Where the program stands

The reset is done. Profit is dated.

Three outside hits moved the August 15 date: a broken tracking link on the buyer's side, a Google block on BAR from a legal filing, and a month-long wait for the client lists the strategy was built on. None of the three were advertising errors. All are fixed or contained, the instruments are honest for the first time, and the engine now runs under a written pricing law: first profitable Search arms by September 1, breakeven or better blended in the first week of September.

Aug 13 · done

Clean data

per-channel tracking live; first honest attribution ever

→

Aug 25

Checkpoint

12 clean days; every arm repriced at its measured value

→

Sep 1

First profit

$1-2k/day of Search at or above ROAS 1.05

→

Sep 7

Verdict week

blended ≥ 1.05; each arm scales or dies on its numbers

The scale target: $20,000/day on Search at ROAS ≥ 1.05, with Search carrying at least 80% of account spend.

What moved the date. None of it ours.

Profitability is a pricing problem, and pricing needs honest measurement, two working accounts, and the audience data the strategy was designed around. All three were taken away at once, and each one directly limited how fast anything could be tested.

Hit 1 · the broken link

Every tracking URL resolved to the Search UserKey

Separate tracking URLs existed for Display and Demand Gen, but on the buyer's platform each one routed to the same Search UserKey. Whatever channel produced a lead, it landed on the Search identity: Search read a false 0.26 while its leads actually price at the top of our book, buyers bid the polluted key down, and Google's bidding algorithms trained on the wrong payouts. Every test run before August 13 was measured on corrupted instruments.

Fixed same day disclosed. Per-channel keys and routes live Aug 13; campaigns recreated; clean attribution since.

Hit 2 · the BAR block

Google suspended BAR over a legal filing

BudgetAutoRates failed business verification on August 6 over an incorrect legal registration prepared outside ad operations. No ad, targeting or landing page was cited. The API refuses every write, and the account is not expected to come back. Half the test surface is permanently gone: experiments that ran in parallel across two accounts now run one at a time.

Written off operationally. The program plans on CAQ alone; nothing in the plan waits on BAR.

Hit 3 · the missing lists

The client lists took a month to reach us

The value strategy prices audiences by what their leads historically earned, which requires the CRM client lists (Clients by earnings tier) as Customer Match audiences. The upload sat with the dev team for about a month, landed July 31, then took Google roughly a week to process, neither step in our control. Until then the entire tiered-targeting program was unbuildable.

Processed ~Aug 7, in live targeting by Aug 12, and the analysis below is where it lands us: we can now see who is worth more.

What the blind stretch cost: on roughly $135k a month of media spend, the 30 days to August 6 closed about $85k underwater across both accounts, nearly all of it steered by the corrupted numbers. Since August 6 the testing cadence runs on one account instead of two. The response since disclosure: the account was re-routed, re-priced from measured lead value, and stripped of dead weight (212 dead campaigns removed) inside four days.

The honest numbers are good news.

Five days of clean attribution, plus a platform-wide pull of the buyer's own marketplace, settle the oldest question in this program: the leads were never the problem.

Platform blend

$22.21

per lead · 624 accounts, 830,614 leads, 30d

Our Search route

$26.73

per lead · clean era, still warming

Sibling brand Search

$40.41

same buyer, warm route: the near-term bar

Best auto at scale

$42.15

proof the band is reachable at volume

Value per lead · by channel route, clean era

Search$26.73

Display$16.03

Demand Gen$13.34

Search, warmed + narrowed$33-45

Search leads price at a premium. The broken link hid exactly this: the premium was being blended away and bid down.

The gap is buying, not quality

95% of market median value: what our leads already sell for. Quality was never the problem. 2.3× market cost per lead: what the corrupted era paid for clicks. This is the whole gap.

The biggest account on the platform runs 62,697 leads a month at $14.92 cost per lead against $18.33 earned, roughly $31k a day of spend at ROAS 1.23. The profitable-at-scale operating point exists on this exact marketplace; the rebuild's whole job is to buy our way onto it.

We now know who the $60+ lead is.

The delayed client lists, once uploaded, unlocked the analysis this strategy was waiting for: our own book, bucketed by what each client actually earned, then profiled. The premium segment is real, sizable, and now targetable.

Clients who earned $100+

4,800

the premium tail is in OUR book

Clients who earned $60+

32,000

about 1 in 18 of listed clients

Listed clients, total

582k

bucketed by earnings tier

Lookalike pools built

6-25M

seeded from high-earners: the reach instrument

The six customers, by what their leads earned

Illustrative portrait of the $0-20 tier customer

$0-20990,000 clients

The active insurance shopper. In-market for auto insurance indexes 3.0x here, the highest of any tier. Price is everything.

Illustrative portrait of the $20-40 tier customer

$20-40440,000 clients

Still a price shopper: life and health insurance searches, tax prep, wheels, tires and repair bills. The volume tier.

Illustrative portrait of the $40-60 tier customer

$40-60100,000 clients

The first owner tier: pickup and truck households, classic vehicles, lawn care, repair-heavy but valuable.

