Designing a Consumer Arbitration Program That Survives Mass Filings: A Practical Guide

Document type: Guide Practice area: Litigation — Arbitration Jurisdiction: United States (federal) Last reviewed: 5 September 2026

An arbitration clause is not a document. It is a system that other people will operate, on their schedule, using capabilities you may not have.

Most consumer arbitration clauses were drafted to answer one question — how do we avoid class actions — and were never tested against the question that actually arrives: what happens when six thousand demands land in a week. The clauses that survive that test share a specific architecture, and the companies that survive it share a specific set of operational capabilities that have nothing to do with drafting.

This guide covers both.

PART ONE — THE THRESHOLD DECISION

Step 1: Ask whether the clause is still worth having

Do this before redesigning anything. The honest answer is not always yes.

What the arbitration clause buys:

No Rule 23 class exposure.
Confidentiality.
Usually lower per-case cost when volume is normal.
No published adverse precedent.
Faster individual resolution.

What it costs:

The company pays most institutional fees, at any volume.
Almost no dispositive motion practice.
Effectively no appellate review of an award.
No global peace. A settlement binds only signing claimants; the next campaign can begin immediately.
Mass filings reproduce aggregate exposure through a channel the company funds.

The decision framework. Estimate three numbers: your realistic class exposure, your realistic mass arbitration exposure, and the value you place on avoiding a public docket. Then ask whether your business has a systemic practice affecting a large population — auto-renewal, fees, disclosures, data practices — because that is exactly the profile mass arbitration targets, and it is exactly the profile where a court-approved class settlement is the only route to finality.

Some companies have removed their consumer arbitration clauses after running this analysis. That is a legitimate outcome, and making the decision deliberately is more important than which way it comes out.

Step 2: If you keep it, decide what you are optimizing for

Optimizing for deterrence — making arbitration burdensome enough that campaigns go elsewhere — does not work and produces unenforceable clauses. Every countermeasure struck down in the last several years was drafted with this goal.

Optimizing for orderly processing — making it possible to handle volume without paying fees disproportionate to the merits — works. Every countermeasure that survived was drafted with this goal.

The distinction is operational: does every claimant retain a real, scheduled path to an individual adjudication? If yes, the provision is likely enforceable. If no, it is likely not.

PART TWO — THE CLAUSE

Step 3: Build the informal dispute resolution provision

This is the highest-value provision, and it is worth getting exactly right.

The elements:

Individualized written notice containing the claimant's name, contact information, account identifier, a description of the dispute, and the specific relief sought.
Signature or verification by the individual claimant, not counsel alone. This is the operative requirement.
A negotiation period of thirty to sixty days.
A telephone or video conference with the individual claimant, at a mutually convenient time, if either party requests it.
Tolling of all limitations periods during the process.
A condition-precedent statement: no arbitration may be commenced until the process is complete.
Mutuality: the company must follow the same process before initiating against a customer.

Why it survives. It is genuinely useful — many disputes resolve there — it applies equally to both sides, and it does not prevent anyone from arbitrating. It merely requires an individualized step before an individualized proceeding.

The condition on which it survives. The company must actually perform. A company that auto-responds and refuses conferences will find the provision read against it, and a court reviewing an unconscionability challenge will ask how many conferences the company actually held.

Build the capability before you rely on the clause. If the provision promises a conference within thirty days, someone must be able to hold several thousand of them. Staff it, script it, log it.

Step 4: Build the mass-filing procedure

Batching:

If 25 or more demands raising substantially similar claims are
filed against the Company by the same counsel or coordinated
counsel within a 90-day period, the demands shall be
administered in batches of no more than [50], with each batch
assigned to a single arbitrator. Batches shall be initiated at
intervals of no more than [60] days until all demands are
resolved. All limitations periods are tolled for any demand
awaiting assignment to a batch. Claimants shall be assigned to
batches in the order their demands were filed.

What makes this enforceable: every claim is adjudicated; the schedule is defined and not at the company's discretion; limitations are tolled; assignment is by an objective rule rather than by selection.

