What to Record About a Supply So the Next Visit Is Shorter
Why this matters
Most supply records in most shops are numbers with nothing around them, which makes them decorative. A voltage written on a ticket with no note of where it was measured, between which conductors, or with what running cannot be compared to anything, so the next visit re-measures from zero and the visit after that does it again. The whole value of writing a supply down is that a future reading can be set beside it, and comparability is a property you have to design into the record, not something the number carries on its own.
The test every field below has to pass is a single question: could a different tech, a year from now, take a reading that is legitimately comparable to this one? If a field is not needed to answer that, it does not belong on the list, and this card spends as much space on what to leave off as on what to put in.
A reading is a value plus five conditions
The value alone is the least useful thing on the line.
- Where. The physical point, specifically enough to return to it: "load side of the main disconnect," "line side lugs of subpanel B," "at the equipment disconnect on the roof." Not "at the panel."
- Between what. Which two conductors. On a single-phase 120/240 V service that means naming line to line or which line to neutral; on a three-phase service it means naming the pair. A number with no pair attached cannot be used to find a bad leg later, which is the most common thing you will want it for.
- Under what load. What was running at that moment, in enough detail to reproduce. "Idle, shop closed" and "compressor running, two production machines on" are both good entries. This is the field that gets omitted most and costs the most.
- When. Date and time of day. Time of day is a proxy for load in most buildings, and it is what lets somebody schedule a comparable visit.
- On what instrument, set how. Meter type, whether it is true RMS, and whether the value is a live reading, a minimum or maximum capture, or an average over an interval. A demand average and an instantaneous reading are different quantities that print in the same units.
Record the value with its unit every time, even where it is obvious, because a record that mixes bare numbers is a record somebody will eventually misread.
The standing facts, written once per address
Separate from per-visit readings, some things about a supply do not change between visits and should be captured once, on the customer's file rather than on a ticket:
- Service configuration and nominal voltage, main rating, and the bus rating from the panel label.
- The panel label's maximum overcurrent device count and permitted device types, because a future add-a-circuit question is answered from that.
- Whether there is a transfer switch, and if so, its coverage stated as a circuit count and the loads that matter.
- The ownership boundary: where the service point is for this address, and who told you. A sibling card covers why that answer varies and where to get it.
- Grounding electrode arrangement as found, in a photograph.
Standing facts are what make a phone quote possible. Standing facts are what let you answer an add-a-circuit question on the phone instead of scheduling a visit to go read a panel label.
What stays off the list, and why
This is the part that decides whether the record survives contact with a busy day.
Adjectives. "Supply good," "voltage OK," "checked power." These read as work performed and carry no information. They are worse than a blank, because a blank prompts the next tech to measure and an adjective persuades them not to.
Percentages you calculated instead of the values you read. Record the three line-to-line voltages, not the percent unbalance you worked out from them. The raw values regenerate the percentage any time; the percentage cannot regenerate the values, and it hides the one thing you most want later, which is which leg was low. The same rule holds anywhere the record is tempted to store a conclusion in place of an observation.
Readings whose load state you do not know. These are not worthless, but they belong in the visit narrative flagged as uncontrolled, not in the comparable record. Promoting one into the record is how a future comparison goes silently wrong, because it looks exactly like a controlled reading and no one can tell the difference afterward.
Acceptance ranges written from memory. Do not write "should be 230 to 240" beside the value. The criterion for a specific piece of equipment is on its nameplate, and the criterion for the installation comes from the adopted code and the utility's stated service voltage range. Freezing a remembered band into the record hands the next tech a gate that may be wrong for the equipment they are looking at.
Anything you did not personally observe. What the customer says the electrician found in the last decade is customer history, which is useful and belongs somewhere else. A supply record holds measurements and observations with your name on them.
Worked example: two years, two visits, one comparison that worked
A small warehouse, 120/240 V single-phase service, intermittent complaint of lights dimming and a rollup door operator that occasionally fails to start.
