of the 787,189 active carriers FMCSA regulated ran a single power unit in 2023
Federal Motor Carrier Safety Administration, 2025Why do databases mix real buyers with look-alikes?
Because an industry code records how a company was classified, and that says little about whether it runs the operation your product serves.
NAICS, the North American industry classification, puts each establishment in the industry of its primary activity.↗ In practice that activity is often judged by proxies such as revenue, shipments or employment.↗ When the US Census Bureau has no survey answer, it falls back on other records, including codes that businesses assign themselves on tax returns.↗ Where it knows only part of a code, it fills in the rest from an establishment with similar payroll and employment.↗
A prospect database’s industry field is a label of the same kind, applied by the data vendor’s own method. A company that runs a network of warehouses and one that only books trucks can carry the same label. A search on that label returns both, and each look-alike costs a seller research time before anyone sees it can’t buy.
So treat the code as a place to start looking, and define fit by what an account does. The same gap opens in any market where your product serves a single operation inside a broad category.
Define fit as operating evidence, then check it at the source
A fit definition works when every criterion is a fact about how the account operates, checkable at a named source. Most definitions need 4 kinds: the operation you serve, who the account runs it for, its size in your product’s unit and what it runs today. The unit depends on what you sell: plants or lines for manufacturing software, branches for a banking platform, warehouses for warehouse management software (WMS).
| Criterion | What counts as evidence | Where to check it | Use |
|---|---|---|---|
| The operation you serve | The account runs it today, at sites you can name | Its own locations and services pages, registers and directories | Gate |
| Who it runs it for | Itself, its customers or a parent company | Services pages, case studies and client lists | Gate |
| Size in your product’s unit | Sites, lines or users, counted individually | Distinct addresses and job posts that name sites | Band |
| What it runs today | The current system and how long it has run | Job posts, integration pages and case studies | Score |
| Each look-alike | The fact that rules it out | Registers of what a company is licensed to do | Exclusion |
The first 2 are gates. An account that fails either is out, however well it scores on anything else. Write down the look-alikes as well, each with the fact that rules it out, so a researcher can reject an account in a line.
Then build the candidate list wide and let the evidence decide. Pull candidates on several codes and keywords. Check each one where the evidence lives: the company’s own locations and services pages, public registers, association directories and job posts that name sites or systems. Buy contact data only for accounts that pass.
Count the size unit site by site: distinct addresses that do the work, without offices or partners’ sites. Record each fit field with its evidence link, the date it was checked and a confidence: verified, inferred or unverified. An account with an unverified count goes to a research queue, never into a tier. Give that queue an owner and a weekly slot, or it turns into a second list nobody reads.
Test the definition on your won and lost deals of the last 24 months before you commit sellers to it.↗ Our audit looks for segments whose win rate is at least 1.5× the average, with 10 or more closed deals behind it.↗ Our fit scores start with 4–6 features, because more of them overfit the test.↗
Size Tier 1 and 2 to the panels your sellers can map
Tier by fit first and capacity second: Tier 1 and 2 hold only as many verified fits as your sellers can research, map and multi-thread. Today’s AI answers to this question tier by score thresholds or by share of the list, such as a top 10–15%.↗ They don’t ask whether anyone can work those accounts.
Our working ranges per account executive are 5–15 accounts in Tier 1 (1:1), 20–60 in Tier 2 (1:few) and 100–500 in Tier 3 (1:many).↗ Within the verified fits, rank by fit score and deal potential, and fill Tier 1 from the top.
Each tier also has a role map of 3–5 required seats per account, so a tier’s size is really a count of seats to reach.↗ That count has to stay small enough to multi-thread, and multi-threading pays: Gong found it lifted win rates by an average of 130% on deals over $50k.↗ Tier 3 gets 1:many plays, and an account moves up when a signal fires.
On a named list, tiers also settle the ABM-or-inbound question. Say you sell a transport management system (TMS) to 400 named shippers on 9-month cycles: the market is finite, so the frame is account-based. Tier 1 and 2 get plays for named seats, and Tier 3 gets content and ads aimed at the named list.
Inbound content still earns its place. In 6sense’s 2025 survey of nearly 4,000 buyers, 79% made the first contact themselves, and the vendor they contacted first won about 80% of the time.↗ Record each account’s busy window as dates on the account, such as a freight RFP season or a year-end close, and book first meetings around it.
Store the fit definition in fields that scoring and routing read
Give every fit criterion its own CRM field, with its evidence link and the date it was checked. Calculate the tier from those fields instead of typing it.
- Fit status: qualified, excluded or research, from a fixed list
- Exclusion reason: which look-alike, from a fixed list
- Size count as a number, with the band it falls in
- Evidence link and the date each fit field was checked
- Confidence for each field: verified, inferred or unverified
- Tier, calculated from the fields, with the date tiered and the owner
- Busy window: the dates the account’s buying seats are hard to reach
- Definition version the account was tiered under
Then make every system read the same fields. The gates run first: a failed gate sets the fit status to excluded, and the score never runs on that account.
Routing reads the tier and the owner. A form fill from a Tier 1 account goes to its owner the same day. One from an excluded account goes to nurture, so look-alikes create no work for sellers.
