Qualified, Adverse, or Disclaimed: How to Report a Study That Did Not Go to Plan

Answer first: allocating rigor, controlling your process and selecting your sample all assume the same thing - that the plan survives contact with fieldwork. It frequently does not. The segment fails to recruit, the event logging turns out to have been broken for the period you cared about, legal blocks access to the accounts that would have answered the question. Auditing faced this so often that it built a graded, named ladder of report outcomes with explicit triggers for each rung. Research has one move: a bullet in the limitations section that nobody reads. That asymmetry is the subject of this article.

The short answer

When an auditor cannot issue a clean opinion, the standards do not offer a free-text box. They offer three named outcomes, collectively called a modified opinion. ISA 705 defines it as "a qualified opinion, an adverse opinion or a disclaimer of opinion."

Which one you must issue is determined by a two-by-two, not by judgment about tone.

Two inputs. Is the problem big enough to matter, and does it contaminate everything or only one part? Answer both and the report type is decided for you.

Pervasive is a defined term, and the definition is the useful part

Most of the power here comes from the fact that "pervasive" is not left to instinct. ISA 705 defines pervasive effects as those that, in the auditor's judgment:

1. "Are not confined to specific elements, accounts or items of the financial report"
2. "If so confined, represent or could represent a substantial proportion of the financial report"
3. "In relation to disclosures, are fundamental to users' understanding of the financial report"

Strip the accounting nouns and you have a test that works directly on a research read-out:

• Not confined. Does the problem touch every finding, or one of them? A moderator who mis-asked one question has a confined problem. A screener that let the wrong population into the whole study does not.
• A substantial proportion. Even if confined, does it cover most of what the study was for? Losing one of six questions is confined and small. Losing the only question the decision actually depended on is confined and substantial.
• Fundamental to understanding. Would a reader who did not know about this problem misread the whole document? If yes, it is pervasive no matter how narrow it looks.

That third clause is the one that catches teams out. Pervasiveness is about the reader's comprehension, not about the size of the defect.

The ladder, translated

The rung that does the most work in practice is the emphasis of matter, and it is the one people misunderstand. An emphasis of matter paragraph is not a modification of the opinion. The work is sound and the finding is correct; the auditor is simply insisting that a particular correctly-disclosed matter gets read. In research terms: the study is fine, and the fact that it was fielded the week of your competitor's launch is not a flaw, but any reader who misses that context will misread every number. That is an emphasis of matter, and it belongs at the top of the document, not in an appendix.

Scope limitations, and the rung nobody wants to use

AS 3105.05 describes the trigger precisely: restrictions on scope, "whether imposed by the client or by circumstances, such as the timing of his or her work, the inability to obtain sufficient appropriate evidential matter, or an inadequacy in the accounting records, may require the auditor to qualify his or her opinion or to disclaim an opinion."

Note the two sources. Circumstance-imposed limitations are the ones research teams recognize: the panel underdelivered, the logging was broken, the holiday period killed response rates. Client-imposed limitations are the ones research teams experience constantly and never name - the stakeholder who will not let you talk to the churned accounts, the legal team that blocks the region, the executive who wants the study but not the enterprise interviews.

Auditing takes a notably hard line on the second. AS 3105 states that when restrictions that significantly limit the scope of the audit are imposed by the client, ordinarily the auditor should disclaim an opinion.

That is a strong rule and it exists for a structural reason: if the party commissioning the work can narrow the scope and still receive a confident conclusion, the conclusion is worth nothing, and the incentive to narrow scope becomes irresistible. The research analogue is exact. A stakeholder who blocks access to the segment that would disconfirm the plan, and then receives a report that quietly hedges in the limitations section, has learned that blocking access is free.

The behavioral fix is not more rigor. It is that the consequence becomes visible and named. "We are disclaiming on the enterprise question because we were not given access to enterprise accounts" is a sentence that appears in the summary, gets read by everyone in the room, and changes what happens next time. The same fact as a limitations bullet changes nothing.

What this is not

It is not a sample-size question. Materiality here is a reporting judgment - is this defect big enough to change what a reader should conclude. It is not a question about how much evidence is sufficient, which is a different problem with its own guidance in how many interviews are enough and survey sample size. Keeping these apart matters, because collapsing them produces the bad habit of treating every under-recruited study as automatically unreportable.

It is not the same as declaring an assurance level up front. Assurance levels are prospective: you commit before fielding to how strong a claim the study will be entitled to make. This ladder is retrospective: it is what you do when the study you committed to did not happen. A study can be honestly limited-assurance by design and separately require a qualification because the SMB cell never filled.

It is not a way to avoid publishing. Rungs 1 through 3 are all reports. A disclaimer is a document that states what was attempted, what blocked it, and what would be needed - which is far more actionable than silence, and considerably more actionable than a confident conclusion drawn from a broken study.

How the three siblings fail into this ladder

Each of the preceding problems has a natural landing rung, which is a useful way to rehearse the judgment.