Visit one record, as written. Measured at the load side of the main disconnect, line to line. Idle, building closed, 8 a.m.: 241 V. With the refrigeration compressor running and the door operator cycled, 9 a.m.: 232 V under that load, a 9 V sag, which is 3.7 percent of the idle reading. Instrument: true RMS meter, live reading. That is five conditions and two values, and it took about as long to write as this sentence takes to read.
Visit two, thirteen months later, same complaint, worse. Same point, same conductors, matched conditions deliberately because the record said what to match. Idle, 8 a.m.: 240 V. Same load combination, 9 a.m.: 224 V, a 16 V sag, which is 6.7 percent of that day's idle reading.
Read it. The idle readings are within 1 V across thirteen months, so nothing about the supply at no load has changed. The sag under the same load went from 9 V to 16 V, nearly doubling. Both sags are expressed against their own day's idle reading, at the same measurement point, under the same load, which is what makes the 3.7 and 6.7 percent figures comparable at all; comparing a sag against a different point or a different load would have produced two honest numbers and one meaningless comparison.
A supply that is unchanged at no load and materially worse under load is a resistance that grows with current, and that is a connection, not a utility-wide condition. The record made a two-year trend visible in the first five minutes of the second visit.
The version without the load field. The second tech opens the file, finds "241 V" from last year, measures 224 V today, and has no way to know whether the supply lost 17 V or whether last year's reading was taken with the building asleep. The honest response to that is to re-measure everything and start over, which is exactly the hour the record existed to save. The dishonest response is to call it a 17 V loss and chase a fault that is partly an artifact of the two readings not being the same measurement.
While you are taking the readings
Live measurement runs through the gate at 29 CFR 1910.333(a)(1): conductors and parts are de-energized before work on or near them unless your employer can demonstrate that de-energizing introduces additional or increased hazards or is infeasible. A load-dependent voltage reading cannot be taken dead, so it is done as energized work under your employer's electrical safety program with boundaries and PPE per NFPA 70E-2021, 130.5 and 130.7 in the edition that program adopts, and anything requiring contact is locked out per 29 CFR 1910.333(b)(2) - not 29 CFR 1910.147, which excludes electric utilization work at 1910.147(a)(1)(ii)(C) - and proved dead with live-dead-live at NFPA 70E-2021, 120.5. On a construction site the electrical lockout counterpart is 29 CFR 1926.417.
Two habits specific to record-taking are worth naming because the recording, not the diagnosis, is what creates them.
Do not manufacture a load condition. Recreating the customer's peak means asking them to run what they normally run. It does not mean starting equipment out of sequence, defeating an interlock, or running something past its duty cycle to get a bigger number. The load you invent is a process hazard you invented.
Do not leave an instrument inside an enclosure that then cannot be closed. A logger sitting on a live bus with the dead front off, or leads run through a door that no longer seals, has defeated the enclosure for the whole recording period. If a recording needs to run unattended, it needs a means listed for that use and an enclosure that closes on it.
Checking that the record actually works
Hand it to somebody who was not there and ask them to go take the same reading. Then count the questions they have to ask you.
Every question is a missing field, and the mapping is direct: "which panel" is the where field, "which legs" is the between field, "was anything running" is the load field, "was that a live reading or a max capture" is the instrument field, and "what time of day was that" is the when field, which is the one everybody assumes the ticket already carries. A record that produces zero questions is finished. A record that produces three has cost the next visit three phone calls, and phone calls between two techs on two sites are the most expensive minutes in a service day.
Do this once, on a real record, with a real colleague. It is faster than arguing about documentation standards and it settles the argument, because the missing fields are not a matter of opinion once somebody has had to ask for them.
References
- NFPA 70 (National Electrical Code), Article 100 definitions and the equipment marking requirements your standing facts are copied from, in the edition adopted by your authority having jurisdiction
- NFPA 70E-2021, 120.5, 130.5 and 130.7, as adopted by your employer's electrical safety program; 29 CFR 1910.333(a)(1) and 1910.333(b)(2), with 29 CFR 1926.417 for construction
- Equipment nameplates and the serving utility's published service voltage range, which together own the acceptance criteria a record should not try to store
- See related: What a Utility Owns and What You Own; The Fault That Was Upstream of Everything Anyone Checked; What a Transfer Switch Decides