The Restrictive Clause Score, to Sign, Narrow, or Refuse

You have an offer packet, and stapled into it are four restrictive terms: a non-compete, an arbitration agreement, an invention-assignment clause, and a non-solicit. You want to sign the job, not a lawsuit, and you have maybe one real negotiation push before you look difficult. This guide gives you a scoring model to triage all four clauses on the same axes, so you can decide clause by clause which to sign as-is, which to push to narrow, and which to refuse or opt out of, and spend that one push where it actually moves risk.

This is written for a candidate without a lawyer in the room, making a judgement call under a signing deadline. It is not legal advice, and where a dollar figure or a verdict turns on your exact state and salary, I say so and tell you what to check locally.

Why score instead of redline everything?

Score because you cannot fight four clauses at once without burning the goodwill that got you the offer. A scoring model tells you which single clause is worth risking the offer over, and which three you can sign or opt out of with no drama.

Most offer guides do one of three things. A reference defines what each clause binds you to. A teardown walks one clause's redline end to end. A sign-ready standard tells you what a clean packet looks like. None of them handle the real situation, which is four restrictive terms arriving together, each with a different enforceability, a different career cost, and a different path to relief. You need a way to compare them on one ruler.

The ruler has five dimensions. Each one answers a different question, and each one tells you something when it lies.

Score each clause on each axis, and the clause with the highest combination of career cost and enforceability is the one that earns your push. The rest sort themselves into sign-as-is or no-risk opt-out.

Score enforceability first so you never spend your one push fighting a clause that was already dead.

How enforceability is graded where you live

Grade a clause void, weak, or likely-enforceable based on your state's law and your salary, and grade this axis first, because a void clause never deserves a fight. Enforceability is the cheapest axis to score and it eliminates the most work.

Start with the bans. Sources disagree on the exact count, but the states with near-total non-compete bans are commonly listed as California, Minnesota, Montana, North Dakota, Oklahoma, and Wyoming. Wyoming's took effect July 1, 2025 and is narrower, voiding non-competes except for trade secrets, sale of a business, and executive and management personnel. Minnesota's ban runs under Minn. Stat. section 181.988, effective July 1, 2023. If your governing law is a ban state and you are an ordinary employee, the non-compete scores void.

Above the bans, twelve states void non-competes below a salary floor: Colorado, D.C., Illinois, Maine, Maryland, Massachusetts, Nevada, New Hampshire, Oregon, Rhode Island, Virginia, and Washington. Here are the reported 2025 thresholds for a selection of them.

Thresholds like these create a cliff, not a slope. One dollar over the Illinois $75,000 floor flips you from "clause void" to "the reasonableness test governs," so a modest raise in the offer can quietly re-arm a non-compete you assumed was dead. Do not treat a threshold as a shield once you are over it. Above the line, courts weigh duration, geography, and scope, and most states treat restrictions beyond two years as presumptively unreasonable, with the common range running six months to two years and one year being the most typical.

There are also industry carve-outs that change the grade. Nevada prohibits non-competes for employees paid solely by the hour. Virginia, beginning July 1, 2025, bans them for all overtime-eligible employees. Maryland, effective June 1, 2024, prohibits them for direct-care healthcare providers earning $350,000 or less. And note that the FTC's 2024 nationwide ban was blocked by a federal court and never took effect, so there is no federal floor to rely on; this is all state-by-state.

What each clause actually costs your career

Career cost is whether the clause blocks the specific next move you would realistically make in the next one to two years, scored high, medium, or low. A clause that bars moves you would never make is low cost no matter how aggressive it reads.

For a non-compete or non-solicit, ask three questions. Does it name or capture the companies you would actually jump to? Does it cover your metro or just a region you would not relocate within anyway? Does it restrict your specific product area or an entire industry you could sidestep? A 12-month non-compete limited to your exact product area is a very different cost from a 24-month blanket restriction.

Geography is where most people misjudge cost, because they score it against nationwide headcount instead of the pool where they would actually look. The depth of your home market changes everything.

Read it this way: a 50-mile non-compete in a deep metro leaves thousands of non-restricted employers within reach, so its career cost is low. The same clause in a thin market can wall you out of your field. In Refolk's index of professional profiles, the gap between a deep and a thin market for the same title runs more than fifteenfold, so score geography against the local pool you would draw from, not against a national number. You can size that pool for your own role and metro with Refolk before you decide how hard the clause really bites.

For invention assignment, the career cost is usually not about blocking a job. It is about whether side projects, open-source work, or a future startup idea get swept into the employer's ownership. That cost is latent and shows up years later, which is why it is easy to under-score.

Arbitration: a calendar problem, not a legal one

Arbitration is the one clause where you often get a no-cost exit, but only if you act inside a short window. The dominant pattern is a 30-day written opt-out, and silence inside that window counts as acceptance.

The typical language gives you written notice to HR within 30 calendar days of signing, and it usually states that opting out will not affect your employment status or any other term. That protection is the point: opting out of arbitration is low-risk, so for most candidates the verdict is "opt out, in writing, with proof." The reason people end up bound is not that they weighed it and chose arbitration. It is that they never read the packet and the 30-day clock ran out. The opt-out is designed to lapse.

