// Benefits brokers · Census cleanup
Employee census files, rebuilt in each carrier's layout with every gap listed for your account manager.
An employer sends a payroll export, an enrollment report or a spreadsheet someone typed by hand, and someone on your team turns it into the layout each carrier quotes from. This tests whether AI can draft that clean census and its list of questions, on copies of 20 to 30 censuses your team already cleaned, people replaced by look-alikes.
- Past censuses
- 20 to 30
- Real names or birth dates
- None, look-alikes only
- Who sends it to carriers
- Your account manager
// The one job
A census comes in the employer's way. A carrier-ready draft and a gap list come back.
Today an account manager or benefits analyst sorts the employer's rows into employees, spouses and children, fixes the date formats, fills in tiers and eligibility classes, and moves it all into each carrier's template. Then comes the email to the employer about the lines that don't add up.
In the proof of concept, AI does that first pass. It drafts each census in the target carrier's layout, with a list of every missing or suspect line and the reason. Nothing goes to a carrier or the employer. Your reviewer, whoever cleans these best, sets each draft beside the census your team actually sent.
and checked
// What makes a census hard
Every employer exports it one way, and every carrier reads it another
Each carrier sets its own columns. QualChoice's small group quote form for Arkansas wants one row per person, the employee first with the spouse and each child beneath, then gender, a birth date written mm/dd/yyyy, the home ZIP code and one coverage code: EO, ES, EC with the number of children, EF, or W for a waiver. It needs no names at the quoting stage, insists on its exact column fields, and warns that missing or wrong details can make the proposed rates invalid. Florida Health Care Plans' census for groups of 2 to 50 adds county and tobacco columns, splits waivers into covered elsewhere and no other coverage, and gives every employee a status: Active, COBRA, FL-COBRA, Waiting Period or Part Time.
Ages are read on a fixed date. For small group pricing, federal rules use each enrollee's age on the date the policy is issued or renewed. A birth date has to be measured against the requested effective date, then, not the day payroll ran the export, and someone whose birthday falls between the two can price at the older age. Tobacco has a narrow definition too: use on average four or more times a week within no longer than the past six months, with religious or ceremonial use excluded. Where tobacco is rated at all, the draft carries the employer's answer as given and lists every blank.
Dependents follow federal rules, and payroll rarely shows them. A plan that offers coverage for children must offer it until age 26, whether the child is married, a student, working, living elsewhere, financially independent or eligible for other coverage. Nothing obliges a plan to cover a dependent's own child. A payroll export usually shows a deduction tier, not the people behind it, so the draft lists every family tier with no dependent rows, every child 26 or older and every grandchild, for your account manager to raise. Whether anyone stays on the plan is for the employer, the plan and the carrier to settle, never the draft.
The census is wider than payroll. New Jersey's employer certification form, posted by the state's Department of Banking and Insurance, asks for everyone who works for the employer and is paid regularly, owners, partners, officers and independent contractors included, eligible or not. It also wants people no longer working who are still covered through continuation or total disability. Each line carries a status letter: F full-time, P part-time, T temporary, I independent contractor, D totally disabled, C continuee under state or federal law, U covered through a welfare arrangement set up under a collective bargaining agreement. Former employees on continuation coverage draw no pay, so a census built from a payroll export leaves them out until someone adds them back.
Counting decides which market a group is in. Moda Health's 2017 Oregon group size form counts, month by month for the prior year, the employees who worked 130 hours or more, adds the hours of everyone below that, capped at 120 per person a month and divided by 120, and averages the result over twelve months. It leaves out sole proprietors and their spouses, partners, temporary, seasonal, leased and contract workers, retirees and former employees on continuation, and treats a controlled group under section 414 of the Internal Revenue Code as one employer. Your reviewer will watch for a waiver with no reason, a hire date after the effective date and a class the employer never defined.
// What you send
Censuses your team already cleaned and sent
20 to 30 raw censuses
Files exactly as employers sent them: payroll exports, enrollment platform reports and at least one typed by hand, from small groups and larger ones.
