A candid summary of what our July audit found, all of it historical and all of it addressable:
None of this is unusual in an account that has passed through several hands, and none of it requires looking backward. It is simply the starting line.
Every keyword, search term, conversion goal, and setting reviewed. Findings and proposed fixes documented and approved before anything changed.
A 456-term negative keyword shield now blocks searches for other institutions, subprime lenders, and junk queries. Automatic settings Google had quietly enabled were turned off. Brand advertising moved to tightly controlled manual bidding at a fraction of its old cost.
August spend reset to the agreed plan, weighted toward the loan products CommonWealth One most wants to fund, with brand held at an efficient floor.
54 more phrase-match negatives added: other institutions people searched by name, subprime and lease-to-own lenders, and Spanish-language searches the site cannot yet serve. The shield now stands at over 500 terms.
Two more negative batches loaded (91 additional terms): competitor institutions searched by name, subprime and desperate-credit lenders, and spelling variants that slip standard matching. August spend is pacing near $8K against July’s $12.2K, with cost per conversion down about 18 percent since the July 27 changes and brand clicks now averaging $0.14 against $0.72 before the cap.
Full review published (see the Aug 21 review on the hub). The shield reached 693 terms after three more rounds this week. Cost per completed application is down 34 percent versus July, and the client has asked to increase budgets: Personal Loans to $4,000, Auto to $5,000, and a new Credit Cards campaign at $1,500, all supported, with the card campaign launching behind its own conversion tracking. The single biggest open item is landing page implementation: the rebuilt pages have been ready since July 28, and four months of data now show non-brand traffic completes applications only when the searches were low-quality ones we have since blocked. Clean traffic needs a page built for it.
Personal Loans moved to $132 a day (about $4,000 a month, the credit union's number). Auto moved to $138 a day as the first of two steps toward its $5,000 target, so the campaign could absorb the increase without its cost per conversion drifting.
The second step, to $164 a day, once the Auto campaign's cost per conversion had settled after the first increase. Negative keyword rounds continued on a two-to-three-day cadence through this stretch; the shield passed 800 terms on September 1 and reached 969 by September 11.
The consumer lending team asked for more application volume to keep underwriting queues active, accepting that a higher volume of applications tends to come with a lower funding ratio. Auto moved to $225 a day for the rest of September so the month lands the full $6,000 despite starting at the lower level, and steps back to $197 a day on October 1, which is $6,000 a month from there on. Through September 15 the Auto campaign was reaching only 13 percent of the searches it qualifies for, so the extra budget has room to work.
A new Search campaign for the Visa Signature Rewards card went live at $125 a day so the full $1,500 spends inside September: exact and phrase keywords only, the same geography as Auto, and a 41-term block on servicing searches (login, payments, activation) where card campaigns typically leak. The landing page HTML is delivered and previewable on the mock site. One item is with the client: cofcu.org lists the card's rewards rate as 1.25 points per dollar on one page and two points per dollar on another, so neither the ads nor the page state a rate until the credit union confirms which is correct.
The Visa Signature Rewards campaign and the rebuilt Kasasa Checking campaign are running on click-based bidding with firm cost-per-click ceilings ($10 and $8) through the end of September, the standard way to seed a campaign that does not yet have conversion history for Google to learn from. Both move to conversion-based bidding on October 1. Auto's daily budget was set to $200 so the month lands on the $6,000 target. Negative keyword round eleven loaded (41 terms); the shield now runs to over 1,100 terms across the account and shared lists.
Every search that spends money is reviewed daily against the shield and new negatives are loaded in rounds, ten rounds so far. Wasted spend is now the exception rather than a line item.
August was the first full month on the cleaned-up account, so it is the first fair comparison: more applications, level fundings, at 70 percent of July's spend. September is deliberately bigger. The Auto and card budgets are new enough that their numbers are still a floor, and the honest read on them comes in early October once the funding uploads catch up. One structural note still holds: most conversions come from people already searching for CommonWealth One by name, and the growth opportunity is non-brand search, which is where the cleanup, the new card campaign, and the rebuilt landing pages are all aimed.
Measurement is the foundation everything else stands on, so we treat it as an audit subject, not an assumption. What has been checked so far: the tracking handoff from cofcu.org to the loan application system was click-tested live and works correctly; every landing page ads point to was verified loading and functional; and on August 3 we cross-checked Google Ads against Alpharank's reporting for July, and the approved and funded counts matched. We also identified a refinement worth making: deposit-account fundings and loan fundings currently share one bucket, and separating them will let each campaign optimize to its own true outcome. We are coordinating that with Alpharank now.
This program touches systems owned by several teams: the website, the loan application platform, and Alpharank's analytics. Our approach is to add value on top of those partners rather than around them. When reporting has not fit what the program needs, we have asked for specific adjustments and gotten them, and as the new landing pages come online we will use them to sharpen both measurement and retargeting. If there is data anyone on the CommonWealth One side wants more visibility into, tell us and we will build the view or request it from the right partner.
Auto finishes the month at its full $6,000 and the card campaign spends its $1,500. The card campaign gets a first read after one week on calls and search terms, since online applications only become visible as the uploads land. The rewards-rate question on cofcu.org gets settled so the card's strongest selling point can go into the ads.
Deposit-account fundings and loan fundings still share one bucket in the upload from Alpharank. Separating them lets the checking campaign optimize to funded deposits and the loan campaigns to approved and funded loans. October is also the first clean month to read the Auto increase against what the lending team asked for: application volume up, and a look at where the funding ratio settles.
The three rebuilt loan pages have been ready since July 28 and the card page since September 18. Every one of them removes the full site navigation and puts the rate and one apply button in front of the visitor. Implementation timing sits with the web team; the day they go live, the ads point at them and non-brand search finally has a page built for it.
Non-brand search expands as its cost per funded outcome comes down. Performance Max stays paused until the account is optimizing for the right person; turning it on earlier would simply buy more of the wrong traffic.