The ladder is real.
The tail is not the prize.

Bucket the CRM book by realised earnings per lead, hand each bucket to Google as a customer-match list, and read the profile back. Seven of eight bands cleared the insight threshold. Every trait that matters moves in one direction across the whole ladder, without a single reversal — which is rare enough in audience data to be worth stating plainly.

$0-20990,000 clients

The active rate-shopper. In-market for auto insurance indexes 3.2x here — the highest of any band — alongside life and health insurance. Oldest band (38% are 65+), most mobile-only (19% on a computer against a 40% US baseline), least likely to be a parent.

$20-40440,000 clients

Still shopping, starting to own. Insurance signal barely down (2.8x) but repair and parts appear: wheels & tires, transmission, collision. The volume band, and the one the Search route currently buys.

$40-60100,000 clients

The first owner band. Auto insurance falls to 2.4x while Chevrolet, classic and used vehicles, engine work and lawn care rise. Truck & SUV enthusiasts peak here at 2.3x. The reachable next rung.

$60-8029,000 clients

Insurance shopping leaves the top ten entirely. What remains is ownership and household: new and used vehicles, tools, BBQs, window treatments — and accounting software makes its first appearance.

$80-1008,200 clients

The established household. Auto parts 1.7x, accounting software 1.5x, home improvement, real estate, avid investors. Affinity skews collapse to mainstream — DIYers and movie fans at 1.3x. Defined by assets, not hobby.

$100+4,800 clients

The small-business owner. 63% male, 53% parents, 51% on a computer against a 40% baseline, only 15% over 65. Accounting software at 1.6x is the single strongest marker anywhere in the book.

Portraits are illustrative composites of each band's measured interest indexes, not real customers. The $60–80 and $80–100 cards each span two measured bands.

A gradient tells you which direction value lies in. It does not tell you how much value is in that direction. Multiply each band by the number of people standing in it and the strategic picture inverts.

The premium lists cannot carry volume

The program's own scale target is $20,000/day on Search at ROAS ≥ 1.05 — about $21,000/day of lead revenue. Divide that by what each pool is worth per lead and you get the leads per day it must produce; divide the pool size by that and you get how long the pool lasts. The numbers below assume every single person on the list converts, exactly once — a hard upper bound no real targeting instrument reaches.

If the bands differ 12x in value, how much of that difference can you actually buy with the targeting controls Google exposes? Compare the spread of the value itself against the spread of the traits that carry it.

If the seed lists are too small to buy from, the lookalikes built on them are the whole scaling story. So it matters a great deal what those lookalikes actually inherited — and it is not what the interest indexes suggest at first glance.

Read the lookalikes on interest and they look spectacular: Commercial Vehicles at 10.4x on Prem70 Narrow, Car Brakes 8.6x, Automotive Electronic Components 9x — indexes three times anything a seed band produced. Read them on demographics and they have fallen back down the ladder. Every trait that defined a high-EPL customer reverts: 65+ swells from +9 points back to 31%, parents drop to 39%, 18–34 roughly doubles.

Ranked by expected value, not by how interesting they are. The first three are worth more than the rest combined, and none of them is a demographic tweak.

1. 01
   Stop treating the premium bands as a media target

   Reframe $60+ from "the customer we are trying to buy" to "the seed and the yardstick." The strategic goal becomes moving the mode — shifting the $0–20 mass into $20–40 and the $20–40 mass into $40–60 — because that is where 90% of the revenue lives. Search already buys at $26.73, the top of the $20–40 band; the next rung is worth +87% per lead and holds 100,000 people.
   15.9% of bookreachable rung
2. 02
   Re-point Prem70 at the commercial angle; hold Core40 back

   Prem70's profile is trades and fleet, not consumers — pair it with the existing SMB-fleet campaign rather than the consumer offer sheet. Core40 carries a 5x auto-insurance-shopper signal, the strongest anti-signal in the book; scaling it buys the cheap band at lookalike prices. Test it against a bargain-hunter hypothesis before giving it budget.
   55%signal retained
3. 03
   Target the named in-market segments directly instead of the lookalikes

   The lookalike indexes name concrete, directly-targetable Google segments — Commercial Vehicles, Auto Parts & Accessories, Brake Service & Repair, Oil Changes — at 5–10x. Targeting them straight gives the same signal with full control, no black box, and no demographic reversion. Discount the segments that index high on every auto audience; the informative ones are those that differ between Core40 and Prem70.
   5–10xindex, direct
4. 04
   Attach the EPL lists to Search in observation mode

   Search currently carries zero audience attachment. Observation costs nothing and changes no bids, and after 30–60 days it yields real per-keyword EPL-band data — the measurement this account does not presently have, and the only thing that converts section 03 from inference into fact.
   $0cost to run
5. 05
   Seed a new lookalike from $60+ combined, not $70+

   42,000 matched members against roughly 23,000 for a $70+ seed. A bigger seed reverts less, and $60–70 still carries 3.3x the blended EPL. Given the reversion measured in section 04, seed size is the main lever available on lookalike fidelity.
   +83%bigger seed
6. 06
   Exclude 18–24 account-wide

   Efficiency 0.37, cost per conversion roughly double every other band. Currently excluded on some campaigns only; the large Demand Gen spenders carry no demographic exclusions at all. Small and certain.
   ~$390/mobad spend
7. 07
   Keep the 65+ and female exclusions only while budget-constrained

   They buy perhaps +7% on mix, and while campaigns are capped by budget rather than reach, narrowing is close to free — excluded campaigns run $26.57 cost per conversion against $27.26 unexcluded, within creative noise. The moment budget scales past what the narrowed pool fills, these become a growth ceiling and should convert to bid modifiers.
   +7%on mix, capped
8. 08
   Close the demographic measurement gap

   The account-level final URL suffix passes campaign, creative, geo, placement and click IDs to PFM, but nothing demographic — so EPL by age, gender or device can only ever be modelled the way section 03 models it. If PFM exposes a spare sub slot, {device} converts the desktop question from inference into measurement. Do not repurpose sub11; it is a live landing-angle label.
   1 slotto measure

The concentration argument is robust. Three other things in this report are not, and it is worth being precise about which is which.

The underlying data, one panel per band, exactly as Google reports it. Bars show this list against the US benchmark; index columns show the top segments Google surfaces, ten maximum.