PAGA IQ: Modeling California Wage-and-Hour Exposure Before It Becomes a Claim

Two weeks ago we wrote about what the 2024 PAGA reform changed. The short version: penalties are now a function of employer conduct, there are caps for employers who took reasonable steps, and there is a cure path that rewards fixing problems fast. All of that is good news for a PEO with California clients, with one condition. You have to be able to do the math.

Not the plaintiff's math, which is a number on a demand letter designed to start a negotiation. The actual math: for each employee, for each pay period, what was the regular rate, what should the overtime have been, which meal and rest premiums were owed and at what rate, and what penalty tier does each miss fall into under the reformed statute. That is the arithmetic PAGA IQ was built to run, and this post explains how it works and why we built it the way we did.

Start with the regular rate, because everything else depends on it

The single most common wage-and-hour error we see in California payroll is not a missed overtime hour. It is overtime paid at the wrong rate.

California requires overtime to be paid on the regular rate of pay, which is not the base hourly rate. It includes nondiscretionary bonuses, shift differentials, commissions, and most other compensation tied to work performed, averaged over the hours in the period those payments cover. A warehouse employee at $20 an hour with a $200 monthly attendance bonus does not have a $30 overtime rate; the bonus has to be folded in, and the overtime on the bonus portion has to be computed and paid, often in a later pay period once the bonus is known.

The same rate governs meal and rest period premiums. In Ferra v. Loews Hollywood Hotel, decided in 2021, the California Supreme Court held that the "regular rate of compensation" in Labor Code section 226.7 means the same thing as the regular rate of pay for overtime, so premium pay for a missed break must include nondiscretionary bonuses and similar compensation, not just the base rate. The court applied the decision retroactively.

PAGA IQ therefore starts every audit by reconstructing the regular rate per employee per pay period from the actual pay data: base earnings, each bonus or differential, the hours those payments cover, and the resulting rate. It then recomputes overtime and double-time at that rate and compares the result to what was paid. The difference is the underpayment, and it is also the first line of the penalty exposure, because an underpaid overtime hour is both a wage claim and a PAGA predicate.

Meal and rest periods: the premium and the predicate

California's meal and rest rules are familiar to anyone running payroll here: a 30-minute unpaid meal period before the end of the fifth hour, a second before the end of the tenth, and a paid 10-minute rest period for each four hours or major fraction worked. A missed, short, late, or interrupted period triggers one hour of premium pay at the regular rate of compensation, with a maximum of one meal premium and one rest premium per workday.

The audit engine looks at punch data where it is available and at the pay records where it is not. Where the client has timekeeping, PAGA IQ identifies each shift where a meal period was not started by the fifth hour or was shorter than 30 minutes, and each shift long enough to require a rest period. Where the client has only pay records, it looks for the absence of premium pay on long shifts as a flag rather than a finding, and says so in the output.

That distinction matters. An audit tool that treats every long shift without a premium as a violation will overstate exposure, and an underwriter who takes that number to a prospect will lose credibility the moment the prospect's HR manager says the employee waived the meal period in writing. The engine reports what it can establish from the data and labels what it is inferring.

Stacking the penalties under the reformed statute

Once the underlying wage calculations are done, the penalty layer applies the post-reform rules. This is where a spreadsheet becomes unmanageable, because the rules interact.

For each violation type and each employee, the engine determines the pay-period count within the one-year lookback, applies the base tier ($100 per pay period by default, $50 where the violation was isolated and nonrecurring within the statutory window, $25 for the technical wage-statement category), and halves the result if the employee is paid weekly. It then applies the no-stacking rules: no separate penalty for derivative final-pay violations, none for derivative section 204 timing violations that were not willful, and none for derivative wage-statement violations that were not knowing and intentional. Finally it applies the reasonable-steps cap, at 15 percent or 30 percent, as a scenario the user can toggle, because whether the client qualifies is a judgment call the software should not make on its own.

The output is a per-employee table and a total, with every intermediate figure visible: hours, rates, underpayment, premium owed, pay periods counted, tier applied, reductions applied. We show the math because the number is only useful if someone can defend it.

Why showing the math is the product

In two decades of selling PEO, the most effective risk conversation I ever had with a California prospect was one where we sat down with their own payroll register and worked a single employee by hand. Not a slide with an industry statistic. One person, one pay period, the bonus they got, the overtime rate they were paid, the rate they should have been paid, and the difference multiplied out across the year. The owner went quiet, then asked how many of his people had bonuses.

PAGA IQ is that conversation for every employee at once. The design rule we set was that nothing in the output could be a black box. If the engine says an employee is owed $38.20 in overtime true-up for a pay period, the user can click into it and see the base pay, the bonus, the hours, the recomputed rate, and the paid rate. If the penalty layer says a wage-statement violation falls in the $25 tier, the reason is stated. A prospect's CFO, a plaintiff's attorney, or a neutral evaluator in a cure conference can follow every step.

This is also why we did not build it as a pure AI tool. The regular rate formula is deterministic. The penalty tiers are statutory. There is no reason to ask a language model to guess at arithmetic it can get wrong, so the calculation engine is code, and the AI's job is around the edges: reading an unfamiliar payroll export, classifying earning codes into bonus versus reimbursement, and drafting the narrative explanation of what the numbers mean.

How it feeds the proposal

The reason a PEO runs this audit on a prospect is not to scare them. It is to make the proposal's risk story specific.

A generic proposal says that California wage-and-hour compliance is complicated and the PEO will handle it. A specific proposal says that in the prospect's own payroll data, overtime was computed on base rate for employees with nondiscretionary bonuses, that this affects a stated number of employees and pay periods, that the modeled exposure under the reformed statute is a stated range depending on whether reasonable-steps caps apply, and that the PEO's payroll configuration fixes the regular-rate calculation from the first pay period. The prospect's own numbers do the selling.

In HALO IQ, the PAGA IQ output flows into the Investment Analysis step of the proposal wizard as a soft-cost line, with the audit detail attached as a supporting exhibit, and the Executive Summary AI draws on it when it drafts the risk section. The Cost of Inaction analysis uses the same figures, which keeps the proposal internally consistent: the number in the executive summary is the number in the audit.

For a PEO's existing book, the same engine runs as a periodic audit, which is the reasonable step the reformed statute explicitly names. A dated audit report, with findings and the corrective action taken, is the evidence a client needs to claim the 15 percent cap. We built the audit to produce that report as a matter of course.

What it does not do

A few limits, stated plainly. PAGA IQ models exposure; it does not predict what a court will award, and it is not legal advice. It works from the data it is given, so a client with incomplete timekeeping gets flagged inferences rather than findings. It covers the wage-and-hour predicates that account for most PAGA activity, not every Labor Code provision that could theoretically support a claim. And the reasonable-steps determination is a scenario toggle, not a conclusion; whether a client qualifies is for the client and its counsel to decide with the documentation in hand.

Within those limits, it turns the hardest part of the California PEO sale, the part where the prospect does not believe the risk is real, into a table of their own numbers. That is the job.