• A risk allocation that was wrong - you accepted high detection risk on a question whose control risk turned out to be high too. See the research risk model. Usually an emphasis of matter or a qualification, because the study is sound but narrower than the decision needs.
• A control that did not operate - the guide was not followed for one segment, so probing depth is not comparable. See process controls vs output checks. Qualified if confined to that segment; adverse if the comparison between segments was the entire point.
• A selection you cannot project - the review pulled the vivid cases and someone computed a rate from them. See choosing what to review. Qualified as to the rate, unmodified as to the existence claims, which is exactly why the two buckets are reported separately.

Writing the modification so it survives the deck

Three mechanics make the difference between a real modification and a decorative one.

Put it above the fold. Auditing requires the basis for modification to appear immediately, in its own section, before the detail. A limitation that appears on slide 34 has not been disclosed; it has been archived.

Name the rung. "Qualified" is a word that makes a reader stop. "Some limitations apply" is a phrase that makes a reader continue. Use a fixed vocabulary and use it consistently so the terms accumulate meaning across studies.

State the remedy and its cost. Every modification should end with what would remove it: which segment, how many participants, what access. This converts an excuse into a decision, and it puts the cost of the limitation in front of the person who caused it.

Does the ladder actually work?

Worth being honest: a vocabulary is not a cure. The PCAOB reported that across all inspected firms, the Part I.A deficiency rate - engagements where the auditor had not obtained sufficient appropriate evidence to support the opinion it had already signed - was 39 percent in 2024, down from 46 percent in 2023. Among the Big Four it was 20 percent, down from 26 percent.

So a profession with a formal ladder, statutory independence, an external inspector and personal liability still signs a substantial fraction of opinions without adequate support. The ladder is not what makes people honest. What it does is remove the excuse that there was no way to say it - and it makes the gap measurable, which is the precondition for the improvement those numbers show.

Where the cheapest fix lives

The most useful property of the ladder is that it makes the remedy specific: a qualification names exactly the cell that is missing. Whether that qualification ships is then a cost question.

This is the honest Koji angle. When re-fielding a missing segment takes weeks of recruiting and moderator calendars, the qualification ships, because the deadline wins. When a study can be launched and completed in hours - AI-moderated interviews in voice or text, adaptive follow-ups, analysis assembling as responses land - the missing SMB cell often becomes a thing you go and get rather than a thing you disclose. The ladder tells you precisely what to buy; a low marginal cost per study is what lets you buy it.

Structured questions help here too, because they make the shape of a gap unambiguous. When every participant answers the same scale, single_choice, multiple_choice, ranking and yes_no items, an under-filled cell is visible as a count rather than as a vague sense that the segment felt thin, and the open_ended material can be honestly reported for the segments that did fill. See the structured questions guide.

Frequently asked questions

What are the three modified research conclusions?

Borrowed from ISA 705, they are qualified (findings hold except for a specified part), adverse (the study measured the wrong thing and should not be used), and disclaimer (you cannot conclude at all). Which applies is decided by whether the problem is a misstatement or an inability to obtain evidence, and whether it is material but not pervasive or material and pervasive.

What does pervasive mean?

ISA 705 defines pervasive effects as those not confined to specific elements; or if confined, representing a substantial proportion; or, for disclosures, fundamental to users' understanding. The third clause matters most in research: if a reader who did not know about the problem would misread the whole document, it is pervasive regardless of how narrow the defect looks.

Is an emphasis of matter a criticism of the study?

No. An emphasis of matter is explicitly not a modification of the opinion. The work is sound and the finding is correct - you are insisting that a particular correctly-reported piece of context gets read, such as the study having been fielded during a competitor launch. It belongs at the top of the read-out.

What should I do when a stakeholder blocks access to a segment?

Treat it as a client-imposed scope limitation and name it as one. Auditing standards say that when restrictions significantly limiting scope are imposed by the client, the auditor should ordinarily disclaim. The point is not punishment; it is that a visible, named consequence changes the incentive, whereas a limitations bullet does not.

Is this just a fancier limitations section?

The difference is that a ladder has triggers and names. A limitations section is free text of unbounded length that readers skip, and it uses the same register for a minor caveat and a fatal one. A named rung placed above the fold forces a decision about severity and communicates it in one word.

Does a disclaimer mean I wasted the study?

No. A disclaimer is still a report. It states what was attempted, what blocked it, and what would be required to answer the question - which is directly actionable, and considerably more useful than a confident conclusion drawn from a study that could not support it.

Related Resources

• Assurance levels for research - the prospective commitment this ladder is the retrospective counterpart to
• The research risk model - allocating rigor before fieldwork
• Process controls vs output checks - catching a control that did not operate
• Choosing what to review - why some findings can be stated but not projected
• The research expectation gap - why stakeholders are disappointed by studies that were done right
• The structured questions guide - the six question types and making a gap visible as a count
• Presenting research findings - getting the modification read rather than archived