The Conditions You Cannot See From a Walkthrough

Why this matters

The walkthrough is the cheapest hour in the whole job and the one that decides whether the price holds. It is also a sampling method, and every sampling method has a coverage boundary. Shops that bleed on hidden conditions are rarely the shops that missed something obvious. They are the shops that walked a site, saw nothing alarming, and then quoted as though seeing nothing meant there was nothing there. "Nothing observed" is a statement about your survey, not about the building.

The finding that leaves the commercial track

If the walk turns up a live hazard, that finding stops being part of the quote conversation and does not re-enter it. On a gas odor, everyone leaves the building immediately, no switch is touched on the way out, no light is turned on or off, no phone is used inside, and the calls to the gas utility and to the customer are made from outside and upwind. On scorching, arc tracking or a panel cover that is warm to the back of the hand, the condition is written down and stated plainly to the customer the same day, and nothing is opened behind that cover unless the circuit is isolated, locked, tagged and verified de-energized under 29 CFR 1910.333(b)(2), which is where electrical work lives because 29 CFR 1910.147 excludes exposure to electrical hazards from work on conductors and equipment in electric utilization installations at (a)(1)(ii)(C). The live-dead-live proving sequence is NFPA 70E-2021, 120.5, and it reaches you through your employer's electrical safety program or a contract, in whichever edition that program adopts, not on its own authority.

A life-safety finding is not leverage and it is not a reason to win the job. Write it, hand it over, and let the commercial decision happen afterwards.

Coverage, not confidence

Think of the survey as producing two outputs, not one. The first is what you saw. The second, and the one almost nobody writes down, is the list of things the method could not reach in the time you spent. A quote built on the first output alone silently promises the second.

Five categories are structurally out of reach of a visual survey, and knowing them by name is what turns a vague worry into a line you can hand to somebody:

• Concealed by construction. Anything behind finished surfaces, under a slab, inside a chase, buried, or above a sealed ceiling. Visual access would require destructive opening, so the survey cannot resolve it at any level of skill.
• Revealed only by sequence. Conditions that appear when something is removed. What the deck looks like under the covering, whether a fastener pattern holds, whether the substrate is sound. These are not hidden by malice; they are hidden by the fact that the thing on top has not come off yet.
• Load or weather dependent. Faults that need a condition your visit did not have. A drain that only backs up under simultaneous demand, a joint that only weeps at full pressure, a circuit that only sags on a hot afternoon with the largest load running.
• Controlled by somebody else. Utility capacity, a landlord's riser, an association's rule, a shared service, a permit authority's interpretation. You cannot resolve these by looking harder; they resolve when a third party answers.
• Historical. Undocumented prior work. Whether the previous installer followed the method, whether a repair was a repair or a bypass, whether a modification was ever inspected. The evidence is often visible in fragments and readable only after you open the assembly.

What comes off the list, and why that matters more

The register is only credible if it is short and disciplined, so the harder half of this is deciding what does not go on it.

An item comes off the unknowns list when it is resolvable within the survey hour by a non-destructive check you already know how to do. Nameplate data, a measured clearance, a run of tape across an opening, a photograph of a label, a count of terminations, a reading you can take without opening anything energized. If you could have measured it and did not, it is not an unknown, it is an omission, and calling it an unknown in your paperwork is the thing that makes a customer distrust the rest of the list.

It also comes off when the answer does not change your price or your method. An unknown that resolves both ways to the same scope is a curiosity, not a risk item. Listing it inflates the register and trains the customer to skim it.

And it comes off when the cost of resolving it is lower than the cost of carrying it. If a small opening, an access panel or an hour with a camera settles the question, the honest move is to price that investigation as its own small piece of work rather than to carry the ambiguity into a fixed price. A shop that carries everything is charging every customer for a risk that only some of them have.

The register that replaces the worry

Six fields, one row per unknown, in the quote itself rather than in your head:

The fifth field is the one that does the work. An unknown with no named owner has an owner by default, and the default is whoever holds the fixed price. The sixth field is what makes the fifth field fair: a customer can accept carrying a risk they can see the size of, and almost never accepts one described only as "additional charges may apply."

