How to Decide Whether to Work on a System You Did Not Install

Why this matters

Most shops decide this on vibes. The system looks rough, the customer seems difficult, the tech has a bad feeling, and the job gets quietly priced high enough to go away. That method throws away good work and accepts bad work at roughly equal rates, because age and appearance are weak predictors of whether a job will hurt you. The strong predictor is whether every unknown in it can be handed to somebody specific before you start. Here is a record you fill out on site that answers that, and a case where the same system is a no-go at one scope and a clean yes at another.

The record, and why each field exists

Seven fields, filled in on the walk, in this order. The order matters because field 2 is what most shops skip, and skipping it makes fields 3 through 5 unanswerable.

1. The ask, in the customer's own words

Write what they said, not what you think they meant. "Make the second floor comfortable." "Stop the noise." "Get this thing reliable."

The reason to record it verbatim is that an outcome-shaped ask and a task-shaped ask carry completely different risk, and translating it into a task in your head hides which one you are being offered. Skip this field and you will quote a task while the customer buys an outcome, which is the single most common way this goes wrong.

2. The depth the ask requires

Component replacement, subsystem intervention, or whole-system evaluation. A sibling article covers what each depth commits you to; here you only need to name it, because depth determines how many unknowns you are about to inherit.

An outcome-shaped ask is always whole-system, even when the work is one part. That is the field's whole purpose: it converts "just add a zone" into "you are being asked to certify a 30-year-old system's behaviour."

3. Every unknown, listed one line each

Not a paragraph about how old it is. A list of specific things you do not know that could change the result. Piping you cannot see, prior modifications by parties unknown, a heat exchanger's internal condition, whether an emitter or fixture was ever sized for its load, whether a circuit was ever loaded to its rating.

If your list has fewer than three lines on a system you did not install, you have not looked, you have glanced.

4. For each unknown, its observability, priced

Beside each line, write what act would resolve it and how many hours that act takes. Some resolve for well under an hour. Some do not resolve at all without demolition. Some resolve partially, and partially is its own answer, not a rounding error.

This is the field that replaces the bad feeling with a number, and it is the reason the record beats intuition: an ugly system with six cheaply observable unknowns is a better job than a tidy one with a single unobservable one.

5. Disposition per unknown: resolve, exclude, allowance, or unassigned

Four possibilities and only four.

• Resolve. You do the observing act and the unknown becomes a known before you price the work.
• Exclude. You state in writing that it is not in scope and not warranted, and the customer keeps it.
• Allowance. Neither of you knows, so you price a stated quantity of work against it and true it up against actual. An allowance is a percentage of a scope and a range of outcomes, not a guess dressed up as a number.
• Unassigned. You have not decided. This is not a disposition, it is the absence of one.

6. The gate

Any unknown still marked unassigned is a no-go at that scope. Not a no-go on the job. Re-scope until every line reads resolve, exclude or allowance, or decline.

The unit of analysis is the individual unknown, not the job, and the test is AND: every line must carry a disposition. One unassigned line is enough, because an unassigned unknown is one you have carried by default, and the shop that carries an unknown by default finds out which one it was at the worst possible moment.

7. Safety findings, on a separate line, never traded

Anything you find that endangers people goes here and does not enter the commercial arithmetic at all. It gets stated plainly and in writing regardless of whether you take the job, decline it, or never hear from the customer again.

Worked example: "make the second floor comfortable"

A hydronic heating system, roughly 30 years old, installed by a company nobody can name. The customer's ask, verbatim: make the second floor comfortable.

Field 2 is immediate: that is an outcome, so the depth is whole-system, not "add a zone valve."

Fields 3 and 4, as filled on site:

Field 5, first pass, against the ask as stated. U1 resolve. U2 exclude the internals, resolve the combustion side. U4 resolve. U5 partial resolve, exclude the rest. U3 has no disposition available. It is unobservable, and it cannot be excluded either, because the thing sold is a temperature and embedded piping is a way that temperature fails. Exclusion only works when the customer can still get what they bought without the excluded item.

Field 6 fires. No-go at the outcome scope.

