Class Certification Toolkit: Damages Models, Daubert Motions, and Certification Records

Document type: Toolkit Practice area: Litigation — Class Actions Jurisdiction: United States Last reviewed: 5 September 2026

Tool 1 — Class definition

The rules:

Objective criteria only. Membership determinable without deciding the merits.
No fail-safe. A definition that excludes anyone who loses on the merits is rejected.
Ascertainable. In circuits requiring it, an administratively feasible method of identification.
Bounded in time, product, channel, and geography.
Exclusions stated.
Match the definition to the evidence — this is the whole game.

Template:

All persons and entities that purchased [Product], in [container/grade/model variants listed], directly from [Defendant] or its authorized distributors, in [territory], during the period [start date] through [end date] (the "Class Period").

Excluded from the Class are: (a) Defendant, its officers, directors, employees, subsidiaries, and affiliates; (b) federal, state, and local governmental entities; (c) purchasers under [specified negotiated contracts / the specified pricing program]; (d) purchasers through [channel] where [the challenged representation was not present]; (e) any person who purchased solely for resale in the form purchased; and (f) the presiding judge, court staff, and their immediate families.

Fail-safe examples to avoid:

"All persons who were misled by the label" ✗
"All purchasers who paid an inflated price" ✗
"All employees who were discriminated against" ✗

Corrected:

"All persons who purchased [Product] bearing the statement '[X]' during the Class Period" ✓

Subclass template:

Subclass A (Contract Purchasers): Class members who purchased under a written supply agreement of more than twelve months' duration. Subclass B (Spot Purchasers): All other Class members. Each subclass is represented by [named plaintiff] and by [separate counsel where interests diverge].

Tool 2 — Certification motion outline

Organize as a proof, not an argument.

I. Introduction — one page: the conduct, the class, the common answer, the model, the relief.

II. Factual background — organized around uniformity: the policy, the label, the pricing mechanism, the conduct. Every assertion cited to evidence.

III. The class definition — objective criteria, exclusions, and why each exclusion was made.

IV. Rule 23(a) A. Numerosity — the number, from data. B. Commonality — the common answer, stated as a question the jury will answer once for everyone (Wal-Mart). C. Typicality — the representatives' claims arise from the same conduct and the same legal theory; disclose and address any differences. D. Adequacy — representatives' understanding and commitment; counsel's qualifications; absence of conflicts, including intra-class.

V. Rule 23(b)(3) — Predominance A. The elements of each claim. B. Element-by-element: what evidence proves it, and whether that evidence is common. C. The mechanism of class-wide injury, in one sentence, then supported. D. The damages model: methodology, data, results, fit to the pleaded theory (Comcast), and member-level results including the number showing no injury. E. Individualized issues identified, and why they are subordinate.

VI. Rule 23(b)(3) — Superiority — including the trial plan (Tool 9).

VII. Appointment of class counsel under Rule 23(g).

Appendices: expert report; class definition with exclusions; trial plan; draft verdict form; state-law variation chart if applicable.

Note. Address the uninjured-member argument in the opening brief. Waiting for the reply concedes that it took the plaintiff by surprise.

Tool 3 — Expert engagement scope

Purpose. To determine whether [injury / impact / damages] can be established across the proposed class using evidence common to the class, and if so to develop and apply a methodology for doing so.

Phase 1 — Feasibility (before the model is built). Review available data; assess whether the data supports a common-proof methodology; identify sources of heterogeneity; and advise on class definition boundaries the data will support. Deliverable: oral findings only.

Phase 2 — Model development. Construct the methodology; document the specification; run sensitivities; run the false positive test on a no-conduct period; produce member-level results; and document all code and intermediate data.

Phase 3 — Report and testimony.

Requirements throughout.

All work reproducible from code by a third party.

Member-level output available on request.

Every discarded specification retained and documented.

Notify counsel immediately of any result that does not support the engagement's premise. The expert is not an advocate and must not be treated as one.

Retention. Code, data, and workpapers retained until final resolution.

Tool 4 — Model specification memorandum

Prepared by the expert, reviewed by counsel, before the report is served.

1. What the model measures, in one sentence a non-economist would understand. 2. The liability theory it is tied to, quoted from the complaint. 3. Data. Source, period, observations, unit of analysis, exclusions applied and why. 4. Dependent variable. Definition and construction. 5. Explanatory variables. Each one listed with the reason for inclusion. 6. Variables considered and excluded, with reasons. (Anticipates the omitted-variable attack.) 7. Functional form, and why. 8. Benchmark or comparison period. Definition, and the evidence that it is uncontaminated. 9. Standard errors and clustering. 10. Results. Point estimate, standard error, significance. 11. Sensitivity table. Results under at least four alternative defensible specifications. 12. False positive test. Model applied to [no-conduct period]; result; interpretation. 13. Disaggregated results. By product, customer type, channel, and sub-period. 14. Member-level results. Distribution; number and percentage with estimates at or below zero; explanation. 15. Limitations, stated candidly.

