Employer research
A broker could narrow the search by territory, industry, recorded size, benefits records and workers’ comp records. Each measure showed its coverage limits.
Employer research.
Self-service diagnostics.
A product experiment.
Working prototype and portfolio demonstration. Not offered as a commercial product.
I built PEO Tools to help PEO and insurance brokers work out which employers to investigate and what their numbers showed before the first call. The research database helped find the businesses. Four live diagnostics let an owner work through a specific question before speaking with a broker.
Nick Angeli, Founder / Solo BuilderLinkedIn
Employer research
- Territory
- Industry
- Recorded size
- Benefits records
- Workers’ comp records
- 20.7 millionBusinesses in the research database
- 2.84 millionWith a recorded size indicator
- 110,740With a health-plan record
Database snapshot: October 11, 2026. Coverage and record dates vary. Recorded size may represent employees, plan participants or an individual worksite.
What was built.
The database covered 20.7 million US businesses. That gave a broker a place to start. The next part was helping the owner work through their own figures, so the first call could start with findings and questions already in hand.
Four working diagnostics
The checks worked through workers’ comp audit exposure, health and auto renewals, and the federal 50-employee count. Calculations ran in tested code. Missing information stayed visible.
From research to a case
I designed the product for delivery under the broker’s own brand, with the findings routed back to that broker. The product explanation followed the same sequence: find the employer, work through the figures, then discuss the case.
Two problems, one product.
A broker needs to know which employers are worth investigating and what a prospect’s numbers actually show. Those were the two problems PEO Tools was built around.
Employer research
An employer might fit the territory and industry a broker worked in. Its recorded size, benefits and insurance records gave the broker more to investigate. A missing record stayed a gap, not a reason to assume a fit.
Self-service diagnostics
Finding the business didn’t tell the broker what its costs or renewal figures showed. The diagnostics let the owner supply those figures and see what needed a closer look.
Designed around the broker
The owner’s answers and the calculated findings belonged in the same case. The broker could then see what was known, what was missing and what to ask next. Advice and placement remained the broker’s responsibility.
Finding the employer was only the start.
An employer record could help a broker decide who to approach. It couldn’t explain what that employer’s own figures showed. That still needed answers from the owner.
The diagnostic collected those answers and ran the calculations before the call. The broker could spend the conversation on the findings and the gaps that remained.
| Research on its own | PEO Tools |
|---|---|
| Records help identify an employer worth investigating | The owner adds their own figures through a diagnostic |
| The broker has the research record | Designed to bring the owner’s answers and findings back to the broker |
| A starting point for discovery | Research, a diagnostic, then a conversation about the findings |
| The broker still needs to collect the owner’s figures | The owner works through guided questions before the call |
The call could start with what the numbers showed and what still needed checking.
How the workflow was designed.
Each step answered the next question: which employer, which check, what figures, what findings, and what the broker still needed to ask.
- Find the employersThe broker filtered by territory, industry, recorded size, benefits records and workers’ comp records to build a shortlist.
- Share a diagnosticThe broker shared a check suited to the question: workers’ comp audit, group health renewal, commercial auto or the federal 50-employee count.
- The owner answersThe business owner worked through the questions in their own time.
- The software calculatesDefined, tested calculations ran on those answers. Missing information stayed visible.
- The broker receives the caseThe broker received the owner’s answers, the calculated findings and the information still missing.
- The first call starts informedThe broker could start with what the numbers showed and which gaps still needed attention.
What the market showed.
The brokers I built this for earned placement and monthly service fees. The diagnostics worked and produced prepared cases, but those brokers didn’t have budgets for a recurring software subscription. Once that was clear, I closed the commercial experiment. The software, database and design remain here as a portfolio demonstration.
How I built it.
My background is in sales, automation and commercial software development. For this product, I started with the work a broker needed to do: find relevant employers, understand their figures and have a useful first conversation.
I used Claude Code in development. I also experimented with Claude to extract information from documents and prepare case summaries. Calculations stayed in tested code, separate from those experiments. If information was missing, the software showed the gap.
The calculations were deterministic: the same inputs produced the same results. That made the findings testable and kept the distinction clear between what the owner supplied, what the software calculated and what was still unknown.
Live artifacts.
These are the four published checks. You can open each one and work through the questions to see how the software uses the answers.
- Workers’ comp audit check
Works through workers’ comp audit exposure and shows which information is still needed.
- Group health renewal check
Works through the renewal figures and shows where the comparison is incomplete.
- Commercial auto renewal check
Works through commercial auto renewal figures and the questions they raise.
- Federal 50-employee count
Works through the federal 50-employee count using the employer’s own inputs.
A view into the employer market
The snapshot shows how much information was available and where coverage was limited. The charts below keep those limits alongside the findings.
- 20.7 millionbusinesses in the research database.
