Turning a quote form into an operational lead pipeline

The owner's request was reasonable: a more commercial-looking site and a better quote form. The paid-search freelancer who made the introduction had heard a different complaint. The ads produced leads, but the office called them inconsistent.

Discovery traced that complaint. Future Built interviewed the owner, operations manager, office coordinator and estimator, compared 60 days of web inquiries against their matching Jobber records, and watched the coordinator work five incoming leads.

The form asked for a name, an email and a message. It captured nothing structured about property type, where the work was, which service was needed, or whether the buyer was a homeowner or a property manager. Each submission landed in a shared inbox as an email, not a request. The coordinator read it, judged whether it fit the service area, called or emailed back for what was missing, then re-typed it into Jobber. Leads that arrived after hours often waited until late the next workday.

Of 126 submissions a month, 84 fit the service area and service mix. Median staffed first response was 2h 47m, and intake work consumed 10.9 office hours a month. A successful form submission still did not produce an operationally usable lead.

Discovery weighed four options. Embedding Jobber's own request form was cheapest and operationally acceptable, but it limited conditional qualification, landing-page control and analytics. Connecting the existing WordPress form through a generic automation service was fast, but added a sync layer to maintain and kept a site the client already wanted replaced. A custom lead or CRM application would have rebuilt what Jobber already does.

Future Built chose the fourth: rebuild the conversion surface and integrate Jobber directly. The website owns landing pages, qualification, conversion UX and measurement. Jobber keeps clients, requests, quotes and jobs, with no configuration overhaul. Just as deliberate is what was left out: no CRM, estimator, scheduling product, customer portal, payment processing or custom CMS. Own what creates differentiation. Integrate what does not.

The new intake asks one question at a time across six steps, starting with the question the coordinator used to answer by hand: what kind of property is it? Residential, HOA/multifamily and commercial each open their own service list, so a facilities manager never scrolls past roof soft-washing. The visitor then gives the service location, which is checked against a deterministic ZIP rule. Out-of-area inquiries get a clear decline and referral path, and no request is created. In-area visitors add contact details, a company name where required, requested timing and an optional note. Missing fields are fixed in the form, not discovered on a callback.

Two parts came from testing, not the first draft. The prototype, clickable in week one, used a single property-address field. Staff testing exposed the flaw: a commercial requester often works somewhere other than the property being cleaned. The schema was split into contact, company and service location before content migration, which also kept address-based client matching from merging unrelated requesters at one property.

The second surfaced at launch. Commercial paid-search visitors kept choosing "Other" because the original services were written for homeowners. Instead of a longer generic list, the form now asks for property class first and scopes the services to it.

The site runs on Next.js and TypeScript on Vercel. The browser and server share a Zod schema, but every submission is revalidated server-side, including the service-area check, before anything reaches Jobber. A valid submission is mapped across the boundary (contact and company to a Jobber client, service location to a property) and created through Jobber's GraphQL API. The OAuth token stays on the server, and client creation records the website as the lead source.

There is no application database. Jobber already holds the durable record, and the site has nothing of its own that needs to persist. Traceability comes from a submission_id, a UUID that travels through structured logs, analytics and the note on the Jobber request.

If the Jobber call fails on a timeout, server error or rate limit, the endpoint retries with backoff, checking for that submission ID first so a retry cannot create a duplicate. If Jobber still fails, the system raises an integration alert and sends the office the complete request with its submission ID. The visitor receives a submission confirmation rather than an infrastructure error, while the coordinator completes intake from the fallback notification. An inquiry can arrive late; it cannot vanish silently.

Analytics events mark each stage, from form start and qualification branch through submit success or integration failure, so form friction and integration health are measured separately.

Results come from a constructed eight-week post-launch scenario, ordered by how much the new system can claim.

The strongest measure the workflow directly. Median staffed first response, from website timestamp to first logged contact, fell from 2h 47m to 26m, an 84% reduction; an after-hours inquiry now waits in Jobber with the office already notified. Office intake, re-keying and chasing fell from 10.9 to 3.2 hours a month. And 97% of web-qualified leads entered Jobber without office re-keying; the rest were integration exceptions or intentional manual handling.

Form start-to-submit rose from 62.1% to 74.6%. Attribution for that 12.5-point gain is medium: new landing pages launched at the same time.

Qualified lead to booked estimate moved from 31.8% to 36.9%. The 5.1-point difference appears in the cohort, but seasonality and lead mix changed too, so it is not presented as a conversion gain the site caused.

The economics stay separated too. The 7.7 released hours, at a modeled $30 loaded rate, represent $2,772 a year of office capacity, not a payroll saving. A secondary opportunity model based on the booking movement adds $15,299 a year. Together, $18,071 implies roughly 4.5-month payback on the $6,800 engagement, or about 29 months if only the capacity model holds. None of these figures is measured revenue.