Two rights survive signing no matter what, so do not trade your opt-out away believing it is your only protection. The Ending Forced Arbitration of Sexual Assault and Sexual Harassment Act, signed March 3, 2022, invalidates pre-dispute arbitration agreements for sexual assault and sexual harassment claims arising after that date. And the Sarbanes-Oxley whistleblower provision, at 18 U.S.C. 1514A(e)(2), makes those disputes non-arbitrable. EFAA's reach is broad: one court held an entire case of eight different claims non-arbitrable once a covered claim was present. That makes the harassment carve-out more valuable than its narrow label suggests.

The concession ladder for a push to narrow

When you decide to push, map the ask to a documented concession rather than inventing one, because asks with precedent get granted and vague ones get refused. Four levers recur across the evidence: duration, geography, competitor definition, and pay.

A 24-month nationwide non-compete can be negotiated down to six months within your immediate market, or eliminated entirely in exchange for stronger confidentiality. Garden-leave pay is both a documented ask and, in Massachusetts, a statutory default: under the Noncompetition Agreement Act, effective October 1, 2018, the employer must pay no less than 50% of your highest annualized base salary over the two preceding years during the restricted period, and the Act caps most non-competes at 12 months. Another clean carve-out to request: the restriction only applies if the company terminates you without cause.

For invention assignment, the single highest-value redline is one word. "I hereby assign" is a present transfer; "I agree to assign" is only a future promise that can be contested years later. Change the future tense to the present, and attach a prior-inventions exhibit listing any work you are excluding. Around ten states cap forced assignment on the California model: Labor Code sections 2870 to 2872 void clauses that force assignment of own-time inventions and require the employer to give written notice, and New York's Labor Law section 203-f, effective 2023, mirrors it. But the own-time carve-out only holds if the invention does not relate to the employer's business or demonstrably anticipated research, so the tense fix and the exhibit remove more certain risk than arguing the business-relatedness language.

Thank you for the offer, which I am excited to accept. I have two small requests on the restrictive terms so I can sign with confidence.

1. Non-compete: the current term is [24 months, nationwide]. I would like to narrow it to [12 months] within [my metro / my specific product area], limited to [named competitors]. If a broader term is important to you, I would ask that it apply only if the company terminates me without cause.

2. Invention assignment: please change "I agree to assign" to "I hereby assign," and attach the standard prior-inventions exhibit so I can list pre-existing work I am excluding.

Separately, I am exercising the arbitration opt-out under Section [X] by this written notice, within the [30]-day window. Please confirm receipt.

The scoring procedure, start to countersign

Run the clauses through these steps in order. Scoring enforceability before drafting asks is deliberate: employment-law sources front-load "check enforceability first," while negotiation sources front-load "narrow the scope," and starting with enforceability stops you from spending a push on an already-void clause.

How this goes wrong

The scoring model fails in predictable ways, and each failure is a false positive: a clause you think you have handled that is still live. Treat this section as the most important part of the guide.

• "It's void, so ignore it." A non-compete void in your state can still trigger deferred-comp forfeiture. The forfeiture is a contractual self-help remedy that is almost always upheld, because the court never reaches enforceability. Senior packages can defer 30 to 70 percent of comp, so the money at stake is real even in a ban state. Score any separate equity or clawback clause on its own axis.
• Missing the opt-out clock. A verbal "I declined arbitration" is not an opt-out. Employers have won on the argument that there was no evidence the employee returned the notice inside the 30-day window. Send it in writing, dated, with proof of delivery.
• Assuming the invention carve-out is automatic. The own-time carve-out only holds if the invention does not relate to the employer's business or demonstrably anticipated research. If it relates to anticipated R&D, California does not protect it, and you are back to needing the tense fix and the exhibit.
• Reading "agree to assign" as harmless. It is a future promise that can defeat your ownership later. Verify present versus future tense before you sign.
• Treating a threshold as a shield when you are over it. The floor only voids the clause below it. Above the line, the reasonableness test governs, so a high earner gets no automatic out.
• Confusing the arbitration opt-out with EFAA-type rights. Sexual-harassment and SOX whistleblower claims are non-waivable regardless, so do not trade the opt-out away thinking it is your only protection.
• Over-narrowing the wrong clause. Spending your one push on a low-career-cost clause wastes it. Re-check that the clause you are fighting actually blocks a realistic next move before you risk the offer over it.

If you want a real-world read on how freely restricted people actually move, Refolk can pull live examples from its index rather than leaving you to guess.

Before you countersign

Run this checklist against the final packet. Every item is something you can verify, not a topic to think about.

Keeping the model current

The scoring axes are stable, but the numbers that feed enforceability change by statute and effective date, so re-check your inputs each time you use this. Wage thresholds are reset periodically; Wyoming's ban took effect July 1, 2025 and Virginia's overtime-eligible ban the same day, which tells you new rules land on fixed dates you can look up. Before you rely on a threshold or a ban, confirm your state's current figure and the effective date against the statute itself, and check whether an industry carve-out like Nevada's hourly-worker rule or Maryland's healthcare rule applies to you.

For anything where real money or a career-defining move is on the line, the verdict from this model tells you whether to involve counsel and what to ask them. The model's job is triage: it narrows four clauses to the one that matters, maps your asks to documented concessions, and keeps you from signing away a right you could have kept for free. That is enough to walk into the negotiation knowing exactly which clause you are willing to risk the offer over, and which three you are not.