What would make it unenforceable: an indefinite stay for later batches, a company right to postpone, or no tolling.

Bellwethers with an off-ramp:

The Parties may agree that [10] demands, [5] selected by
Claimants' counsel and [5] by the Company, proceed as bellwether
proceedings. Within [30] days after the last bellwether award,
the Parties shall participate in a global mediation before a
mediator jointly selected. If the mediation does not resolve the
remaining demands within [60] days, the batching procedure above
shall apply, and no bellwether award shall have any preclusive
effect on any non-bellwether claimant.

The last clause is essential. A bellwether that binds non-participants is the provision courts have squarely rejected.

Step 5: Add the supporting provisions

Small claims carve-out. Either party may bring an individual claim in small claims court instead of arbitration. Cheap, pro-consumer, reduces the fee-generating pool, and materially improves the unconscionability posture.

Meaningful opt-out. Thirty days from acceptance, by a simple written or in-app mechanism, with no adverse consequence. Honor it in practice and keep the records. Opt-out rates are consistently negligible.

Mutual fee-shifting for frivolous filings. An arbitrator may award fees against a party whose claim or defense is frivolous or brought for an improper purpose, applying a standard comparable to Rule 11. Mutual, or it is one-sided.

Claimant verification. Require that each demand include a statement that the named claimant authorized it and an account identifier. This tracks what the institutional rules now require anyway.

Class waiver with severance. All claims arbitrated individually; if the waiver is unenforceable as to any claim, that claim is severed to court and the rest proceed individually. The severance sentence prevents class claims from ending up in arbitration.

FAA fallback. If a court holds the FAA inapplicable, a named state arbitration act governs.

Consumer fairness terms. Company pays arbitrator fees above a court filing fee; full statutory remedies; no shortened limitations period; venue in the claimant's home county or by video; reasonable discovery; a written reasoned award; no gag on discussing the underlying facts.

Institution selection. Choose a provider whose mass arbitration procedure has been tested and whose rules provide individual adjudication and a role for claimants in arbitrator selection. A provider chosen because its protocol is cheap and claimant-unfriendly is a liability.

Step 6: Do not do these things

Do not amend the clause after filings begin and apply it retroactively.
Do not draft indefinite stays.
Do not let bellwether results bind absent claimants.
Do not impose costs on claimants exceeding a court filing fee. Green Tree Financial Corp.-Alabama v. Randolph, 531 U.S. 79 (2000) puts the burden on the challenger to show prohibitive costs, but a clause that plainly imposes them invites the showing.
Do not assume a delegation provision insulates an unconscionable protocol; courts have held the delegation fails with it.
Do not draft the clause without asking whether the company can perform every obligation it creates.

PART THREE — THE CAPABILITIES THAT MATTER MORE THAN THE CLAUSE

Drafting is the visible part. The companies that handle mass arbitration well do so because of four operational capabilities, none of which is a legal deliverable.

Step 7: Build the claimant-matching capability

What it is. A documented, tested process that takes a list of names, email addresses, phone numbers, and any other identifiers from a mass filing and returns, for each: whether an account exists, when it was created, what was charged and when, which terms version was accepted and on what date, which interface version the user saw, and whether the user previously contacted support about the issue.

Why it matters more than anything else in this guide. In every mass filing, a meaningful fraction of the demands do not correspond to a chargeable account. Twenty to forty percent is a common range, and it can be higher. That fraction disappears from the case once you can prove it — but only if you can prove it in weeks rather than months, because the process arbitrator's schedule will not wait.

What building it requires:

A stable identity resolution approach across email, phone, device, and account identifiers.
Access to historical billing records at the transaction level.
Terms acceptance logs tied to user identifiers and terms version.
Interface version history mapped to date ranges.
A person who owns the query and can run it on two days' notice.

Test it annually with a synthetic list. A capability that has never been exercised is a capability you do not have.

Step 8: Build the formation-evidence package

The same package needed for any motion to compel, maintained continuously.

