What does a payroll provider actually do?

It calculates gross-to-net using current tax tables, pays employees, deposits federal and state payroll taxes on schedule, and files the returns — Forms 941, 940, W-2s, and state equivalents. Full-service providers do this under their own automation; what they do not do is classify your workers, verify your data inputs, or absorb your legal liability.

If my payroll provider makes a mistake, am I still liable?

Yes. The employer remains legally responsible for deposits and filings even when a provider is engaged — the IRS says so explicitly. Reputable providers reimburse penalties caused by their own errors, but errors caused by your inputs (late hours, wrong wage, wrong state) are yours, and unpaid trust fund taxes can reach responsible individuals personally regardless of who ran the software.

Is DIY payroll ever a reasonable choice?

For one or two salaried employees in a single state, paid the same amount every period, DIY with careful use of Publication 15 and EFTPS is workable. The equation flips fast with hourly staff, tips, multiple states, or benefits — each adds tables, deadlines, and failure modes that a provider handles for roughly the cost of one small penalty per year.

Payroll provider or DIY: what you're actually buying

A payroll provider sells three things: arithmetic (gross-to-net under current tax tables), logistics (paying employees and depositing taxes on schedule through EFTPS and the state portals), and filings (941s, 940, W-2s, state returns). That is a real product — the tables change every year and the deadlines never move. What it is not is an outsourcing of responsibility: the employer remains liable for its payroll taxes no matter whose software ran them.

What stays your job either way

Whether you run payroll yourself or pay a category-leading service, these never leave your desk:

Classification. Employee or contractor, exempt or non-exempt — the provider processes whatever you tell it; see /bookkeeping/payroll-compliance/employee-vs-contractor.
Inputs. Hours, rates, new hires, terminations, state of work, benefit elections. Wrong in, wrong out, your penalty.
Registrations. New-state withholding and unemployment accounts are yours to open (providers file into them but most won't create them) — see /bookkeeping/payroll-compliance/state-payroll-registrations.
The sweep-up items. Taxable fringes, S corp owner health premiums, tip declarations — someone must tell the system before year-end.
Verification. The IRS's position is blunt: employers are responsible even when using a third party. Check that deposits actually happen (EFTPS lets you view your own account) and that filings match your records — the ledger tie-outs in /bookkeeping/payroll-compliance/payroll-journal-entries are the check.
Cash. The trust-fund exposure for unremitted withholding is personal and does not transfer — /irs-controversy/trust-fund-recovery-penalty.

The risk math

DIY payroll fails in predictable places: a deposit a week late, a table not updated in January, a state return forgotten, a W-2 box miscomputed. Each failure carries a penalty tiered by lateness, and failure-to-deposit penalties reach a meaningful percentage of the deposit within days (the tiers are in Publication 15 and in /bookkeeping/payroll-compliance/penalty-notices-payroll). Provider fees for a small employer typically run in the low hundreds of dollars per month. So the comparison is not fees versus free — it is fees versus your time plus your error rate:

Illustrative annual cost: 5-employee payroll$

Illustrative figures for comparison of structure, not a market survey; DIY assumes owner time at a modest rate and one moderate deposit penalty in the year.

The structure of that chart is the whole argument: DIY wins only in the zero-error row, and the zero-error row is the one that requires reading Pub 15 for fun. Complexity multiplies the failure modes — hourly staff, tips (/bookkeeping/payroll-compliance/tips-reporting-basics), multiple states, benefits — and each one moves the equation further toward the provider.

If you do go DIY

Single state, one or two steady salaries, no benefits complexity: workable. Build the system deliberately — Pub 15 for the tables, EFTPS enrolled and tested before the first deposit is due, every deadline on the calendar (/bookkeeping/payroll-compliance/compliance-calendar-small-business), and a quarterly self-audit tying the 941 to the ledger. And revisit the choice at every hire; the setup that was fine at two employees is rarely fine at eight.