The Review Step That Catches the Most In a tired October week, the review step is the first thing a firm cuts — and the one week it catches the most. A practitioner's case for protecting the step that decides whether quality survives the deadline.

There's a step in every tax return that costs almost nothing when the week is calm and gets quietly deleted when the week is not. It's the review — the moment a licensed person looks at a number that's already on the return and confirms it against the document it came from. In March, with a full desk and a clear head, that step is second nature. In the second week of October, half-staffed and forty returns deep into an extension stack, it's the first thing to go. Which is exactly backwards, because October is when the review step catches the most.

I want to make the case for protecting it — not as a compliance nicety, but as the single highest-leverage habit a firm has in peak load.

What the review step actually catches

Ask a preparer what review is for and you'll often hear "typos." That undersells it. The errors that matter in an extended return are rarely dramatic. They're plausible. A number transposed from a brokerage statement that's off by a digit but still looks like a reasonable figure. A K-1 that arrived as a photograph and got keyed under time pressure. A wage figure that was typed once by a payroll company, printed onto a form, and typed a second time into a field on your screen — and the second typing is where it went wrong. None of these announce themselves. They pass every smell test except the one that matters: does this number match the source in front of me. The review step is the only place that question gets asked out loud.

Why speed makes it more necessary, not less

Here's the uncomfortable part. The instinct under a deadline is that speed and verification trade off — that a firm moving fast has to accept a little more risk. The opposite is true. Speed doesn't just make errors more likely; it makes them more expensive, because a fast, tired, half-staffed week is exactly when a wrong number sails through unchallenged. The failure-to-file penalty runs at five percent of the unpaid tax per month, up to twenty-five percent — the expensive one, ten times the failure-to-pay rate. An extended return that goes out wrong, or goes out late because it had to be redone, walks straight toward it. In October the cost of a missed error isn't a quiet amended return in the spring. It's the deadline you extended specifically to avoid.

The number that's confidently wrong

This is also where I'd caution any firm about how it thinks about automation, including the kind I build. A machine can read a document beautifully and hand you a clean, well-formatted number that is wrong, and it will never once raise its hand to say it wasn't sure. Confidence is not accuracy. In a calm week a preparer catches that easily. In the crunch, when the same preparer is moving fast and trusting the tidy output, a confidently wrong number is the most dangerous object in the building — precisely because everything about it looks finished. The lesson isn't "don't automate." It's "never let automation retire the review." The reading can be delegated. The confirming cannot.

Extraction is not verification

That distinction is the whole architecture question underneath this. There are really two jobs hiding inside "getting the data onto the return." The first is extraction: reading a document and pulling the field values off it. The second is verification: a licensed person confirming each of those values against its source before the number is trusted. Most of the noise in tax technology is about the first job, because reading a W-2 is the part that demos well. But the first job was never where a firm's exposure lived. A number that has been extracted but not verified hasn't saved you anything — it has just moved the risk downstream to a return with your signature on it. The design that survives a deadline keeps those two jobs separate on purpose: the machine reads and places, a person confirms every field. The speed comes from the extraction. The safety comes from the review. Collapse them and you've built the thing that fails quietly in October.

What the client is really paying for

It's worth remembering what the review step is from the client's side of the desk, because it reframes the cost. A client cannot tell whether their return was assembled carefully or hastily. What they experience is the absence of a problem — no notice, no amended return, no letter in the spring asking about a number that didn't match. That quiet is the product. It's easy, in the crunch, to treat review as internal overhead the client never sees and therefore never values. It's the opposite. It's the one part of the work that decides whether the client ever has a bad month because of you. The firms that keep clients for a decade are not the fastest ones. They're the ones whose returns simply don't come back — and that reputation is built one verified field at a time, never faster than in the October weeks when the temptation to skip is highest.

How to protect the review step in peak load

So, practically, three things — the ones I'd do this week if I were back at a firm staring at the stack.

First, separate assembly from judgment before you touch a single return. Sort the extended pile by what each return is actually waiting on. Most of what's left in October isn't waiting on your expertise; it's waiting on assembly — a document to arrive, a number to be moved, a moved number to be checked. When you see the stack that way, the returns that genuinely need your judgment are a short list, and the rest is a transcription problem wearing a deadline costume.

Second, take the transcription off your licensed people without taking them off the return. The keying — reading a document, pulling the fields, moving them into the software — is the part that was never judgment and is now the part that can be lifted. That's the genuinely new fact about this year: a preparer can stop being the one who types without stopping being the one who verifies. It's the only way to buy back review time in a week where you can't buy more hours.

Third, make verification structural rather than optional. The single change that protects the review step under load is putting the source document in front of the person, field by field, so that confirming means looking — not scrolling past a tidy list and clicking approve. A review that's easy to skip will get skipped in the crunch. A review that's built into how the number gets confirmed survives the week it matters most.

None of this is glamorous. The review step is the least impressive part of the whole workflow — no one demos it, no one markets it, and it's the first thing a tired team is tempted to cut. But it is the cheapest insurance a firm buys all year, and the only week it truly earns its keep is the one most firms are in right now. The extended returns on your desk don't need you to work the wall harder. They need the one step that's easy to skip and expensive to lose. Protect the review, and October stops being the month your quality quietly drops to meet the deadline.

Sources referenced: IRS penalties guidance — failure-to-file penalty (5% per month, up to 25%) and failure-to-pay penalty (0.5% per month, up to 25%), irs.gov; IRS tax calendar — extended individual (Form 1040) and calendar-year C-corporation (Form 1120) returns due October 15, 2026, irs.gov.