Worked example: fourteen hours and six question marks

A replacement-type job. Base scope estimates at 14.0 labour hours from the shop's own history on that job type. The walk took 1.5 hours and produced six candidate unknowns.

Two came off the list during the walk itself. One was a clearance question, settled with a tape in about two minutes. The second needed an access panel opened on a piece of equipment that was already isolated, locked and tagged with the absence of voltage verified before the cover came off, and that added 0.5 hours to the survey. Both were resolvable, so neither belonged in an exclusion.

Four stayed:

1. Concealed routing between two points, hidden by finished surfaces. Assumption: the existing path is reusable. If it is not, the shop's own history on this job type says the rework runs 4.0 hours.
2. Shared circuit capacity, where the load calculation depends on what else is actually on the circuit behind a wall. Assumption: capacity is adequate. If it is not, 3.0 hours plus a material change.
3. Prior work permitting, which only the authority having jurisdiction can answer. Assumption: the existing installation is accepted as-is and is not required to be brought up to current adopted code as a condition of the permit. If the AHJ rules otherwise, the scope grows by 2.0 hours at minimum and possibly far more, which is why this one is written as a customer-carried item rather than priced.
4. Substrate condition under the covering, unknowable until removal. Assumption: sound. If not, 2.0 hours.

If every one of the four broke the wrong way and the third resolved at its floor, the additions run 4.0 plus 3.0 plus 2.0 plus 2.0, which is 11.0 hours against a 14.0-hour base, or 79 percent. That number is the whole argument for writing the register down. Nobody prices a 79 percent swing into a fixed number, and nobody should be asked to.

Now the allocation. Items 1, 2 and 4 are risks the shop can partly read and partly manage, so they are carried by the shop up to a stated hour cap and converted to a change beyond it. Item 3 has no shop-side skill that changes the outcome, so it is carried by the customer explicitly, with the assumption stated and the consequence named.

The shop's records on the last twelve jobs of this type show item 1 breaking the wrong way on three of them, which is 25 percent. Sizing a contingency on that gives 0.25 times 4.0 hours, or 1.0 hour, added to the 14.0-hour base. Read what that actually means before you use it: on three-quarters of these jobs the contingency is unearned, and on the other quarter it covers a quarter of the damage. A contingency prices the average across a run of jobs. It does not protect the individual job, and a shop that treats it as job-level protection will feel robbed the first time item 1 lands. Sizing method and the distribution behind it are covered in the sibling on the contingency line; the point here is only that the register is what makes the sizing possible at all, because you cannot put a rate on an unknown you never named.

How to tell the register is honest

Three checks, and none of them is a re-reading of the list.

Count the resolvable ones you resolved. If a job's register has zero items that came off during the walk, the walk was probably a look rather than a survey. Two came off in the example above and both were cheap. A register that only grows is a register being used as a disclaimer.

Check that each owner could actually have acted. A risk assigned to the customer that the customer has no way to investigate, price or influence is not an allocation, it is a transfer, and it reads that way in a dispute. Item 3 above passes because the customer can call the permit office and get the same answer you would. Item 4 would fail if you assigned it to them, because only the crew doing the removal can see it.

Look for the unknown you removed by assumption rather than by evidence. The failure mode is quiet: an item gets written as an assumption during the walk, then in the second draft of the quote the assumption reads as a fact and the row disappears. When the job opens up and the assumption was wrong, you are arguing about a sentence that is no longer in the document. Keep the row even when you are confident, because the row is the record of who agreed to the assumption.

References

• 29 CFR 1910.333(b)(2), selection and use of work practices for electrical work, including lockout, tagging and verification of de-energization; 29 CFR 1910.147(a)(1)(ii)(C) for the carve-out that sends electrical utilization work there
• NFPA 70E-2021, 120.5, the process for establishing an electrically safe work condition including live-dead-live verification, which applies through an employer electrical safety program or a contract in the edition that program adopts
• See related: Scoping the Hidden-Conditions Risk, The Contingency Line and How to Size It, The Existing Conditions Photo Record, Estimating the Unknown: The Conditions Clause