For every version of the terms: the text, the effective dates, the version identifier, renderings of every acceptance screen as it appeared, the acceptance logging schema, the retention policy, the amendment notices and how they were delivered, and a named declarant plus a backup.

The failure mode is universal: a company produces today's registration screen for a 2022 acceptance, opposing counsel produces the archived 2022 version, and the motion to compel is denied pending formation discovery. Archive the renderings at the time of each change; they cannot be reconstructed later.

Step 9: Build the informal resolution function

If the clause promises a thirty-day negotiation and a conference, the company must be able to deliver several thousand of them.

What that requires:

An intake channel for IDR notices, separate from general support.
A triage rubric distinguishing claims by product, period, and issue.
Trained staff who can conduct a settlement conversation within defined authority.
Settlement authority delegated in advance, with tiers.
A logging system recording every notice received, every contact attempt, every conference held, and every outcome.

The log is the point. In a later unconscionability challenge, the company that can show it held 2,900 conferences and resolved 1,100 disputes is in a completely different position from the one that can show it sent 2,900 form responses.

Step 10: Build the claims-triage function

When filings arrive, someone must determine which claims are real. This is product and operations work with legal supervision.

The output is a segmentation:

Tier 1 — verified account, charged, and the alleged defect was present for that user's period and interface version. Real claims. Pay them.
Tier 2 — verified account and charged, but the defect was not present, or the user's conduct defeats the claim. Defensible.
Tier 3 — verified account, never charged, or claim outside the limitations period. Weak.
Tier 4 — unmatched, duplicate, or unverified. Contest administratively.

Paying Tier 1 promptly is not weakness. It is the cheapest resolution available, it removes the strongest claims from the campaign, and it removes the moral force from the rest. Companies that settle the whole population at a blended number overpay Tiers 3 and 4 to avoid dealing with Tier 1.

PART FOUR — THE RESPONSE PLAN

Step 11: Days 1 through 5

Obtain the claimant list in machine-readable form.
Issue a litigation hold covering interface versions, billing records, support tickets, and internal communications for the claim period. Make it broad enough to cover a class action if the clause is later held unenforceable as to any group.
Notify insurers. Several policy forms respond, and notice provisions are strict.
Stand up the response team: legal lead, data lead, product lead, support lead, finance lead.
Calendar the institution's deadlines.
Brief senior management with a range, not a number.

Step 12: Days 5 through 30

Run the claimant list. Produce the four-tier segmentation.
Map each claimant to the terms version in effect on their acceptance date. In a company with multiple versions live over the claim period, different claimants are governed by different clauses.
Assess condition-precedent compliance, honestly, in both directions.
Assemble the formation-evidence package for each relevant version.
Set a preliminary reserve.

Step 13: Days 20 through 60 — engage the process

Appear before the process arbitrator or process administrator.
Raise administrative issues with data: unverified filings, duplicates, condition-precedent noncompliance, fee allocation, and proposed sequencing.
Propose a batching schedule consistent with the clause.
Do not argue that mass arbitration is impermissible. It is permissible, the argument loses, and it costs credibility on the issues you can win.
Where the clause's IDR provision was not satisfied, seek a stay for compliance rather than dismissal. A stay achieves the same thinning with far less risk.

Step 14: Days 45 through 120 — resolve

Make prompt, fair offers on Tier 1.
Negotiate the process for the rest: batch schedule, bellwether selection, mediation timing.
Prepare the bellwether cases properly. A bellwether lost badly reprices everything.
Track resolution rates and update the reserve monthly.
Keep the settlement structure claim-by-claim; there is no class release available.

Step 15: After it ends — the debrief

What did the clause do well, and where did it fail?
What could the data capability not answer?
Which product defect generated Tier 1, and has it been fixed?
What did the IDR log look like, and would it survive scrutiny?
Update the clause prospectively, with notice and acceptance.
Schedule the next annual review.

PART FIVE — WORKED SCENARIOS

Scenario A: the program that worked

Wrenfield Streaming received 8,900 arbitration demands alleging that its free-trial-to-paid conversion violated automatic renewal statutes in four states.

