The electrical count takeoff guide

Start at the symbol legend

Every electrical set opens with a legend sheet mapping symbols to devices — and no two engineers draw them identically. A duplex receptacle, a GFCI, a quad, a floor box, and an isolated-ground receptacle can be five nearly identical circles with different tick marks. Before counting anything, read the legend end to end and note the symbols that look alike. When a symbol on the plan isn't in the legend (it happens constantly), write an RFI or flag it as a scope question — don't guess. The legend is the contract between you and the drawing; a takeoff that ignores it is a list of shapes, not devices.

Count devices by type, one pass per type

Resist the urge to count everything in a room and move to the next room. Counting one device type across the whole sheet — all duplex receptacles, then all GFCIs, then all switches — is faster and dramatically more accurate, because your eye is pattern-matching a single shape instead of re-deciding what each symbol is. The usual passes:

• Receptacles: duplex, GFCI, quad, floor, special-purpose (ranges, welders, equipment), usually split by mounting height where the drawings note it.
• Switches: single-pole, 3-way, 4-way, dimmers, occupancy sensors, low-voltage switching.
• Lighting fixtures: by fixture type letter (A, B, C…) against the fixture schedule — the schedule defines what each letter costs.
• Data and low voltage: data/phone outlets, AV, security devices — often on separate T-series or E-series sheets, sometimes carried by a different sub. Know whose scope they're in before you count them into yours.

Mark every symbol as you count it. On paper that was a colored pencil dot; digitally it's a count marker. An unmarked symbol is a symbol you'll either miss or count twice.

Home runs and branch conduit

Counts don't capture the wire. Home runs — the arrow-tailed lines from the last device back toward the panel — get measured as linear feet, and the plan only shows them schematically, so the convention is: measure the drawn run, then add the vertical drops and rise at each end, and apply your standard slack allowance. Branch wiring between devices is often estimated by device-count formulas (an average LF of conduit and wire per device, tuned by building type from your history) rather than traced line by line — a defensible shortcut as long as you state it and apply it consistently. Feeders and larger conduit runs deserve real linear measurement, floor by floor, with vertical rises added explicitly.

Reconcile against the panel schedules

The panel schedules are your built-in answer key. Each panelboard schedule lists its circuits with descriptions and loads; if you counted 46 receptacles fed from panel LP-1 but the schedule shows receptacle circuits that could plausibly serve 30, one of you is wrong — and it's worth twenty minutes to find out which. Fixture counts reconcile the same way against the lighting fixture schedule totals, when the engineer provides them. Schedules also expose scope the plans understate: spares, space-only breakers, and equipment connections that appear in the schedule but are easy to miss on a crowded plan.

The double-count trap

The most expensive systematic error in count takeoffs is counting the same device on two different sheets. It happens in predictable places:

Verification passes

A count takeoff isn't done when the counting is done; it's done when it's been checked. Three passes that catch most errors: a visual sweep of every sheet looking for any unmarked symbol; a ratio sanity check (receptacles per SF and fixtures per SF against similar past jobs — an office floor with one receptacle per 400 SF should make you suspicious); and the schedule reconciliation above. If the project has addenda, re-check the affected sheets specifically — a one-line addendum note like "add GFCI at each column" can move a count more than a redrawn plan does.

Counting with software: AI proposes, you dispose

Everything above is method, and it works with a highlighter. Software's contribution is speed with an audit trail. In Groundwork Takeoff, the Count tool drops a visible marker on every symbol you click — color and shape per device type — so the "mark everything you count" discipline is automatic, totals accumulate live in the estimate grid, and every count traces back to marks on the sheet. Snap and the crosshair keep dense fixture layouts precise.

For the long repetitive passes, AI Detect finds candidate symbols matching one you select — and then, critically, it proposes rather than commits. You review the found instances and approve them; nothing enters your count because a model was confident. We built it that way deliberately: your stamp goes on the bid, so you stay the estimator of record. AI proposes, you dispose. It's included on every seat, not an add-on tier — see pricing, or how that compares to other tools' AI pricing on our comparison hub. When counts are approved, export the whole thing to Excel or CSV, split by device type, ready for pricing. And before any bid goes out, run our 12-point accuracy checklist — the count-once rule is item five.