Covalynt vs Pattern Data: how they compare in 2026

The models drive the flagship product within a wider data engineering suite. ClaimScore's scoring runs on what the vendor calls a proprietary expert system AI whose criteria are weighted by neural network machine learning trained on cohorts of known valid and invalid claims, and every claim's score and rejection turns on it. The rest of the suite is described as data science and identity resolution rather than model work: ClassResolution's record reconciliation for certification, DeepValidation's contact enrichment for notice, and bespoke data engineering. It would function without the scoring models.

The product reads records, extracts exposure, injury and treatment data, scores every claimant against litigation criteria and decides eligible claims automatically. Pattern Data has sold nothing else since its founding, and it names the models it tailors to each litigation as the product.

The scoring system page explains that each claim starts at 1,000 points, loses points on more than 65 weighted criteria, is tagged with deduction codes naming each failed criterion, and is rejected below 700. Accuracy is measured through regular control studies, with data split into learning and testing sets. The ROI page shows 99.5 percent identification of valid claims and 98 percent fraud identification accuracy, but inside a worked example case rather than as a labeled result with a sample, period and method. The site also has an Accuracy Explained section.

Retrieval augmented generation over language models underpins the platform, Pattern Data's FAQ states, with case reviewers validating the findings the AI aggregates. The platform page states clear sourcing and human validation at every step and that every result can stand up to review, audit or settlement. A third party review gives a figure of 98 percent accuracy, which is not the vendor's own. No figure, test set or evaluation from the vendor is published. The product reads records and cites no primary law.

Automated eligibility decisions are described with a threshold and reasons, but no human review position is published. The scoring system marks any claim that falls below 700 points as rejected. The product page describes automated claims processing producing eligibility decisions, delivered in real time to claims systems through the API. Deduction codes give an administrator the reasons for each score, which is a real review point. Nothing published says whether a person reviews rejected claims, whether rejection is final or a recommendation to the administrator, or whether a claimant flagged in error can cure. These decisions determine who in a class is paid.

Eligible claims are decided automatically by the proprietary AI while complex or outlier cases are flagged for human review, the FAQ states, and case reviewers validate the findings the AI aggregates. On its home page, Pattern Data states that every extraction, classification and score is built to support the team's review rather than bypass it. The threshold is the eligibility and allocation logic set by the settlement criteria, recalculated when criteria change. The review surface shows validated data, record status and what each case needs next, with sourcing on every result. Flagged outliers go to a person. Nothing states a confidence level below which automatic decisions are withheld, beyond the criteria themselves.

Named deployments are published without measured outcomes from them. The case experience page lists about twenty matters with docket numbers and courts, including Lopez v. Apple, Katz-Lacabe v. Oracle, Brooks v. Thomson Reuters, In re Broiler Chicken Antitrust Litigation and Kessler v. Quaker Oats. It names some thirty federal, state and Canadian courts where the vendor has served, and describes its work as court appointed neutral and consultant. The home and product pages add further matters, including mass torts. None of these is paired with a result, and the ROI page's savings figures come from an illustrative example case, so no outcome is tied to a named matter with a method.

More than thirty litigations and 1.4 million cases sit on the platform, Pattern Data's FAQ states, alongside roles appointed by courts in major settlement programs, settlement submissions in one day rather than fifteen and review costs cut by up to five times. None is attributed to a named firm or program, and the testimonials on the home page are unnamed. The chief executive's conference biography names the 3M Combat Arms Earplug and Philips CPAP litigations as deployments, without figures.

Confidentiality rests on an audited control set, while the only published policy covering the Services permits selling or sharing personal information. The security page says the SOC 2 Type II examination covers the Confidentiality trust services criterion and describes role based access controls. The privacy policy states that it applies to the vendor's class action settlement services. It says that identifiers such as name, postal address and email, along with commercial and online activity information and inferences, will be sold or shared with advertising and analytics partners, and it permits deidentified use to improve the Services. No customer agreement setting confidentiality terms for class member data is published, and privilege and work product are not addressed.

Confidentiality measures in Pattern Data's FAQ are TLS encryption, environments that comply with HIPAA for protected health data, multifactor authentication, stringent access controls and staff training. No customer agreement is published, so nothing binds a training position. No statement is published on training use, retention, deletion, segregation by matter or docket, third party model providers, or privilege and work product.

Nothing published on the product's role relative to legal judgment. The product produces claim eligibility decisions presented as able to withstand court scrutiny, but no statement addresses who is responsible for those determinations, whether they constitute advice to counsel or the court, or what counsel and administrators must verify. The published terms govern the website only.

No advice line or supervision statement is published. The product is sold to law firms and to administrators appointed by courts. Pattern Data's home page states the AI amplifies rather than replaces human judgment. No page states that outputs are not legal advice or how the product supports a supervising lawyer's duties, and no customer agreement is published to carry such a statement.

A testing practice is described, without a governance framework or any disclosure on uneven impact. The scoring system page says accuracy is measured through regular control studies, with criteria weighted by machine learning on known valid and invalid claims and separate learning and testing sets. Nothing names an accountable owner or governance process. Nothing addresses whether fraud criteria such as shared contact attributes or name variations flag some groups of legitimate claimants more than others, which is the bias question a court or class counsel would ask.