- 37,998businesses with a recorded workers’ comp policy end date in the next twelve months. Policy end dates are recorded for a small number of states.
Database snapshot: October 11, 2026. Coverage and record dates vary.
Six market charts
Each chart answers a different question. The definitions explain who was counted, which records were available and what the figures can tell you.
Where are employer businesses concentrated?
Territory planning
Useful when deciding where to focus outreach.
Businesses with a recorded size of 10 to 249
New York-Newark-Jersey City leads with 66,477 businesses, and the ten largest metro areas hold 29% of the 1,086,307 businesses in this population.
About these figures
- Population
- Businesses with a recorded size of 10 to 249 in the 50 states and DC, 1,086,307 in total. The ten metro areas with the most businesses are shown.
- Coverage
- Recorded size may represent employees, plan participants or an individual worksite, and many businesses have no recorded size. Businesses outside a metro area are not shown.
- Snapshot
- October 11, 2026.
When were these businesses registered?
Business age
Helps separate newer businesses from established ones when building a list.
Active businesses with a recorded registration date
2025 is the largest registration year among active businesses with a recorded date, at 868,703; the counts show businesses active today, not registrations filed each year. The hatched bar is the current year to date.
About these figures
- Population
- Active businesses with a recorded registration date: 6,673,534, or 32% of the database. Business applications, registrations and operating employers are different populations, and this chart shows only the last.
- Coverage
- Registration dates exist for a limited set of states, and the two largest account for 94% of the dated businesses. Earlier years are smaller partly because more of those businesses have closed.
- Snapshot
- October 11, 2026. The current year is partial.
How has the number of employers with plans changed?
Market sizing
Shows how many employers already sponsor a plan you may be asked to compare against.
Employers with a recorded plan, by plan year
Health plans
Retirement plans
Employers with a recorded health plan rose from 49,863 in 2012 to 87,711 in 2024, and employers with a recorded retirement plan from 649,743 to 835,396. Hatched bars are plan years still filling in.
About these figures
- Population
- Distinct employers with at least one recorded health plan, and separately with at least one recorded retirement plan, in each plan year. An employer with both appears in both charts, and each chart has its own scale.
- Coverage
- Health plans with fewer than 100 participants are mostly not recorded, so the health series understates smaller employers.
- Snapshot
- October 11, 2026. Plan year 2025 is incomplete because records are still arriving.
How are recorded health plans funded?
Plan funding
Shows how often the conversation starts from an insured or a self-funded plan.
Share of recorded health plans, by plan year
- Self-funded
- Insured
- Insurance and own assets
- Not reported
Among recorded health plans that report a funding method, self-funded plans were 9% in 2024, down from 19% in 2012, while 30% of 2024 plans report no method at all.
About these figures
- Population
- Recorded health plans in each plan year. These are plans, not employers, and an employer can have more than one.
- Coverage
- A growing share of plans report no funding method. They are shown as not reported and are not assigned to a category.
- Snapshot
- October 11, 2026. Plan year 2025 is incomplete.
Which industries are adding jobs?
Industry changes
Points to the sectors where headcount, and the employer costs that follow it, are moving.
Market-wide employment change, 2025 versus 2024
- Health care and social assistance+780K
- Construction+116K
- Arts, entertainment, recreation+62K
- Transportation and warehousing+56K
- Education services+42K
- Information-54K
- Wholesale trade-69K
- Retail trade-88K
- Administrative and support-92K
- Manufacturing-153K
In 2025 versus 2024, health care and social assistance added the most jobs, about 780 thousand, and manufacturing lost the most, about 153 thousand.
About these figures
- Population
- Private-sector employment across the United States, by industry sector. This is a market-wide measure, not a measure of the businesses in the research database, and an industry adding jobs does not mean any one business is hiring.
- Period
- Annual average employment, 2025 versus 2024. The five largest gains and five largest declines among the industry sectors are shown.
- Snapshot
- October 11, 2026.
When do businesses’ recorded workers’ comp policies end?
Outreach timing
Shows when recorded policies come up, so outreach can be timed once the dates are confirmed.
Businesses with a recorded policy end date, by month
1.4K2.5K2.7K5.4K3.0K3.4K3.8K3.1K3.5K3.9K2.8K2.5K
37,998 businesses have a recorded workers’ comp policy end date in the next twelve months, and January 2027 is the busiest month, with 5,443 businesses. The hatched bar is the current month from the snapshot date.
About these figures
- Population
- Businesses with a recorded workers’ compensation policy end date: 40,227 in all, and 37,998 with an end date in the next twelve months. Each business is counted once, by its recorded end date. An end date does not establish an intention to switch.
- Coverage
- Policy end dates are recorded for a small number of states, and the two largest account for 94% of the records. This is not a national picture.
- Snapshot
- October 11, 2026.