How to Build an Administrative Record That Holds Up in Court

What actually goes in the record

The standard comes from the Administrative Procedure Act. Courts require what's called the "whole record" and this includes every document that might have influenced the decision, including materials that were reasonably available to staff involved in the process even if nobody specifically reviewed them before signing the final document ¹. The EPA's 2011 administrative records guidance reinforces this ¹⁷.

There's also an important distinction between the "whole administrative record" and the "certified administrative record." The whole record is everything used by the agency during the decision process. The certified record is the subset the agency actually hands to the court. If something from the whole record doesn't make it into the certified version it's treated as omitted record evidence and not extra record evidence ³. That distinction matters because extra record evidence is almost impossible to get admitted while omitted record evidence just needs to be identified.

One thing agencies consistently get wrong is leaving out materials that cut against their position. These are things like internal disagreements, unfavorable data and memos where someone raised a concern that didn't make it into the final document. Courts expect all of that to be there and it is important to know agencies can't clean up their record to make the narrative tidier. Wilson and Brammer wrote about this in the Public Land and Resources Law Review in 2024 and multiple circuit courts have held that sanitizing the record is grounds for remand ²¹.

The contemporaneous problem

A record has to reflect what was before the agency at the time it made the decision and not something that was made up after the record was challenged.

The DOI's 2006 guidance recommends agencies maintain a "decision file" as the project progresses ². This is a running repository of the documents, data and correspondence that show how the decision was made. If they build the record while the work is happening they won't end up scrambling to reconstruct it later, which is what happens most of the time. They're focused on the analysis itself and the record is all over the place until it is challenged.

When records get compiled after being challenged, courts look at them with suspicion. Reports or affidavits prepared specifically for litigation are generally treated as post hoc rationalizations ⁸. They don't count as part of the whole record. If the existing record is so thin that it can't explain the agency's reasoning a court might allow testimony from officials, but that's treated as an exception. The EPA's guidance is to build the record as the action develops and not after the record is challenged ¹⁷.

What the record actually needs to contain

Technical reports and modeling data are the biggest parts but for complex reviews, whether it's a NEPA EIS, a Section 404 alternatives analysis, a FERC licensing proceeding or a CERCLA remedial investigation, agencies need to archive modeling inputs and outputs at every stage not just the final results ¹⁸. This includes the intermediate runs, the calibration data, the assumptions that changed between draft and final. Courts have found that factual modeling information like acoustic impact models or air dispersion modeling is generally not protected by deliberative privilege. It's considered objective data and the court has a right to see it.

Correspondence is just as important and gets lost more often this includes emails where someone explains why one methodology was chosen over another, meeting minutes from interagency consultations, memorialized phone calls with USFWS or NMFS about species findings, notes from SHPO coordination under Section 106. All of this needs to be captured. When an agency changes its position on something, whether it's a determination of effect under the ESA or a consistency finding under the CZMA, the correspondence showing why that shift happened is one of the most critical material in the record.

Rejected alternatives need to be documented with enough detail to show the agency actually considered them ⁴. This applies across permitting programs. A NEPA EIS requires an alternatives comparison matrix. A Section 404 permit requires a practicable alternatives analysis. A BLM right of way application requires plan consistency findings. In each case the record needs to show not just what was chosen but what was considered and why it wasn't. A thin alternatives analysis is one of the most reliable ways to get a decision overturned.

Privilege and what it doesn't cover

Internal drafts, preliminary opinions and deliberative memos are complicated as the deliberative process privilege protects these materials from disclosure in many circumstances. But the privilege doesn't cover factual information. If a deliberative memo contains data, measurements or survey results that information is considered "reasonably segregable" and has to be produced ¹⁷. The EPA recommends pulling factual data out of deliberative drafts and putting it into standalone technical support documents specifically to avoid this problem.

Communications with outside contractors add another layer. When an agency hires a consulting firm to prepare a biological assessment or run air quality modeling those communications may be protected under what's called the "consultant corollary". It treats the contractor as if they were internal agency staff for privilege purposes. However the factual work product, the data, the model outputs, the survey results, none of that is protected regardless of who produced it.

Some circuits require agencies to produce a privilege log identifying every document withheld on deliberative grounds. Others, including the Ninth and D.C. circuits, have traditionally not required one. But the trend is moving toward requiring logs ¹¹. Without them there's no way for a court to tell whether the privilege was properly applied or whether the agency is holding back material it shouldn't be.

Breaking the presumption

When an agency submits its certified record the court gives it a "presumption of regularity", meaning the court assumes the agency compiled the record properly ¹⁰. This is an advantage for agencies and it's also why sloppy records are so damaging. Once a plaintiff shows that specific documents were before the decision makers and aren't in the certified record that presumption starts falling apart.

The standard for breaking it is concrete, non-speculative evidence. A plaintiff can't just argue that documents probably exist somewhere in the agency's files. They need to point to specific materials. But the bar for showing bad faith is lower than agencies assume. Evidence that a decision was pretextual, that the agency ignored its own experts, or that the outcome was decided before the analysis was done can all open the door to extra record discovery ⁸. The Center for Biological Diversity v. Burgum order and the BMBP v. Jeffries decision in the Ninth Circuit both dealt with these questions recently ^5,19.

The mechanical failures nobody talks about

Some of the most preventable record problems are physical and digital. These include color coded GIS maps that get scanned in black and white and become unreadable, version control breaks down and the record contains a draft that was superseded three revisions ago, website content and social media posts that documented public engagement efforts disappear because nobody archived the pages with dates and URLs.

The AASHTO Practitioner's Handbook and the FTA's 2019 guidance both recommend labels with dates, authors and version numbers on every document ^16,9. Final memos and notes should be converted to non editable PDFs. Dynamic content needs to be captured with timestamps when it's posted, not when challenged ²⁰. Although these are dull operational details, they're also the reason records fall apart under scrutiny.