Average Estimate Value Counts Drafts and Conversion Does Not

Why this matters

These two figures are almost always printed side by side, and they are computed over two different sets of estimates. One counts the quotes that were never finished; the other cannot see them at all. Nobody notices, because both numbers are individually correct and neither one announces its population. Every conclusion drawn from the pair is then unsound, and the conclusions shops reach from it are expensive ones: raise prices, chase bigger work, stop quoting small jobs.

Two figures, two populations

Average estimate value, as it is normally built, is the mean total across every estimate raised in the window. A draft that was started at the counter, half-filled, and abandoned is an estimate that was raised. It carries a total. It is in the mean.

Conversion rate is computed over estimates that got past draft. An unfinished quote never reached a customer, so it was never an opportunity to win, and excluding it is correct.

Both decisions are defensible on their own. The problem is that the two figures then sit on the same line, get read as a pair, and get multiplied together in someone's head.

The same shop, both ways

A quarter: 80 estimates raised. 24 never left draft. 56 were sent.

Index this shop's smallest routine repair quote at 1.0 unit, and read everything in those units.

• The 24 drafts average 0.8 units each, so they carry 19.2 units in total.
• The 56 sent estimates average 3.0 units each, so they carry 168.0 units.
• All 80 together carry 19.2 plus 168.0, or 187.2 units, and 187.2 over 80 is a reported average estimate value of 2.34 units.

So the headline figure is 2.34 units and the average quote an actual customer received is 3.0 units. The reported number sits 0.66 units below it, which is 22.0 percent of the 3.0. The shop believes its typical quote is about a fifth smaller than the one it actually sends.

Of the 56 sent, 21 were approved, so conversion reads 21 over 56, or 37.5 percent.

The sentence that joins them is the defect

Here is what gets said in the review: our average quote is 2.34 units and we close 37.5 percent of them, so every quote we write is worth about 0.88 units of booked work.

Two things are wrong with that, and the second is worse than the first.

The populations do not match. The 2.34 is per estimate raised, over 80 records. The 37.5 percent is per estimate sent, over 56 records. Multiplying them produces a figure that is per neither. If you want a per-sent figure, both terms have to be per-sent: 3.0 units times 37.5 percent is 1.125 units.

The rate is not independent of the value. Even the corrected 1.125 is wrong, because multiplying an average by a rate quietly assumes large and small estimates convert at the same rate, and in almost every shop they do not. The honest figure is the approved value divided by the sent count, and that needs the split below.

Value split by outcome

Take the 56 sent estimates apart by what happened to them.

The totals close: 39.9 plus 117.6 plus 10.5 is 168.0 units, and 21 plus 28 plus 7 is 56 estimates.

Now the real number. Booked value per sent estimate is 39.9 over 56, or 0.7125 units. Against that, the review's 0.88 is 23.2 percent high and the corrected-pair figure of 1.125 is 57.9 percent high. The larger error came from the more careful calculation, which is the part worth remembering.

And the finding the headline pair could never produce: a rejected estimate at this shop averages 4.2 units against an approved one at 1.9, a ratio of 2.2 to 1. This shop closes its small work and loses its large work. That is a pricing, trust or presentation problem on big tickets, and it is the single most useful thing in the quarter's estimating data. No arrangement of the average value and the conversion rate reveals it.

The 7 undecided estimates are the truncation artifact a sibling card owns: they are in the denominator because the window closed before the customer decided. Read this split on an aged cohort and most of them resolve.

Two win rates, and they are not the same number

From the same table:

• By count, the shop wins 21 of 56 sent estimates, 37.5 percent of the count.
• By value, it wins 39.9 of 168.0 units quoted, 23.8 percent of the value.

Both are shares of what was sent, so they are comparable, and the gap between them is not noise. It is the size skew stated as a number: whenever the value share sits below the count share, the wins are smaller than the average quote. The reverse gap, value share above count share, says you are landing the big ones and losing the filler.

Print both. A shop that reports only the count share and grows its quoting effort on large jobs will watch the count share hold steady while booked value goes nowhere, and have no figure that explains it.

Where the wall is: conversion by size band

The approved-against-rejected comparison says there is a size problem. Banding the same 56 sent estimates says where it starts.

The bands sum to the totals already stated, 56 sent and 21 approved, which is the first thing to check on any table like this.

Read the bands honestly. Every one of them sits under the 30-record floor a rate needs before it can be read as a rank, and the bottom row is 6 estimates, so 0.0 percent is an observation to extend, not a close rate. Roll two or three quarters together before you compare one band against another as a number. What one quarter can carry is the direction, and the direction here is clean: the rate falls at every step across four bands covering all 56 estimates. The size of any one step is not readable yet - 14 records against 14 will not carry a ratio - so pool the quarters before anybody quotes one band against another.

That shape locates the wall somewhere around 3 units for this shop, and the wall is where the decision changes. Below it the customer is buying a fix and the thing that wins is being reachable and being there; above it the customer is making a considered purchase and the things that win are the options presented, the terms, and whether anybody followed up. A price cut aimed at the whole book lowers margin on the 36 estimates in the two bands that were already closing well, to chase the 20 in the two bands that are not closing for reasons a price cut may not touch.

Cut the bands off your own sent distribution rather than at round numbers. Quartiles of one quarter's sent values give four bands with similar counts, which is what keeps the top band from being six records.

What changes the answer: which way your drafts skew

The example above has drafts running small, and there is a reason for it. The quick counter or phone quote is the one that gets started and abandoned when the customer says they will call back. The large job gets finished because somebody is chasing it.

The opposite shop exists, and the error flips with it. Where an estimator starts the big jobs, then stalls waiting on a site visit or a supplier price, the abandoned drafts are the large ones, and the reported average estimate value is inflated rather than depressed. That shop reads a rising average and congratulates itself on moving upmarket while its actual sent quotes hold still.

The tell is one calculation, and it takes a minute: compute the average value of drafts separately from the average value of sent estimates. If the two are close, the headline figure is harmless. If they are far apart in either direction, the headline figure is mostly telling you about your drafting habits.

There is also a degenerate case worth keeping in mind, because it exposes what the figure really measures. The fastest way to raise your average estimate value is to clear the abandoned drafts off the books. Delete all 24 and the reported average moves from 2.34 to 3.0 units without one thing changing about what you sell or what you charge. Finishing them instead moves nothing: a finished 0.8-unit quote is still an estimate raised at 0.8 units, and the average sits at 2.34 where it started. A number that a filing habit can move by a fifth is not a market signal.

Which figure sees which rows

Only the last two change a decision on their own. The first one changes a decision more often than any of them, and it should not.

References

• See related: Estimate Conversion Rate and the Cohort Problem - why the undecided column is there and how to age it out
• See related: Days From Estimate to Job and What a Long Tail Means - how long an undecided estimate is genuinely still live
• See related: The Close Rate Improved Because They Stopped Quoting - the same mix mechanism moving both of these figures at once
• See related: Pipeline Value Is Not a Forecast Until You Weight It - what an entered estimate value is worth before a quote exists