Now re-scope rather than walk. The shop offers the 3.0-hour heat loss and measurement study, plus the 1.5-hour circulation check, as a deliverable in its own right: a written report of what the system is and what it would take to reach the customer's goal. No temperature is promised. U3 becomes excludable the moment the deliverable is a report rather than a result, because a report can accurately state that embedded runs were not inspected and the customer still receives exactly what they paid for. Every line now carries a disposition. Go.

Two hazards live inside those 4.5 hours and neither is covered by a generic entry block. Taking surface temperatures on operating hydronic pipe means hot metal, so readings are taken with a non-contact instrument or with gloves rated for the surface temperature, never with a bare hand steadying the pipe. Do not pull, cut or peel pipe insulation to reach bare metal. Thermal system insulation of that vintage is presumed to contain asbestos under the construction asbestos standard, 29 CFR 1926.1101, with a general-industry counterpart at 29 CFR 1910.1001, and disturbing it releases respirable fibres that no glove or safety glass addresses; if insulation genuinely must be disturbed, it is assessed and handled under that standard by people trained and equipped for it, not opened with a utility knife to get a reading. For the combustion analysis, a personal CO monitor goes on before entering the equipment room and stays on for the whole visit, because the analysis requires the appliance to fire and a firing appliance is exactly the condition under which a flue defect spills into the room.

What the study found

U1 resolved clean. Both upstairs loops circulating, supply-to-return temperature drops within a normal band, no air, no closed valve. A shop working on instinct would very likely have sold a balancing job here and delivered nothing.

U4 is the finding. Measured emitter surface upstairs, rated at the system's existing 160 F supply water against a 70 F room, produces about 62% of the calculated design-day heat loss for those rooms. The system is not out of balance. It is short of radiation, and it always was.

Test the obvious fix before selling it. Raising supply water from 160 F to 180 F raises the water-to-air temperature difference from 90 F to 110 F. Natural-convection emitter output rises roughly with that difference raised to about the 1.3 power, an approximation that describes finned and cast emitters over ordinary residential temperature differences and should be replaced by the manufacturer's published output at your actual water temperature wherever that data exists. That gives about a 30% increase, so 62% becomes roughly 80%. Still short, and now with two live constraints: on a condensing appliance, pushing supply temperature up drives return water above the range where flue gases condense and gives away the efficiency the appliance was bought for, while at the other end, on a non-condensing appliance, return water below roughly 130 F condenses flue gas and corrodes the heat exchanger from the inside.

So the honest answer to the ask is added emitter surface, priced as its own project, and the study said so in one page. The customer bought that project. The instinct version of this job, "old system, difficult ask, price it high," would have either lost the work or won it as a balancing call that could not possibly deliver, followed by a callback nobody could win.

The failure mode

The one that costs real money is not taking a bad job. It is taking a good job at the wrong scope. Nothing about that system was dangerous, badly installed or unworkable, and a shop that declined it on appearance would have declined a project that paid. The defect that would have hurt was quoting a temperature against an unobservable pipe run.

The second failure mode is treating the record as paperwork. If a tech fills fields 1 through 5 and then a manager overrides field 6 because the customer is a good account, the record has made the decision worse rather than better, because now the shop has documented in its own hand that it knew about the unassigned unknown.

How to verify you got this right

• Re-read field 1 against your quote. If the customer's words describe an outcome and your quote describes a task, one of the two has to move before you send it.
• Count the unknowns and count the dispositions. Equal, or you have not finished.
• Check that at least one unknown was resolved rather than excluded. A record where everything is excluded is a shop transferring all risk to the customer, which is a legitimate offer and rarely a competitive one.
• Check that any allowance names its quantity and how the true-up works. An allowance with no stated basis is an unassigned unknown wearing a better word.
• Ninety days later, pull the record on any job that went sideways and find the line that predicted it. If the line is there and was marked resolve or exclude, your gate works and your execution slipped. If it is not there at all, your field 3 is too short, and that is the fix.

References

• 29 CFR 1926.1101, asbestos (construction), including the presumed-asbestos-containing status of thermal system insulation; general-industry counterpart at 29 CFR 1910.1001
• See related: What You Take On When You Touch Somebody Else's Work
• See related: Scoping the Hidden Conditions Risk
• See related: The Contingency Line and How to Size It