A specialist glazier: quote-to-order lag down from two weeks to one day
A specialist glazier in Norfolk was taking two weeks to turn a measured survey into a quote. By the time the quote arrived, the customer had often moved on.
The reading
A specialist glazing contractor on the north Norfolk coast, serving country houses, listed buildings, and the occasional commercial project. Eighteen employees, two directors, one estimator who had done the firm's quotes for fourteen years. Jobs typically began with a site survey, which fed into a quote, which fed into an order.
The survey took a day; the quote took a fortnight. The estimator produced quotes by taking the surveyor's hand-drawn sheets, cross-referencing supplier price lists held in three ring binders, calculating glass sizes, waste, and fitting hours, and typing the quote into Word.
The directors had measured, over a quarter, that the firm lost about one quote in three to a competitor who answered faster. The lost quotes were, disproportionately, the larger jobs.
Fourteen years of ring binders
The estimator was not slow. He was accurate. The firm's quotes were widely respected and almost never had to be revised downward in execution. The problem was that the inputs to his work were in paper form, the price lists updated monthly on paper, and the calculation was done by hand because the previous attempt at a spreadsheet had been found to produce wrong answers in edge cases.
The directors' instinct had been to hire a second estimator. They had been unable to find one who met the standard.
- 1 estimator, 14 years in post, 3 ring binders of pricing
- Average 11 working days from survey to quote
- Approximately 1 in 3 quotes lost to faster competitors
- Larger jobs disproportionately represented in the losses
- No digital record of past quotes for reference
What we built
We digitised the price lists (which took two weeks and the estimator's active cooperation), modelled the firm's standard calculation methodology, and built a quoting application that the surveyor used on an iPad on site. The surveyor entered dimensions and glass specifications as they measured; by the time they returned to the office, a draft quote was already generated.
The technical choice worth noting: the estimator remained the signoff. The surveyor's iPad produced a draft; the estimator reviewed it, applied his judgement for the awkward cases (a listed window with a difficult beading detail, say), and issued the quote. He went from spending ten days per quote to spending about forty minutes.
Stack: a React Native application for the surveyor, a Rails backend with a rules engine for the pricing, Postgres for the catalogue, PDF generation for the quote itself.
The numbers at ninety days
| What | Before | After |
|---|---|---|
| Survey-to-quote lag | 11 days | 1 day |
| Quote win rate | 65% | 83% |
| Estimator time per quote | 10 days | 40 min |
| Average quote size (won) | £8,400 | £11,200 |
| Past quote reference time | 20 min | 5 sec |
The judgement is scarce
The estimator's skill was never the bottleneck anyone thought it was. The bottleneck was the paper around his skill. Remove the paper, and his judgement, which was the actual asset, became available for many more quotes per month.
This is a specific version of a general observation: experienced people are usually faster than their systems allow them to be. The system-first move is to find out which parts of their day are their skill and which parts are the paper around their skill. We try to remove the paper and leave the skill.