Wrenfield's advantage was that its general counsel, Sanjay Okoro-Lindqvist, had run a tabletop exercise eighteen months earlier and had built three things: the claimant-matching query, the formation-evidence package, and an IDR function staffed by six people in the support organization.

Day 3. The claimant list was matched. Results: 6,200 verified with charges; 1,400 verified without charges (free trial only, never converted); 800 unmatched; 500 duplicates.

Day 9. Terms mapping showed that 5,100 claimants had accepted the 2023 terms containing the IDR provision and the batching protocol; 3,800 had accepted the 2021 terms, which had neither.

This split drove everything. The 5,100 were subject to the IDR condition precedent, which the claimants' bulk letter had not satisfied. The 3,800 were not.

Day 22. Before the process arbitrator, Wrenfield sought a stay of the 5,100 for IDR compliance and challenged the 1,300 unmatched and duplicate filings on verification grounds. It did not contest the 3,800.

Outcome of that phase. The stay was granted. Of the 5,100, about 3,300 completed individualized notices. Wrenfield's IDR function conducted 2,700 conferences over ten weeks and resolved 1,900 disputes at an average of $118, mostly by refund plus a small credit — total about $224,000.

The remaining population. 3,800 under the 2021 terms plus roughly 1,400 who completed IDR without resolution. Wrenfield's triage found that a specific conversion screen used from March to July 2022 lacked a required disclosure in two of the four states. That defect covered about 1,600 claimants.

Resolution. Wrenfield paid the 1,600 at statutory levels — about $1.1 million — and bellwethered fifteen of the remaining claims. It won eleven, lost four at modest amounts, and resolved the balance at low value.

Total: about $2.9 million, including roughly $900,000 in fees and defense costs, resolved in fourteen months.

Sanjay's assessment. The IDR provision did most of the work, and it did it because the support function could actually perform. "If we had had the clause and not the people, the provision would have been read against us."

Scenario B: the program that did not

Calloway Household Goods had an arbitration clause drafted in 2019 with a class waiver, a delegation provision, and nothing else relevant. It received 4,100 demands over a fee-disclosure issue.

What went wrong, in order.

No claimant-matching capability. It took eleven weeks to determine which demands corresponded to accounts, and the answer was incomplete because billing records for one acquired business unit were in a legacy system nobody could query.

A retroactive amendment. In week four, Calloway pushed a terms update adding a batching protocol and a bellwether provision, and asserted it applied to the pending demands. The process arbitrator rejected the application to pending claims, and claimants' counsel used the amendment in a subsequent unconscionability filing as evidence of bad faith.

No IDR provision. Nothing thinned the population.

A blended settlement. Unable to segment, Calloway settled the entire population at an average of $640, including the roughly 900 demands that did not correspond to any charged account.

Total: about $3.4 million in settlements plus $1.6 million in fees and costs, and — the part that stung — a second campaign of 2,700 demands nine months later, because the settlement bound only the signing claimants and the underlying disclosure had not been changed.

Three lessons. Segment or overpay. Never amend retroactively. And fix the product defect, because a mass arbitration settlement buys no peace at all if the practice continues.

PART SIX — BUDGET, STAFFING, AND GOVERNANCE

Budget

Item	Range
Clause redesign and terms update	$40,000–$120,000
Building the claimant-matching capability	$60,000–$250,000 (mostly engineering)
Formation-evidence package, initial build	$30,000–$100,000
IDR function, annual run rate	$150,000–$900,000 depending on volume
Mass filing response — first 90 days	$250,000–$800,000
Institutional fees, 5,000 demands, mass rules	$250,000–$1.2M
Bellwether arbitrations (10–20 cases)	$400,000–$1.5M
Settlement of meritorious tier	Highly variable

The item with the best return is the claimant-matching capability, because it reduces the population against which every other cost is multiplied.

Staffing

A legal owner who understands both the clause and the operations, and who has authority to settle within tiers.

