Change Value as a Share of the Original Job

Why this matters

Total change order value for the quarter is not a finding. It is a quantity with no base, and it goes up when you are busy, down when you are quiet, and sideways for reasons nobody can act on. The same is true of the average change order value, which additionally averages a small correction on a small job with a major addition on a large one. The number that does work is the ratio of change value to the original contract value, read per job, and what it says depends almost entirely on the shape of its distribution rather than on its size.

A total with no base is not a number yet

Divide the quarter's change value by the quarter's original contract value and you have a shop-level figure. Divide each job's change value by that job's own original contract value and you have a distribution. Both are useful and they are not the same statement, and a shop that quotes only the first will eventually compare it against a quarter with a different job mix and conclude something false.

Three conventions have to be fixed before either figure means anything.

The base is the original. A job's change ratio is measured against the contract value at the point the work was authorised, not against the value as it stands after earlier changes. The second convention compounds and it hides growth. Take a job with three changes, each written at about 10 percent. Measured against the original, the job grew 30 percent. Measured against the running value, each change is still reported as about 10 percent while the job actually grew to 1.1 times 1.1 times 1.1, which is 1.331, so 33 percent. The reported per-change figures are identical and the job is a different job. Use the original as the base, and report cumulative growth as its own number rather than leaving a reader to add the increments.

The unit is the job, not the change order. A job with four small changes and a job with one large change can carry the same ratio, and the ratio is the thing that tells you about your estimating. Change order count per job is a separate question that the approval rate card handles.

The clock starts at authorisation. An estimate revised three times before the customer accepted it produced no change orders, because the scope moved before any work was authorised. That is correct treatment and it is also the loophole: a shop that keeps re-quoting until the number sticks will show an excellent change ratio and still be estimating badly, since the same missed items were found in the same way, just earlier. Count pre-authorisation estimate revisions as their own figure beside the change ratio. If revisions per accepted estimate are climbing while the change ratio falls, nothing improved and the work moved upstream.

What the total quietly includes

Three populations ride along inside a headline change value, and each one has to be a deliberate choice rather than a default:

• Rejected and withdrawn changes. Including their value measures what was proposed; excluding it measures what was billed. Both are legitimate and they answer different questions. The proposed figure is the better estimating signal, because a change that was needed and refused still says the scope was short. The approved figure is the better money figure. Pick per use and label it, because the gap between them is invisible once they are added together.
• Deductions and credit changes. Scope removals carry negative value and they net against additions. A job with a 20 percent addition and a 15 percent deduction nets to 5 percent of its own original value and reads as a clean job, when in fact the scope was wrong twice in opposite directions. Report gross additions and gross deductions separately and then the net. The netted figure alone will hide a shop that quotes badly in both directions.
• Jobs with no original contract value. Time-and-materials work has no base, so it cannot enter the ratio. Exclude it explicitly and state how many jobs you excluded, or the distribution is quietly computed over a narrower population than the job count beside it.

Two distributions, one headline

Two shops each close 40 jobs in a quarter. Both report total change value at about 8 percent of their quarter's total original contract value. Nothing else about them is alike.

Shop A, the consistent small uplift. Thirty-four of the 40 jobs carried a change, and the jobs are of broadly similar size. Each of those 34 changes ran between 6 and 12 percent of its own job's original contract value, averaging about 9.4 percent. Because 6 of the 40 jobs carried nothing, the quarter total is 9.4 percent times 34 over 40, which is 8.0 percent of the quarter's original contract value.

Shop B, the rare large change. Three of the 40 jobs carried a change and 37 carried none. Those three are among the largest jobs of the quarter, together holding about 20 percent of the quarter's original contract value, and their changes averaged about 40 percent of their own original contract values. So the quarter total is 0.40 times 20 percent, which is 8.0 percent of the quarter's original contract value.

Same headline, and they are not the same problem. Shop A has an estimate that is short by roughly a tenth on almost everything it quotes. Shop B estimates well and then, on one kind of job, finds something expensive after the price is set.