Why this exists. Every item on this list is something the other side will ask about. An expert who has written it down before serving the report answers from a document; one who has not, improvises in a deposition.

Tool 5 — Member-level results table

The single most consequential exhibit in modern certification practice. Prepare it whichever side you are on.

Segment	Members	Mean estimate	Median	% with estimate > 0	% ≤ 0	% statistically significant > 0
All
Product A
Product B
Contract purchasers
Spot purchasers
Period 1
Period 2

For plaintiffs: produce it, explain the negative estimates as estimation error around a common effect rather than as evidence of no injury, and offer the narrowed class if the numbers require it.

For defendants: produce it from the plaintiff's own model and code, so it cannot be dismissed as your expert's construct.

Tool 6 — Test protocols

False positive test. Apply the plaintiff's model, unmodified, to a period during which no conduct is alleged, preferably before the alleged conduct began and after any acquisition or structural change. Report: estimated "overcharge," significance, and comparison with the class-period estimate. A model that finds an overcharge when none is alleged has not been shown to measure the conduct.

Disaggregation test. Re-estimate separately for each product, customer type, channel, and sub-period with sufficient observations. Report the estimate and significance for each. Segments with no significant effect indicate that the aggregate masks members with no injury.

Specification sensitivity. Re-estimate adding and removing each control variable individually, and under at least two alternative functional forms. Report the range. A result that swings materially under defensible alternatives is fragile.

Benchmark contamination test. Identify, from the plaintiff's own documents and the defendant's records, any conduct during the benchmark period. Re-estimate excluding contaminated sub-periods. Report the change.

Replication. Run the plaintiff's code on the plaintiff's data and confirm the reported results are reproduced exactly. Report any discrepancy.

Tool 7 — Data production protocol

1. Scope. Transaction-level records for [products] for [period], including: transaction date; customer identifier; ship-to and bill-to; product identifier and description; quantity; unit of measure; gross price; discounts, rebates, and allowances by type; net price; freight; contract identifier; and channel. 2. Format. Delimited text or database export, one row per transaction line, with a defined delimiter and text qualifier. UTF-8. No spreadsheet row limits. 3. Data dictionary. Required with the first production: field name, data type, definition, permissible values, units, treatment of nulls, and any coding changes over the period. 4. Completeness certification. A declaration identifying the systems queried, the query logic, and any known gaps. 5. Technical contacts. Each party designates one technical contact authorized to resolve format and content questions directly. 6. Validation window. The receiving party has [30] days to identify anomalies; the producing party responds within [14] days. 7. Anomaly log. Maintained jointly, with each item's status. 8. Supplementation. Rolling productions permitted; each accompanied by a description of what changed. 9. Confidentiality. Attorneys'-eyes-only; expert acknowledgments processed within [10] days of submission. 10. Preservation. Both parties preserve source extracts and any transformations applied.

Tool 8 — Rule 702 motion outline

I. Introduction. The model is the plaintiff's only evidence that injury can be shown class-wide. It is unreliable. Without it, predominance fails.

II. Legal standard. Rule 702 as amended in 2023: the proponent must establish admissibility by a preponderance, and the opinion must reflect a reliable application of the methodology to the facts. Daubert gatekeeping; Kumho Tire extension to all expert testimony; Joiner on the analytical gap. Daubert applies at certification where the evidence is critical to the Rule 23 decision.

III. The model does not measure what the theory requires. Comcast. [The model measures X; the certified theory is Y.]

IV. The methodology is unreliable on its own terms. A. False positive: applied to a no-conduct period, the model finds an "overcharge" of ___%. B. The result depends on [specification choice], which is unsupported and untested. C. [Benchmark contamination / aggregation masking / analytical gap].

V. The methodology was not reliably applied. A. Data errors: [specifics]. B. The benchmark period includes conduct the plaintiff itself alleges. C. [Population mismatch / survey design defect].

VI. The opinion cannot be replicated. [Code produced does not reproduce the reported results / code not produced.]

VII. These are admissibility, not weight. The defect is in the methodology and its application, not in the choice between two reliable approaches.