A data owner who can run the match and defend the methodology.

A product owner who knows which interface version was live when, and can say honestly where the defects were.

An IDR lead who runs the conference function and owns the log.

A finance partner who maintains the reserve and can explain the range to the board.

Outside counsel with actual mass arbitration experience, engaged before the first filing rather than after.

Governance

Annual review, on the calendar, covering:

Is the arbitration clause still the right choice? Run the class-versus-mass comparison.
Has the institutional rule set changed?
Have the year's decisions required a clause change?
Is the formation-evidence package current for every live terms version?
When was the claimant-matching query last tested?
What does the IDR log show, and would it survive scrutiny?
What is the reserve, and what assumptions support it?

A tabletop exercise every two years. Simulate a 5,000-demand filing. Run the match. Assemble the formation package. Draft the first process arbitrator submission. The exercise finds the gaps at a cost of a few days rather than a few million dollars.

PART SEVEN — MISTAKES THAT TURN A PROBLEM INTO A CRISIS

Refusing to pay institutional fees. Courts compel arbitration under the company's own clause, and in some jurisdictions nonpayment waives arbitration entirely.

Amending the clause after filings begin. Ineffective as to pending claims and evidence of unconscionability as to everything else.

Arguing that mass arbitration is improper. It is not. The argument loses and it costs credibility on the arguments that win.

Settling the whole population at a blended number. Overpays the meritless, and it happens whenever segmentation is impossible.

Treating IDR as paperwork. A provision the company does not perform will be read against it.

Producing the current interface for a historical acceptance. The archived version will be found.

Choosing a provider for a claimant-unfriendly protocol. The Ninth Circuit has already invalidated one such arrangement, delegation clause and all.

Failing to fix the underlying practice. A settlement without a fix guarantees the next campaign.

Never testing the response plan. The first time the claimant-matching query runs should not be the week the filings arrive.

PART EIGHT — THE EMPLOYMENT AND GIG-WORKER VARIANT

Everything above applies, with four adjustments.

Claim values are larger. Wage and hour claims — unpaid overtime, meal and rest premiums, off-the-clock work, expense reimbursement, misclassification — commonly run from several thousand to tens of thousands of dollars per worker. The merits exposure is real, and the fee-leverage dynamic is secondary rather than primary. Segmentation still matters, but paying the meritorious tier is a much larger number.

The FLSA collective is opt-in. A mass arbitration campaign and an FLSA collective action are functionally similar in a way a mass arbitration and a Rule 23 class are not, which weakens the "we avoided the class device" rationale considerably.

Representative claims may not be arbitrable. Viking River Cruises, Inc. v. Moriana, 596 U.S. 639 (2022) permits an employer to compel arbitration of an individual PAGA claim notwithstanding a state rule against splitting; the Court's further suggestion that the representative claims would then fail for want of standing was rejected as a matter of state law by the California Supreme Court. Plan for a bifurcated proceeding, and address representative claims expressly in the clause.

The EFAA carves out a category. Claims of sexual assault or sexual harassment may be brought in court at the claimant's election, notwithstanding the agreement and notwithstanding any delegation provision. A campaign may therefore have a court-side component the employer cannot compel.

Drafting adjustments for worker agreements:

The IDR provision must be performable by a worker without counsel: no requirement to produce payroll records the worker does not have, no requirement to quantify damages precisely.
Fee allocation must meet the applicable state standard, which in several states exceeds the federal baseline.
Batching should account for claims large enough to warrant individual treatment; a fifty-claim batch works for $200 consumer claims and poorly for $40,000 wage claims.
Address representative claims expressly, including what happens if the individual claim is compelled and the representative claim is not.
Confirm the § 1 exemption analysis for anyone who moves goods or people, and include a state-law arbitration fallback.

And one operational point. The claimant-matching capability in the employment context is a payroll and scheduling data question rather than a billing question. The systems are different; the requirement is the same. You must be able to say, within days, whether a named claimant worked for you, when, in what role, in what jurisdiction, and under which agreement version.