The definition has a single owner and a version number, so a change to a band or a gate is a decision with a date. Sales and marketing agree it once, in writing, and the fields are where that agreement lives. Re-tier every quarter by a written rule.↗ An AI pass may propose moves with reasons, and the list owner approves them.↗
Measure the list against a baseline. Our audit scores the share of new pipeline that comes from Tier 1–2 accounts.↗ Under 20% is the bottom of its scale, and over 75% is the top.↗ Take that share before you rebuild the list, and report it every quarter after.
How do you build the list when the database mixes 3PLs with brokers and carriers?
Gate on warehouse evidence before anything else: in US logistics, the look-alikes are counted in hundreds of thousands. In 2023, 418,526 of the 787,189 active carriers FMCSA regulated ran a single power unit, and 28,351 property brokers were registered.↗ Any logistics database inherits that mix.
More than half of the carriers FMCSA regulated in 2023 ran a single power unit
Data behind this chart
| Item | Value |
|---|---|
| 1 power unit | 418526 |
| 2 power units | 123353 |
| 3–10 power units | 168507 |
| 11–100 power units | 56715 |
| Over 100 power units | 5062 |
| None or unreported | 15026 |
Part of the screen is free. FMCSA’s public company snapshot shows a registered company’s entity type, such as carrier or broker, with its operating authority and power units.↗ You can drop carriers with a single truck before anyone opens their website.
Broker authority alone shouldn’t exclude an account, since some warehouse operators broker freight too. A missing warehouse should.
Take a fictional WMS vendor with 3 account executives (Illustrative).↗ It exports 2,400 US companies that its data provider labels third-party logistics.↗ The warehouse gate looks for sites on the company’s locations page, in association directories or in job posts.
It rejects 1,530: 1,050 carriers and 480 brokers or forwarders with no warehouse.↗ Another 90 go to research, and 780 pass.↗ The third-party gate drops 170 that store only their own goods, leaving 610.↗ Counting distinct warehouse addresses, without offices and cross-docks, puts 214 in the 5–30 band, 352 below it, 26 above it and 18 unverified.↗
So about 9% of the export fits.↗ At 10 Tier 1 and 40 Tier 2 accounts per seller, the vendor tiers 30 and 120 of the 214.↗ The other 64 go to Tier 3, and the 352 smaller operators wait for a new-site signal.↗
At a 3PL, by our estimate, the managing director or COO signs, with finance, IT and the operations lead in the room.↗ Those 4 seats per Tier 1 account make 120 seats, or 40 per seller.↗ Calling all 214 accounts Tier 1 would mean 856 seats, about 285 per seller.↗
A shorter list of verified fits produces more pipeline
We think a shorter list of verified fits produces more pipeline than a long, loose one. The first reason is seller attention, which we see as the scarcest input in account-based work. Every look-alike in Tier 1 takes research, mapping and touches that a real buyer needed. In our view, a seller who meets a few look-alikes in a row also starts to trust the list less.
The second is the score. A fit score ranks whatever it is fed, so gating the look-alikes out first lets it compare real buyers with each other. On a loose list, much of what it ranks is noise.
The third is automation, which repeats whatever the list says. Routing, sequences and ads all read the tier, so a tier that rests on a label spreads that label’s mistakes faster. That’s why we clean the fields an automation reads before it goes live.
We don’t think there is a single right list size. It depends on how many sellers you have and what a deal is worth. So we size tiers to capacity rather than to a share of the list.
In practice
How we do it at Panelhop
In the Panel Check (GTM audit · 2–3 weeks), 4 of the 32 checks cover targeting: whether a named list lives in the CRM, whether fit comes from won and lost deals, whether the tiers match the team’s capacity and how much new pipeline Tier 1–2 accounts produce. Where they fall short, Leak Fix (we build the fixes) builds the list from the client’s own data providers and public sources. It writes the tier, fit score, date tiered and owner on each account, and leaves a written rule for the quarterly re-tier.
Signal Desk (in-market accounts, weekly) then watches the tiered accounts and puts the ones in market into the CRM with a brief the rep can act on the same day. Every tool stays in the client’s name.
Questions buyers ask about this
We sell a TMS to about 400 named shippers on 9-month cycles. ABM or inbound demand gen?
Account-based, because 400 named shippers is a finite list: plays for named seats in Tier 1 and 2, and inbound content and ads aimed at the named list for the rest. Record each shipper’s RFP season on the account, say September–November. A 9-month cycle that starts in December then ends as the next season begins, while one that starts in February ends inside it.
Can we still use industry codes?
Yes, as a place to start looking. Pull candidates wide on several codes and keywords, then let the operating evidence decide. A code that agrees with the evidence is fine, but a code alone qualifies nothing.
What if an account runs the operation we serve and a look-alike business too?
Judge it on the operation you serve: if it runs enough of it, it passes the gates whatever else it does. Record its other lines of business in a business-model field, so plays and messages can use them.
How often should we re-check the fit fields?
Re-tier every quarter by a written rule, and re-check an account’s evidence before it moves into Tier 1. Re-check sooner when a signal says something changed, such as a new site or an acquisition.
How many accounts should Tier 1 hold?
As many as your sellers can map and work, which is why we don’t give a single number. Our working range is 5–15 per account executive for 1:1 work. Count the required seats as well, because those are what your sellers actually have to reach.