The AI does not replace human judgment and every output supports review, Pattern Data's home page states, and the FAQ describes validation by people in the loop. No responsible AI framework, ISO 42001 or equivalent, accountable owner, results of testing before release, or statement about uneven output across litigations or record types is published.

Security practices are listed while the data terms are generic and permissive. The security page describes encryption in transit and at rest, role based access, WAF and DDoS protection, backups, disaster recovery, penetration testing, vendor risk management and an incident response framework. The privacy policy keeps personal information as long as reasonably necessary, names no subprocessors beyond categories, sets no incident notification timeline, and permits sale or sharing of personal information with advertising and analytics partners. No retention or deletion commitment for customer claim files is published.

TLS in transit, storage that complies with HIPAA for protected health data, multifactor authentication, stringent access controls and continuous staff training make up the safeguards in Pattern Data's FAQ, with adherence to SOC 2 and regular audits. No retention period, deletion commitment, subprocessor list or practice for notifying incidents is published. No customer agreement is published.

Nothing published on liability for the product. The Terms of Use govern the website and refer product use to a separate license agreement and ordering documents, which are not published; no warranty, indemnity, cap or insurance position for the scoring services is published.

No liability position is published. The site's navigation and footer carry Privacy, Security, Careers and an AI search page, and no terms of service or customer agreement. The FAQ describes pricing as tailored per firm, so the agreement is a negotiated document that is not published.

ClaimScore connects in real time through an API key to an administrator's digital claim form and sends results directly to claims or client management systems. A retro review path takes bulk claim files by secure transfer and returns result files with match and validation fields. The developer page describes API keys per case that keep each case's claim data separate, webhooks returning a score, determination and deduction codes for each claim, usage alerts, validation status updates, a mock server, and sandbox, staging and production environments. The API reference itself sits behind a login at docs.claimscore.ai. No specific claims administration or case management system is named.

The navigation carries an Integrations page and a dedicated Litify page. Pattern Data's FAQ states an open API with customizable scripts for automated actions, bulk imports from cloud storage systems and drag and drop upload, and describes integration with existing case management systems. The platform page states that settlement packets are generated in the formats claims administrators require. What syncs with Litify, and in which direction, is not described in the FAQ or on the platform page.

Cloud delivery is described without a named host, tenancy or region. The security page refers to trusted cloud partners for redundancy, and the privacy policy says the company and its sites are based in the United States for U.S. users. No hosting provider, data location or tenancy model for claim data is stated.

The platform is delivered as a cloud service. Pattern Data's FAQ states that protected health data is stored in secure environments that comply with HIPAA and that bulk import from cloud storage is supported. No hosting provider, region, residency option or tenancy model is named.

A named attestation is published with scope and a request route, without auditor or period. The security page states SOC 2 Type II examination by an independent AICPA accredited firm covering the Security, Availability and Confidentiality trust services criteria, with current reports available on request under a mutual NDA. It also lists penetration testing and vulnerability assessment. The auditor and report period are not named.

Pattern Data's FAQ states adherence to SOC 2 compliance, regular comprehensive security audits and full HIPAA compliance. It does not state that a SOC 2 report has been issued. No report type, auditor, coverage period or route to a report is stated. The footer links a Security page.

The model architecture is described as built in house, without versions or change notification. The vendor says it owns its own data and builds its own models, and describes ClaimScore as a proprietary expert system of more than 65 criteria with weights learned by neural network machine learning on cohorts of known valid and invalid claims. No third party model provider is indicated, and no model version or update cadence is published. Nothing commits to telling customers when scoring criteria or weights change during a case.

Pattern Data's FAQ states a retrieval augmented generation approach and large language models, and the litigation pages describe models tailored to specific litigations. No provider, model, inference location or commitment to notify customers of changes is named.

Cost is described only relative to value, without a rate or unit. The ROI page says the cost of ClaimScore averages under five percent of the amount saved from fraudulent claims regardless of case type, and walks through an example case, but no price, unit or pricing model is published and buying runs through a demo request.

Pattern Data's FAQ describes a transactional pricing model, with fees based on case volume and the specific services used. It names case analysis, settlement award allocation and settlement packet generation as priced services, tailored per firm to caseload and budget, and claims review fees cut by up to five times against manual review. No figure, rate card or pricing page is published.

The vendor serves settlement administrators, class counsel, defense counsel and courts across class actions, mass torts and mass arbitrations. It covers class certification data, notice, claims processing and payment, and lists consumer, privacy, antitrust, securities and pharmaceutical matters. It does not state case sizes, claim types or jurisdictions it handles less well.

Buyers are plaintiff firms, defense firms, settlement administrators and special masters. The litigation pages name Roundup, AFFF and PFAS, Camp Lejeune, Depo-Provera, hair relaxer, GLP-1, Paraquat, social media, talc and Zantac, with more than thirty litigations supported. Pattern Data states the platform is built for mass tort and multidistrict litigation. No jurisdiction outside the United States is addressed.