PART EIGHT-B — WORKING WITH THE PROCESS ARBITRATOR

The process arbitrator (AAA) or process administrator (JAMS) is the most important decisionmaker in the first six months, and most respondents handle the relationship badly.

What the role covers. Administrative and procedural questions arising from the mass filing: whether demands satisfy the filing requirements, whether claimants have provided required information, conditions precedent, fee allocation and payment disputes, batching mechanics and sequencing, and coordination of common procedural issues. It does not cover the merits.

How to approach it.

Bring data, not adjectives. A submission stating that "many of these claims are not legitimate" accomplishes nothing. A submission attaching a declaration from the data lead explaining the matching methodology, the identifiers used, the match rate, and the specific counts in each category accomplishes a great deal.

Concede what you should. A respondent that contests everything is treated as contesting nothing. Conceding that the 4,850 verified charged claimants are properly before the institution buys credibility on the 1,550 that are not.

Ask for stays, not dismissals. Where a condition precedent was not satisfied, a stay for compliance is both more likely to be granted and more effective. Claimants who cannot or will not complete an individualized notice fall away on their own.

Propose a schedule. A respondent that arrives with a workable batching proposal, consistent with its own clause, is far more persuasive than one that only objects.

Do not relitigate the enforceability of your own clause. The process arbitrator is not the forum, and challenging the framework you drafted reads badly everywhere.

Keep the record clean. Everything filed here will be attached to a later court filing by someone. Write as though a district judge will read it, because one will.

PART NINE — FREQUENTLY ASKED QUESTIONS

We just got 4,000 demands. What do we do first? Get the claimant list in machine-readable form, issue the litigation hold, notify insurers, and start the match. Everything else follows from knowing who these people are.

Can we stop paying the institutional fees while we sort this out? No. Nonpayment closes the cases, claimants move to compel under your own clause, and courts compel. In some jurisdictions nonpayment waives arbitration and sends everything to court.

Our clause has no IDR provision. Can we add one now? Prospectively, with notice and acceptance, yes. Applied to pending or accrued claims, no — and doing it now will be used against you.

Is a batching provision enforceable? Generally yes, if every claim is adjudicated on a defined schedule with limitations tolled and assignment by an objective rule. Generally no, if later batches are stayed indefinitely.

Can bellwether outcomes bind the rest? No. That is the clearest line in this area.

How much of a typical filing population is unverifiable? Twenty to forty percent is common; higher where the claimant acquisition was purely advertising-driven. You cannot know your number without the matching capability.

Should we pay the meritorious claims quickly? Yes. It is the cheapest resolution available and it removes the strongest claims and the moral force from the campaign.

Does settling give us peace? Only with the claimants who signed. If the underlying practice is unchanged, expect another campaign.

Should we just drop the arbitration clause? Run the analysis in Step 1. For businesses with a systemic practice affecting millions of users and no realistic path to global peace through arbitration, the answer is sometimes yes.

What is the single best investment? The claimant-matching capability. It costs a few weeks of engineering and it reduces every other number in the matter.

PART TEN — A ONE-PAGE PROGRAM SUMMARY

The clause must contain: individualized IDR as a condition precedent with claimant signature and tolling; batching with a defined schedule; bellwethers with a mediation off-ramp and no preclusion; a small claims carve-out; a meaningful opt-out; mutual frivolous-filing fee-shifting; claimant verification requirements; a class waiver with severance; an FAA fallback; and consumer fairness terms.

The company must be able to: match a claimant list against its records in days; produce the formation-evidence package for every terms version; conduct thousands of informal resolution conferences and log them; segment claims into merit tiers; and settle within pre-delegated authority.

The response plan must: hold and preserve in week one; segment by week four; engage the process arbitrator with data by week six; resolve the meritorious tier by month four; and debrief with a product fix.

The governance must: review annually, tabletop biennially, and revisit whether the clause is worth having at all.

