Hiring an examiner right now is a project: weeks of posting, two candidates who look real, one of whom takes the other offer. Then they show up Monday and the plan is: sit next to whoever’s least buried. That isn’t a plan. It’s a seating chart. Eight months later they leave and the shop decides the labor market is brutal. The market is brutal. That’s not why they left. Back in June I argued that examiner retention rides on the quality of the job, not the size of the wage. This is the operational half: if the job experience is what keeps people, the first ninety days is where you build it or lose it.

Osmosis Is a Curriculum Written by Your Order Queue

Shadowing isn’t useless — watching a good examiner work is how a lot of this gets learned. But unstructured shadowing has no syllabus. What they learn is a function of which files came through that week. Hire during a stretch heavy on refis and they’ll be sharp on refis, blank on everything else. Hire the next one into two commercial files and a messy estate, and they get a different education entirely. Neither of them knows which one they got. Neither do you. And it costs you twice: the senior examiner you parked them beside is now doing two jobs badly instead of one well, behind by Thursday on their own queue. That’s the person you can least afford to lose.

The Part Nobody Wrote Down

Procedure is writable, and most shops have written it — a binder, a shared-drive folder, a wiki somebody built in 2019. Nobody reads it, and it barely matters: procedure is what a competent adult absorbs in a few weeks. The hard part is judgment. Which anomalies matter. When a gap in the chain is a real defect and when it’s an artifact of how a county indexed for a decade. When to call the underwriter — and when calling makes you look like you can’t do the job. That knowledge lives in three or four heads in your building, and those people can’t easily tell you what they know. Ask a twenty-year examiner why they flagged something and you’ll get “it just looked wrong.” They’re not being cagey; compressed pattern recognition doesn’t decompose into sentences. So the bottleneck isn’t documentation — it’s that there’s no ordinary mechanism for transmitting judgment, and we’ve agreed to let time handle it.

Deep First, Then Wide, Then Ugly

So what should ninety days look like? Not a rotation. Start narrow: one transaction type — whatever your bread and butter is — and a lot of it. Judgment comes from pattern density. Twenty residential purchase files in a week teaches more than twenty different things once. Every one gets reviewed, the same day — the part that gets cut first and shouldn’t be. Early on, feedback frequency beats feedback depth — ten fast “yes, that’s right” or “no, here’s why” calls a day beat a thorough Friday review, because by Friday they’ve spent four days practicing whatever they got wrong on Monday.

Then widen, one transaction type at a time, and only once the last one is unaided. Then, deliberately, the ugly files. Don’t wait for a broken chain or an unadministered estate to wander in on its own. Pull files you’ve already closed and hand over the mess on purpose, with someone watching. Otherwise the first hard one arrives unannounced, in a volume week, with a closing date on it.

The Highest-Value Hour in the Whole Ramp

Take a file the senior examiner already cleared and have them walk the new hire through it out loud. Not the exceptions they raised — the whole read. What they looked at first. Where they slowed down, and why. And above all, what they dismissed. New examiners see what got flagged. They never see the twelve things the senior ruled out in four seconds, because ruling something out leaves no trace on the file. The dismissals are where the judgment lives. “I saw that, here’s why it doesn’t matter” is a sentence a senior examiner can actually say — even when “here is my model of risk” isn’t.

Day 30 and Day 60 Have to Mean Something

If you can’t say where someone is at day 30 versus day 60, you don’t have a program. You have a hope with a start date. Track which transaction types they clear unaided, against a schedule set in advance rather than discovered afterward. Track rework rate, watching direction more than the number. And track escalation appropriateness, which cuts both ways: escalating nothing at day 45 isn’t confidence, it’s unawareness; escalating everything at day 75 means nobody gave them permission to decide. I’m not handing you target numbers — I don’t have defensible benchmarks, and neither does anyone quoting them at you. What matters is that the numbers exist and somebody looks at them on a date set ahead of time.

My Bias, Stated Plainly

The people side of this business is not my strongest suit. I’m an engineer by instinct — twenty-six years in, my first move on a human problem is still to reach for a system. That’s a bias, not a virtue. I’d defend the frame anyway. The standard advice to a shop bleeding new examiners is some version of invest in mentorship. It isn’t wrong. It’s just not actionable for an owner working sixty hours who is also the escalation path for every hard file in the building. “Try harder” is not a plan. Structure is what you give people when willpower is already fully committed. A sequence, a review that happens whether or not anyone feels like it, two dates on the calendar — those beat good intentions because they don’t depend on anybody’s Tuesday.

Where the Software Earns Its Keep

All of this rests on getting what your senior examiners know into a form someone learning can use, and I won’t pretend a tool solves that outright. But that reasoning around work is what Scribe and Digital Abstracts capture, so the thinking behind a decision doesn’t evaporate when the file closes. The value isn’t that software teaches them. It’s that a record of how experienced people worked a file is there at 4:30 on a Thursday, when the person who knows is in a closing. You spent three months finding this person. Spend the next ninety on purpose. The ramp is the retention plan — very few people quit a job they can feel themselves getting good at.

See how Scribe and Digital Abstracts preserve how a file was worked →