: blended $510, the account’s proven April best.
One question worth deciding together: how much brand should this account buy? Two valid strategies exist: maximize cheap brand capture, or hold brand at its floor and put every incremental dollar into making non-brand fund at an acceptable cost. The right answer depends on what CommonWealth wants to grow and what they measure internally, and it is worth asking them directly.
Tracking health check, tested live the morning of Aug 3. Good news first: we clicked through the site to the application system and confirmed the cross-domain handoff generates fresh tracking at click time, so the funded-loan matching above stands on solid ground. One cleanup item remains: the apply links in the site HTML carry old copy-pasted tracking parameters from 2023–2024; the live handoff overrides them for normal visitors, but they leak through for visitors without JavaScript and anyone sharing a copied link, so they are worth removing (small site-side edit, already specced). Every landing page the live ads point to was also checked this morning: all load correctly with working apply links.
The four buckets, as loaded August 4
What stays live, on purpose
Real demand searches are untouched: "auto loan," "line of credit," "personal loans near me" and everything like them stay live, several converted this week. We also deliberately left off a few ambiguous terms, like "virginia federal credit union": someone searching that may genuinely want a federal credit union in Virginia, which is exactly who we want to reach. When in doubt, we leave demand alone.
0 of 4 buckets approved
The email you send is the written approval. Checking boxes alone approves nothing.