Large projects average 796 RFIs, and each one takes 8 hours and $1,080 to review and answer.

The Navigant Construction Forum looked at 1,362 large projects for its 2013 study, Impact and Control of RFIs on Construction Projects. Most of them were in Australia and New Zealand. The average project carried 796 requests for information. Each one took about 8 hours and $1,080 to review and respond to, and the median reply came back in 9.7 days.

Those are big projects, and they are not the jobs on my desk. But I recognise the shape of it. I'm 18 and I count lighting fixtures by hand at a preconstruction desk. The RFIs I have caused were never about design intent. They were about a number I could not defend.

Here is how it usually goes. Type F comes to 41 by my count. The schedule tag on E-2.1 reads 38. Maybe I counted a fixture twice where it sits across a match line. Maybe the drafter missed a tag. Maybe the engineer added three fixtures and never updated the callouts. From my desk I cannot tell which one is wrong, so I write the RFI and keep going with the rest of the set.

Nine days later the answer comes back. See revised plan. By then the bid has moved on, and whatever I priced in the meantime was priced on a guess.

The recount itself is cheap. Going back over one fixture type on a handful of sheets is a small job. The nine days are not cheap, and neither is an RFI that never needed to be written, because the question was really mine to answer.

What LumenTakeoff does about it

So the review by fixture card shows the numbers separately and never merges them. Detected instances. Production count, which is the exact-signature count and stays unchanged by grouping. In bid scope. Exact signatures. And tag on sheet.

Tag on sheet is how many times the schedule tag appears on the drawing. It is shown as a document-stated figure with its discrepancy, never as a count. If the symbols say 41 and the tags say 38, the card shows both and shows that they disagree. It does not quietly pick one.

That disagreement is where the work is. From there I can open the plan viewer, zoom into the actual sheet, and see every member of the family ringed where it sits, with the set's own fixture schedule beside it. Three extra symbols with no tag next to them is a question for the engineer. Three symbols that turn out to be a different fixture is something I settle at my own desk.

Nothing gets blended into one total along the way. When I lock in results, the reconciliation check runs first, and export is refused rather than exporting a wrong number. The XLSX carries an Evidence sheet saying where every figure came from, and the proof images ring every counted instance.

When my 41 and the sheet's 38 disagree, I see it before bid day instead of mailing it out as a question after. Some of those questions still go to the engineer. They just go with a sheet number and a ringed symbol attached, and the ones I can answer myself, I answer at my own desk.

The machine counts. You decide.