On this page
The formula fits on a sticky note. Revenue potential equals potential units times price per unit.
The hard part is everything around it. Which unit? Where do you find it for thousands of companies you have never spoken to? How do you know the number isn't nonsense? And what do you do when the data for half your accounts is missing?
If you're still deciding whether you need this number at all, start with our answer on what revenue potential is and why it beats ranking people by clicks. This page is the how-to.
What is the formula for revenue potential?
Revenue potential equals the number of units an account could buy times your price per unit. The unit is whatever you actually sell or charge for: seats, sites, machines, lanes, employees covered. The price is your typical price per unit for an account that size, not your list price.
At Chaos, a design software company, the unit was seats. Revenue potential was potential seats times EUR 550. That one field ranked every account in the database by what it could be worth, and it changed how the team saw its own customer base.
The unit changes by business:
- Software: seats or users, sometimes locations.
- Manufacturing: plants, production lines or annual volume of the parts you make.
- Logistics: shipping volume, lanes or facilities.
- Professional services: employees, entities or engagements a year.
When you're not sure, look at your ten biggest customers and ask what made them big. That's almost always your unit.
How do you estimate potential units for accounts you don't know yet?
Use facts you can find for every account and turn them into units with a simple rule. Company size, the roles that use your product, the number of locations or the industry all work. Check the rule against customers where you already know the real number, then apply it everywhere.
At Chaos, potential seats came from company size and the kinds of roles that use their software. A company with a large design team gets more potential seats than one the same size with a small one. You don't need perfect inputs. You need inputs that rank accounts the same way your sales team would.
Three places the inputs come from:
- Data vendors in a waterfall. One vendor after another until one finds the value. Clay's waterfall enrichment works exactly this way.
- AI research on the company's own site. Careers pages, locations pages and product pages often say more than any database. Esko used AI agents to scrape company sites for custom industry tags and revenue potential, because the standard data was weak on global coverage and on their industries.
- Your own CRM and product data. Existing customers, trials and past deals tell you what a real account looks like.
Stage it company first. A media monitoring company we worked with sized its market at about 92,000 accounts, 31,686 already in the CRM and about 60,400 net-new. We enriched companies before contacts, because there's no point paying for contacts at accounts that won't make the cut.
Revenue potential, one account at a time
How do you know if the estimate can be trusted?
Give every enriched field a confidence score, only sync fields above your bar, and run a sample before the full database. Then have the people who know your customers check a few hundred records by hand. If they agree with the ranking, it's ready.
Here's how we did it on Chaos:
- Confidence on every field. Each enriched value got a score from 1 to 10. Only fields above the bar synced to the CRM.
- A record-level check. A separate score looked at the whole record and blocked junk records from getting in at all.
- A sample first. We ran a 1,000-record sample and hit an 80% or better match before running the full database.
- Batches after that. About 174,000 records ran in batches of 50,000, so problems showed up early and cheaply.
- A human spot check. The client checked 200 to 250 records by hand, because the people who know the customers should sign off.
The sample also told the board something they didn't expect. They believed the customer base was about 90% architecture, engineering and construction. The data showed it was closer to 50%. That's the kind of finding a revenue potential project often produces before it produces a single number.
If cost is the worry, our answer on what a Clay waterfall enrichment project costs covers how to keep the data spend predictable.
How do you test revenue potential against your best customers?
Pull your best customers and see where the model ranks them. If they land near the top, the model works. If big customers rank low, the unit or the inputs are wrong, so fix those before anyone uses the number to plan.
This test catches the most expensive mistakes. At Chaos, a signup with one seat and a Gmail address looked like the smallest lead in the database. It was Parsons Corporation, with room for 407 seats. A model based on who signed up would have ignored it. A model based on what the company could buy put it near the top.
A few checks to run:
- Top customers rank high. Most of your biggest accounts should sit in your top tier.
- Personal emails get resolved. At Chaos, 37% of accounts had a personal email and no company attached. Matching those to real companies changed the ranking.
- Old lists get re-run. Elastic Path re-ran 4,887 accounts from an old ICP list through a two-signal check and found 543 worth targeting.