Illustrative portrait of the $60-80 tier customer

$60-8029,000 clients

The vehicle enthusiast: shopping new vehicles, boats and water sports, tools and outdoor gear, performance and luxury.

Illustrative portrait of the $80-100 tier customer

$80-1008,200 clients

The established homeowner: home improvement and decor, financial services, active investors, family households.

Illustrative portrait of the $100+ tier customer

$100+4,800 clients

The small-business owner: accounting software is the strongest single marker anywhere, real estate, software, investors.

Portraits are illustrative composites of each tier's measured interest indexes (Customer Match audience insights, Aug 14), not real customers.

The $70+ signature (found Aug 14)

High-earning clients index on small-business and homeowner signals: accounting software, real estate, home improvement, investing; plus a vehicle-enthusiast band (trucks, SUVs, performance) in the $40-80 tiers. This is a buyable profile, and it is now wired into the live audience boxes.

The inversion nobody guessed

People actively shopping for auto insurance index 3.0x among our cheapest clients and vanish above $70. The head query buys the bargain hunter. The premium lead is the owner who also needs insurance, not the person hunting the lowest price. That single fact reorients the whole buying strategy.

Finding, clean era	Number	What it becomes
Specialty demand converts cheapest: salvage and rebuilt title, commercial vehicles	$9-23 /conv	MARGIN LAYER
Senior demand isolated into its own arm, priced at its measured value ($26 vs $27.31 earned)	live Aug 17	PRICED HONEST
Time of day tested: lead value is flat by hour; no hidden daypart problem exists	$16-24 band	MYTH RETIRED
Junk classes identified with real numbers: carrier-brand shoppers, out-of-vertical queries	10.7% of spend	BEING CUT

The ceiling, found. And the unlock.

Two structural facts from the buyer's own data explain why earnings per lead plateau near $27, and what raises the ceiling rather than the mix.

Fact 1 · the buyer prices the route, not the lead

Every one of our Search routes clusters at $26-28 per lead regardless of which audience produced the lead. Narrower targeting alone cannot outrun a route's price: premium leads sent down the same pipe sell at the same rate. Premium lanes need their own routes to be priced on their own merit.

Fact 2 · we sell to one buyer of 67

The marketplace carries 67 buyer networks. 100% of our revenue flows through one. Operators who auction each lead across many buyers take the highest bid on every lead; single-homing takes one buyer's list price. This is the most plausible mechanism behind the $100+ accounts, and it is a routing change, not an advertising change.

The program therefore climbs on two legs. Curation feeds better leads: the owner profile, specialty classes, lookalikes of proven high-earners, junk cut, every arm priced at its measured value divided by 1.05 so profit is the constraint and volume is the variable. Monetization raises what the same leads sell for: premium lanes split onto their own routes, then multi-buyer auctioning. Curation is live today. Monetization is scoped and waiting on a routing decision, not on ad work.

The plan. Exact dates, hard gates.

Money now moves only through gates. Losses are capped by calendar, not by hope, and every date below is already in the operations journal.

When	What happens	Gate
Aug 17 · done	Every Search arm priced at measured value ÷ 1.05; query hygiene applied; senior arm split out.	LIVE
Aug 19	Discovery gate: the two raised probes multiply impressions or their raises revert.	AUTO-CHECK
Aug 21	Demand Gen value pilot beats its siblings or dies. Loss already capped.	KILL-LINE
Aug 25	Checkpoint 1, first 12-day clean window: every arm repriced on settled earnings; schedule carves applied only where the data convicts; budget follows winners.	DECIDES
Sep 1	First profitable arms: $1-2k/day of Search holding ROAS ≥ 1.05.	TARGET
Sep 7	Verdict week: Search blended at or above 1.05; every arm scales or dies on its own numbers.	TARGET
Late Sep	Scale readout: $3-6k/day near breakeven; the ramp toward the $20k/day Search target set by measured price curves around Oct 1, not promised.	READOUT

LAWPricing law: every arm's bid target is capped at its settled earnings per lead ÷ 1.05. An arm grows only by proving its leads earn more.
TRUTHRevenue truth: verdicts use buyer-reported revenue over Google-reported cost, nothing else. Google's own value estimates already produced one false diagnosis and are excluded.
DISCIPLINEChange discipline: every mutation dry-run first, applied only on explicit approval, verified by readback, journaled. One move per arm per four days so every effect stays attributable.

Aug 19 gate Aug 21 kill-line Aug 25 checkpoint Sep 1 first profit Sep 7 verdict Oct 1 scale decision

How you will track it.

Five numbers, reported every Monday on clean attribution. They move, or they do not.

Metric	Today	First week of September
Search route value per lead	$26.73 · warming	into the $33-45 band
Search cost per lead	~$82 · relearning	$26-34 viable band
Arms holding ROAS ≥ 1.05	0 · all in learning until ~Aug 22	every surviving arm
Search net per day	−$300-580 · buying clean data	positive
Search share of account spend	~30% today	climbing toward ≥ 80%

Past September 7, two numbers carry the story: value per lead climbing as routes warm, lanes split and the owner profile takes share; and spend rising toward the $20,000/day Search target only where the 1.05 constraint holds, with Search taking at least 80% of the account. Both reported the same way, every week.