One consequence worth stating because it is the reason to look at the distribution at all: in Shop A the mean of the 40 per-job ratios is also about 8 percent, while in Shop B it is 3 percent, since three jobs at 40 percent and 37 at zero average to 1.20 over 40. Shop B's mean of ratios and its total-over-total ratio sit 5 points apart because its changes are concentrated on its large jobs, which is spread in both size and rate; the average-of-averages card owns why that gap opens.

Telling which one you are in

Three figures separate them immediately and the total cannot:

The median is the cleanest single tell. A median near zero with a real total means the money is concentrated in a few jobs, full stop, and no template change will reach it. A median that sits close to the shop-level ratio means every job is carrying it and the template is where the fix lives.

There is a third shape, and it is the most common in practice: a bimodal distribution with many small ratios and two or three large ones. Do not average those into a single correction. Split the jobs at the gap, treat the lower group as Shop A and the upper group as Shop B, and apply both fixes independently.

The fix each distribution calls for

Shop A needs a scope correction, not a conversation. A consistent uplift across most jobs is an item or an allowance the estimate leaves out. Find it by reading the change order line items rather than the totals: in a distribution like this, one or two descriptions will appear on most of the changes.

Shop B needs a better survey on one job type, not a template change. The three jobs are the finding; the other 37 are evidence the estimating is sound. Identify what was discovered on each of the three and when. If the answer is something that would have been visible at survey with more time or a different access point, that is a survey scope problem. If it was genuinely undiscoverable before opening up, the honest fix is commercial rather than technical: quote that job type with a stated allowance for the unknown, or quote the opening-up as its own phase and price the rest afterwards.

The sizing for Shop B is a comparison of two different kinds of quantity, so put both on the quarter rather than against each other. That job type was quoted 9 times in the quarter, and 2 more hours of survey on each quote is 18 hours. Against that sits 8 percent of the quarter's original contract value arriving as work nobody planned, on three jobs. The 18 hours is small, fixed and known; the other is variable and has already fired three times.

Sizing the Shop A correction two ways

Read the line items on Shop A's 34 changes and one item appears on 27 of the 34 changed jobs, accounting for about 6.0 of the 9.4 percentage points of the average change ratio. Twenty-seven of the 40 jobs in the quarter is about 68 percent of them, so it is needed on most jobs and not on all. Two routes land the quarter's change ratio in the same place:

• Fold it into the template. It is quoted and performed on every job of that type. The remaining average change is 9.4 minus 6.0, which is 3.4 percent of each changed job's own original value, and the quarter total becomes 3.4 times 34 over 40, about 2.9 percent of the quarter's original contract value. The cost is that 13 of the 40 jobs are now quoted for work they will not need, at about 6 percent of each of their own values, which is 13 times 6 over 40, or about 2 points of the quarter's contract value quoted for nothing.
• Determine it at survey and quote it as a named line. The same 2.9 percent lands, without overpricing the 13 jobs, at the cost of the survey time and a longer quote.

The decision rule is the hit rate, and it is worth stating as a number you tune: fold it in when the item is needed on more than roughly 85 percent of that job type, and price it at survey when it is needed on a clear majority but not nearly all. At 27 of 40, about 68 percent, this one goes to survey. If the item were appearing on 38 of 40 jobs, folding it in would be plainly right and the 2 jobs mispriced would be noise.

There is a third route, which is to change nothing and keep billing it as a change. It works commercially, since the approval rate on a genuinely needed item is high. What it costs is not money but friction: every one of those 27 jobs carries a mid-job conversation, a signature, a schedule adjustment, and a customer who learns that your quoted price is an opening figure. That is the cost that never appears in the ratio, and it is the reason the ratio is worth pulling to zero rather than simply pricing around.

References

• See related: Gross Margin Percent Is an Average of Averages - why a mean of per-job ratios and a total-over-total ratio separate, and what it takes to open the gap
• See related: Change Order Approval Rate and Why a High One Is Not Good News - how often changes are raised, and the story the approval rate cannot tell on its own
• See related: Phase Completion Rate and the Skipped Phase Problem - the same scoping fault seen from the phases that get removed rather than added
• Trade-standard practice: change orders priced against the original contract sum, with cumulative contract growth reported separately from each increment