VIII. Relief. Exclude the opinion; deny certification.

Tool 9 — Trial plan and verdict form

TRIAL PLAN 1. Claims tried on a class basis: [list]. 2. Elements and common proof: table — element | evidence | witness or exhibit | common? 3. Order of proof and estimated length: [days]. 4. Individualized issues and their management: [bifurcation / claims process / special master / sub-trials]. 5. Defendant's defences and how they will be presented — addressing the "trial by formula" objection directly. 6. Damages: aggregate determination method; allocation plan; distribution mechanism. 7. Verdict form: attached.

DRAFT VERDICT FORM

Do you find that Plaintiffs proved, by a preponderance of the evidence, that Defendant [engaged in the conduct]? □ Yes □ No (If No, stop.)

Do you find that Plaintiffs proved that the conduct [caused the class-wide effect]? □ Yes □ No (If No, stop.)

Do you find that Plaintiffs proved that each member of the Class [suffered injury of the type alleged]? □ Yes □ No

State the aggregate damages sustained by the Class: $________

Why the verdict form matters. It is the clearest possible test of whether the case is triable on common evidence. If question 3 cannot honestly be answered yes or no for the whole class on the evidence proposed, the case is not certifiable — and both sides learn that faster from a draft verdict form than from fifty pages of briefing.

Tool 10 — Findings request

Submit with the reply or at the hearing. Both sides benefit.

Plaintiffs [Defendant] respectfully request that the Court make findings on the following disputed questions, which are material to the Rule 23 analysis and to any review under Rule 23(f):

Whether [Expert]'s methodology is a generally accepted method for [estimating but-for prices / measuring price premium / measuring price impact].

Whether the benchmark period used is free of the conduct alleged.

What proportion of proposed class members show [no injury / an estimate at or below zero] under the model.

Whether the model measures damages attributable to the theory on which certification is sought.

Whether the results are robust to [the specifications identified].

Whether the model, applied to [the no-conduct period], produces a spurious result.

Whether individualized inquiry would be required to identify uninjured members, and if so its scope.

Why both sides want findings. A certification order that recites the parties' positions without resolving them is reviewed deferentially and tells neither party where it stands.

Tool 11 — Rule 23(f) petition structure

Fourteen days. Calendar it the day the order issues. In most circuits a reconsideration motion does not extend it.

I. Question presented — framed as a legal question, in one sentence. II. Relief sought and jurisdiction — Rule 23(f); date of the order; date of the petition. III. Statement — the claims, the class, the order, in three pages. IV. Reasons to grant. A. The order turns on an unsettled question of law important to class action practice — [the circuit split / the absence of controlling authority]. B. The order is effectively dispositive. [Denial ends the case as no individual claim is economically viable / Grant creates settlement pressure the defendant cannot resist, so the legal question will otherwise escape review.] C. The order is manifestly erroneous. [The court applied the wrong standard / failed to resolve a factual dispute it was required to resolve / accepted expert evidence without conducting the required analysis.] V. Conclusion.

Framing note. Abuse-of-discretion review is deferential. Frame the petition as legal error — wrong standard, failure to make required findings, misapplication of Comcast or Wal-Mart — rather than as disagreement with how the evidence was weighed.

Tool 12 — State-law variation chart

Required for any multistate class asserting state-law claims.

Element	Majority rule	Variations	States	Material?
Reliance required
Scienter standard
Damages measure
Limitations period
Discovery rule
Pre-suit notice
Privity requirement
Available remedies

For plaintiffs: group states into subclasses by legal standard, and be prepared to show that the groupings are workable at trial. For defendants: identify the variations that are outcome-determinative and show that they cannot be managed by grouping.

Tool 13 — Settlement class approval checklist

Rule 23 satisfied independently of settlement (Amchem) — manageability aside
Intra-class conflicts identified: present vs future claimants, injury types, strong vs weak claims
Structural protections where conflicts exist: subclasses with separate counsel
Class definition drafted with litigation-class discipline
Allocation plan treats similarly situated members similarly, with a stated rationale
Recovery assessed against the strength of the claims, candidly
Realistic claims take-up rate estimated, and the fee measured against actual benefit conferred
Clear sailing and reversion provisions justified or removed
Notice plan by a notice expert, with a projected reach percentage
Plain-language notice; functioning website and telephone line
Opt-out procedure and deadline; blow-up provision considered
Objector process, and a plan for handling serial objectors
Cy pres, if any, justified and the recipient's relationship to the class explained
Service awards for representatives supported by evidence of actual service
Release scope matched to the claims released — no broader
Preliminary approval papers candid about the case's weaknesses