The layouts your team filled
Each carrier's census template or quote request form, blank and as completed, so every draft is held to the same columns and codes.
What went out, and what you asked
The census your team sent to each carrier and the questions your account manager put back to the employer. Side by side, those are the answer key.
NDA first, then look-alike people
The NDA is signed before any file moves, and we never log in to payroll, an enrollment platform or a carrier portal. Names, Social Security numbers and street addresses are removed or replaced with look-alikes, and birth dates become look-alike dates that keep each age on the effective date. Health questionnaires and claims detail stay with you.
// Before it starts
Agreed before the first census moves
The owner settles four points with us in writing before day one.
- Reviewer
- Named by the owner, usually the account manager or analyst who cleans the most censuses.
- Success criteria
- Written first: which columns must match what your team sent, and which gaps the list has to catch
- Sample
- Picked by your reviewer across carriers' layouts, group sizes and the messiest files
- Price
- A range for the live month, given up front
// How the test is judged
The census your team sent, beside the AI draft
| On the same 20 to 30 censuses | Your team's census | The AI draft |
|---|---|---|
| Each person on the right row: employee, spouse or child | Scored on your past censuses | Scored on your past censuses |
| Coverage tier codes written in the carrier's own key | Scored on your past censuses | Scored on your past censuses |
| Ages measured on the requested effective date | Scored on your past censuses | Scored on your past censuses |
| Waivers, continuation members and waiting-period hires placed correctly | Scored on your past censuses | Scored on your past censuses |
| Eligibility class and hire date on every employee | Scored on your past censuses | Scored on your past censuses |
| Missing and suspect lines listed, each with its reason | Scored on your past censuses | Scored on your past censuses |
// After a yes
About 30 days on the censuses that arrive next
Once your reviewer signs off, every new census that comes in over the next month or so goes through the same job.
Your account manager checks each draft and decides what to ask the employer and which carriers receive the file. The AI never sends a census, never contacts an employer, an employee or a carrier, and never decides who is eligible or which tier anyone belongs in.
The reviewer from the test keeps scoring live drafts against the criteria you wrote down at the start.
// Later, one at a time
What could follow census cleanup
- 01Next
New group installs
The cleaned census carried into a sold group's enrollment forms and waivers, checked against the carrier's new group list.
- 02Then
Renewal analysis
Quotes that come back on the clean census set side by side for the employer, each with its conditions kept.
- 03After that
Open enrollment change files
Elections made at open enrollment turned into each carrier's add, drop and tier-change layout for your team to send.
// One hard limit
A licensed person decides who is eligible and sends every census.
The proof of concept works only on copies and drafts a census with a list of questions. It never requests a quote, never sends a file, never contacts an employer, employee or carrier, and never decides who is eligible or which tier they belong in. Names and identifiers are removed or replaced with look-alikes, and nothing on this page is legal, tax or benefits advice.
Priced per job: a range before the proof of concept, a fixed price before you sign.
You hear the range on the call, once we know how many censuses your team cleans a year and how many carrier layouts each one goes into.
// Questions
What account managers ask about census cleanup
Does it need real names and birth dates to work?
Our employers send censuses in every format. Is that a problem?
Will it decide who is eligible or which class someone is in?
Does it handle dental, life and disability censuses as well?
Will it send the census to carriers or back to the employer?

Rob Eason, founder
// Book a proof of concept call
Bring the messiest census you cleaned this year.
A 30-minute call with Rob Eason. The booking form asks what you do at the agency, which job you want tested and about how many censuses cross your desk each year. Hold every census file until the NDA is signed.
- Co-owned a 60-employee financial services firm and sold his share to his partner.
- Ran GOSO Technologies, a Boston solar company rated 5/5 across 24 Google reviews.
- 5/5 across 83 reviews in HighLevel's Certified Admin Directory.
- 4.3/5 from 6 Google reviews of Massmarket AI.
- Every directory and Google review, critical ones included, is shown word for word at massmarket.ai/reviews.
- Takes every proof-of-concept call himself.
