How we built a micro-SaaS estimator to automate quotes for a blue-collar business (and boosted conversions by 35%)

We replaced a three-day manual quoting process with a five-question web form that returns a price range in under a minute, and quote-to-job conversion went up 35% in the first quarter.

Not 35 percentage points. Relative: 11.2% of quote requests turned into signed jobs before, 15.1% after, on almost identical traffic.

The product is a single-purpose estimator for a small excavation and foundation crew. It costs about $18 a month to run, and it took six weekends. Here is the whole build, including the parts that flopped.

What the estimator actually does

It takes five answers from a homeowner and hands back a price range plus a booked site visit. That is the entire scope. No takeoff, no blueprints, no login, no dashboard for the customer.

Someone lands on the contractor's website, taps "get a ballpark", and answers five questions: project type, rough square footage, site access, soil condition, zip code. About forty seconds later they see a range, something like $9,400 to $12,800.

Then the form asks whether they want a firm number and offers three site visit slots. Behind that, a rules engine prices the job off the contractor's own cost book, an SMS hits the owner's phone in under thirty seconds, and a two-page PDF lands in the homeowner's inbox with every assumption written out.

Why quoting is the bottleneck in a blue-collar business

Because the person who prices the work is the same person running the excavator. That is the problem in one sentence.

In software, a quote is a pricing page. In site work, a quote is the owner in his truck at 8:40 pm with a legal pad, a supplier price sheet, and eleven voicemails.

Every hour on estimates is an hour off a machine, so estimates wait. While they wait, the homeowner dials the next three numbers on Google.

Ask any foundation contractor in Lancaster, PA and the shape is the same: thirty to fifty inbound requests a month, roughly half get a real quote, and the ones that do get it two or three days late.

The lead response research has been saying the same thing for a decade. Speed beats price, beats reviews, beats your website. Most trades still answer in hours or the next morning. The bar is embarrassingly low — which is exactly why a tiny tool can move a real number here.

The build: five questions and a rules engine

Version one was a JSON price book and a form. No AI, no CRM, no integrations.

The stack, boring on purpose:

• Next.js on Vercel for the form and a one-line embed script
• Postgres on Supabase for leads and the pricing table
• Twilio for the instant SMS to the owner and a confirmation text to the homeowner
• Resend for the PDF estimate email
• One serverless function that renders the PDF and writes the lead row

The code was never the hard part. The hard part was two evenings at a kitchen table turning what the owner does in his head into arithmetic.

We got it down to four multipliers: access (can a truck reach the pad), soil (clay, rock, or sandy loam), haul-off (does spoil leave the site), and compaction.

We borrowed the scoping order from site layout and planning specialist "Site Prep" and copied that split straight into the form.

Why a range and not a number

Because a number you cannot honor is worse than no number at all.

The width of the range carries the uncertainty. Known soil and easy access → the range tightens. Unknown soil → it widens, and the PDF says why in plain English.

The homeowner gets transparency, the contractor keeps his margin, and nobody argues on day one of the dig. The range also filters. If someone was picturing $3,000 and the tool says $11,000, they leave. That is a win.

What actually moved the conversion rate

Three things, in this order: response speed, price transparency, and cutting the form from fourteen questions to five.

Traffic barely moved. This was never a marketing play — it was a leak in the bucket. Half the requests used to die unquoted because a human ran out of evening.

The counterintuitive part

Ungating the price cost us about a third of our email addresses and won us more booked site visits.

The first version held the range hostage behind an email field. When we showed the range first and asked for contact details after, fewer people submitted and far more of them picked a visit slot.

We were optimizing for a list. The contractor needed a calendar.

What did not work

Four experiments, all dead and buried:

• AI takeoff from satellite imagery: Measured the driveway fine and missed a septic field — off by 40% on one job (in this trade, that’s not a bug, it’s a lawsuit).
• The fourteen-question "accurate" estimator: 71% drop-off by question six.
• A homeowner portal with project updates: Nobody logged in twice.
• Deposit collection inside the form: Excavation is not a checkout.

Is this worth building as a micro-SaaS?

Yes, but only if you have one real operator to build it with.

We now license it to six other contractors at $79 a month (\~$500 MRR). Nearly zero churn, because the thing is welded to their lead flow. Trades don’t buy software — they buy booked jobs.

Before you build one, check that:

• The operator is losing jobs to response time, not to price
• The pricing logic collapses to fewer than ten variables
• The owner will actually sit with you for two evenings, no exceptions
• There is existing traffic to convert

What I would do differently

I would ship the SMS auto-reply first, alone, in a single day, and watch the number move before writing a line of pricing logic.

Half the lift lives in the first thirty seconds after someone taps submit. The calculator is what makes the story sound clever. The text message is what actually beat the other guy to the phone.