And the guiding principle: design for orderly volume, not for deterrence. Every provision that has been struck down was drafted to make arbitration unavailable. Every provision that has survived was drafted to make it manageable.

PART ELEVEN — THE CASE FOR FIXING THE PRODUCT

The final step is the one lawyers are least often asked to lead and most often should.

Mass arbitration campaigns target systemic practices. They are economically viable precisely because the same defect affects a very large number of people in a very similar way. That is also the definition of a fixable problem.

Every mass filing should produce a product review. Not a defensive one — a genuine question about whether the practice at issue is one the company wants to keep.

Is the disclosure clear to an ordinary user, tested rather than assumed?
Is the cancellation flow as easy as the sign-up flow? Several state statutes now require this expressly, and it is the single most common source of auto-renewal claims.
Are fees disclosed before the user commits, in the total price?
Does the renewal notice go out on the schedule the statute requires, through a channel the user actually reads?
Do support scripts resolve the complaint, or route around it?

The economics favor fixing it. A campaign of six thousand demands over a practice that continues will be followed by another campaign. The settlement bought nothing durable. A fix ends the claim population prospectively, and it usually costs less than the second campaign.

And it changes the posture of the current matter. A respondent that can say it identified the defect, fixed it in the current release, and is paying the affected users is in a materially better position before a process arbitrator, a mediator, and a court than one defending the practice.

Legal's role here is to make the finding legible to the business. The claims data produced by segmentation is, incidentally, some of the best product-quality information the company will ever receive: a large, self-selected sample of users describing exactly what confused or harmed them, organized by interface version and date. Companies that treat it that way get more out of a mass arbitration than they lose.

PART ELEVEN-B — WHAT A GOOD TABLETOP LOOKS LIKE

Run this every two years. It takes one day and it finds the gaps.

The scenario. A plaintiffs' firm has filed 5,000 arbitration demands with your named institution alleging a defect in a practice you actually have. Give the team a synthetic claimant list of 5,000 rows with realistic identifiers, including deliberate duplicates, non-customers, and users from a legacy acquired system.

Hour 1 — the match. Can the data team return the four-tier segmentation? How long would it take in reality? Which identifiers failed? Which systems could not be queried?

Hour 2 — the terms mapping. For each claimant, which terms version applied? Can you produce the interface rendering for that date? Who is the declarant, and are they still employed?

Hour 3 — the conditions precedent. What does the clause require? Did the hypothetical claimants comply? Did you perform your side? What does the IDR log show for the last twelve months?

Hour 4 — the process arbitrator submission. Draft the first two pages. This exercise reveals immediately whether you have the facts to support the arguments.

Hour 5 — segmentation and reserve. Which tier is Tier 1, and what would you pay? What is the range you would give the board?

Hour 6 — the debrief. List every gap. Assign an owner and a date to each.

What tabletops usually reveal: a legacy billing system nobody can query, interface renderings that were never archived, an IDR provision the support organization has never heard of, no named declarant, and no pre-delegated settlement authority. All five are cheap to fix in advance and expensive to fix under a deadline.

PART TWELVE — WHERE TO GET HELP

Outside counsel with actual mass arbitration experience. Not general arbitration experience. The process arbitrator practice, the institutional relationships, and the segmentation approach are learned by doing, and the first campaign is expensive to learn on.

Your data engineering team, before anything is filed. The claimant-matching capability is a data project with a legal use case. Framing it that way gets it built; framing it as a legal request does not.

The institution's case management team. The AAA and JAMS mass arbitration groups will explain their rules, their fee schedules, and their expectations. That conversation is more useful than reading the rules alone, and it is free.

A mediator experienced in mass claims. Global mediation after bellwethers is the standard off-ramp, and mediators who have run these processes bring structure that the parties cannot supply themselves.

Your insurance broker, early. Coverage for mass arbitration varies considerably across policy forms, and notice requirements are strict. The analysis should happen before a filing, not after.

And your product organization. The most valuable output of a mass arbitration is a list of exactly what confused several thousand users, organized by date and interface version. Someone should own turning that into a fix.