How we measured and improved the engineering firm's retail workflow

We followed the hours through to the finished store.

Their design automation saved time, but the margin per finished store hadn't improved. We therefore measured the work after the first drawing issue as part of a three-week diagnosis on one live retail program.

Interviews with leaders and delivery roles guided the study. Team members logged work and waiting time, and we checked those entries against timesheets. We also traced retailer requests through emails and meeting notes to see which became paid amendments.

To compare the live work with completed projects, we reviewed 18 months of finished retail jobs. We compared planned and earned fees, hours after first issue, revision rounds, and permit resubmittals. We separately checked use of their existing automation and the time spent reviewing its output. Together, these records showed where later work consumed the time saved in design.

We used existing records to catch changes and repeat drawing issues.

We connected their project and time records, revision and permit-comment logs, and the retailer-change inbox for the selected program. We reused their own automation where it fit, then added two steps around the losses the study identified.

The first flagged retailer requests in emails or meeting notes and drafted a change notice. This gave their project manager a prompt to decide whether to bill the change.

The second compared a drawing set with past permit comments for the relevant city and their quality checklist before licensed review. That brought previous review findings into the check before a sealing engineer received the set.

We tested the additions against their engineers' judgment.

Their senior engineers supplied a reference set of real examples showing correct decisions. We tested our additions on unseen examples against the baseline from the diagnosis, so the checks could be judged against their work.

We documented rules for AI use on client work and kept the professional decisions with their team. Project managers decided whether to bill a change. Licensed engineers reviewed and sealed the drawings.

The reported annual margin has two distinct sources.

The figures below follow our founder's account of the engagement and the results reported by the firm's leadership. The pilot measured a reduction in unpaid rework; the annual margin figure combines an estimate based on that reduction with billed changes.

• Pilot rework: unpaid rework fell by one-third, rounded to 33% for display. This covered revisions after first issue, repeat review comments, and permit resubmittals.
• Estimated annual benefit from less rework: $784K, calculated from the pilot reduction using the retail rework-hour share and a blended labor rate, applied across retail after rollout.
• Billed changes: $1.2M a year in priced and billed work the firm previously absorbed.
• Reported annual retail margin: $784K + $1.2M = $1.984M, rounded to ~$1.98M/year in the opening panel.

The $784K component estimates an annual benefit from the pilot result. It doesn't measure expenses removed over a completed year. The combined result is reported annual retail margin; it isn't an audited earnings figure or a measured net-profit result.

Our founder confirmed that the two benefit lines don't overlap. Revenue from redeploying the same freed hours isn't added as a separate gain.

They collected payment sooner in the following quarter.

Cash collection moved five days sooner on average in the following quarter. The measure weights each collection by invoice value, so a larger invoice contributes more to the average than a smaller one.

This measures how soon the firm collected money it was owed. It stays separate from revenue and the annual margin calculation. The five-day collection result also measures a different event from the delivery of sealed drawing sets.

The firm reported $1.1M in cash collected earlier, calculated from its observed receivables and the collection-time improvement. This is a working-capital timing result, not additional revenue or profit.

We transferred the tools and the method their team went on to use.

Their program team used both additions on live stores. We reported use and results weekly and helped project managers review pilot-store budgets each week, so they could see overruns while there was still time to act.

We transferred the tools and measurement method to their developers and research team, who took over maintenance. Their own team then ran the same discovery on mission-critical work without us leading it.