Checks that tell you the number can be trusted
| Check | How | Example |
|---|---|---|
| Sample first | Run a small batch before the full database | Chaos: 1,000 records, 80% or better match |
| Confidence per field | Score each value, sync only above the bar | Chaos: scores from 1 to 10 |
| Human spot check | People who know the customers review the ranking | Chaos: 200 to 250 records |
| Best customers rank high | Look up your biggest accounts in the model | One-seat Gmail signup: Parsons, room for 407 seats |
| Re-run old lists | Put past target lists through the same model | Elastic Path: 4,887 accounts, 543 worth targeting |
How do you use revenue potential once it's calculated?
Write it to the CRM as a field on every account, set tiers from it, and use it to shape territories, decide which accounts get a person and which get automation, and find expansion. Then refresh it on a schedule so new data and new accounts keep it current.
Where it shows up:
- Tiers that update themselves. Tier one, two and three by revenue potential, recalculated as data changes.
- Fair territories. Balance total potential, not just account counts. Our answer on designing sales territories that scale shows how.
- The human lane and the AI lane. High-potential accounts get a rep. The long tail gets an agent that hands off when an account replies or shows a strong signal. We cover where to draw that line in human reps versus AI agents.
- Expansion. For customers, the gap between what they buy now and their potential is your expansion list. Chaos used this to target trials with potential for three or more seats.
- Reactivation. Old deals and lost accounts with high potential are worth another look. See how to run a closed-lost reactivation campaign.
Esko's plan shows the full shape: about 25,000 net-new accounts, each with revenue potential and a product-fit tag, plus contacts with 90% or better valid emails, scoped at 30 to 45 days. That's a list reps can start on day one.
Where revenue potential goes once it's calculated
Inputs
- Company size and roles
- Locations and sites
- Industry tags
- Your customers and trials
Revenue potential
- Units times price
- Confidence on every field
- Tier from the value
Company first, then contacts
Where it is used
- Tiers that update themselves
- Fair territories
- Human lane or AI lane
- Expansion and reactivation lists
Frequently asked questions
What if our price per unit varies a lot?
Use the price a typical account of that size actually pays, not your list price. If prices differ by segment, use one price per segment. The goal is a number that ranks accounts correctly, not an invoice. Most teams start with one blended price, test the ranking on their best customers, and only add segments if the ranking looks wrong.
How often should revenue potential be recalculated?
Recalculate when the inputs change: new accounts, new enrichment, or a change in your pricing. Most teams refresh new and changed records continuously and re-run the full database a few times a year. Only enrich records that changed, so you're not paying to recalculate numbers that stayed the same.
Do we need perfect data to start?
No. Start with a proxy you can find for most accounts, like employee count or number of locations, and test it against your best customers. If the proxy ranks them near the top, it's good enough to use. Score confidence on every field, so reps can see which numbers are solid and which are estimates.
Should revenue potential be a dollar figure or a tier?
Store both. The number lets you add up territories, compare segments and size your market from the bottom up. The tier is what reps actually use day to day, because "tier one" is easier to act on than a figure. Keep the number in the CRM even if most people only ever look at the tier.
Sources
- Chaos, waterfall enrichment and revenue potential, 2025 to 2026: revenue potential as potential seats times EUR 550, confidence scores from 1 to 10, a record-level check that blocks junk records, the 1,000-record sample at 80% or better match, about 174,000 records in batches of 50,000, 200 to 250 records checked by the client, 37% of accounts with personal emails and no company, the base at about 50% AEC against an assumed 90%, the one-seat Gmail signup that was Parsons Corporation with room for 407 seats, and trials with potential for three or more seats. See the Clay enterprise case study.
- Esko data projects, August and September 2026: AI agents scraping company sites for custom industry tags and revenue potential, and about 25,000 net-new accounts with a product-fit tag and 90% or better valid emails in 30 to 45 days.
- A media monitoring company's territory design, May 2026: about 92,000 accounts in the market, 31,686 in the CRM and about 60,400 net-new, enriched company first.
- Elastic Path: 4,887 old ICP accounts re-run through a two-signal check, leaving 543 worth targeting.
- Clay, waterfall enrichment and Actions and Data Credits documentation: how a waterfall checks providers in order, and how enrichment is charged.
- RevOps Shop, Clay for RevOps: how we build revenue potential and tiers in your CRM.