How DFC Home Improvement Cut Estimate Time by 80% with AI Software

DFC Home Improvement is a roofing and exterior contractor based in the Mid-Atlantic region. They do residential roofing, siding, gutters, and storm damage restoration — the kind of work where being first to the job site, and first to get the estimate in the client's hands, often decides who wins the job.

Before Ontrakt, the estimating process looked like this: drive to the site, take photos on a phone, come back to the office, pull up a spreadsheet, manually enter every line item, do the math, format a PDF, email it out. Two to three hours per estimate. For a company doing 30-40 estimates per month, that's 60-90 hours of office time — nearly two full work weeks — just on estimating.

Here's what changed.

---

The Problem: Estimating Was the Bottleneck

DFC generates leads from Thumbtack, Google, and word of mouth. During storm season, lead volume spikes — sometimes 15-20 new inquiries in a single week. Each one needs a site visit and estimate.

The bottleneck wasn't the site visits. It was everything after: organizing the photos, building the estimate, formatting the PDF, following up. The estimator was spending more time on paperwork than on the jobs themselves.

"We were winning less than 30% of our estimates," said Ezra Sopher, founder of DFC. "Part of it was pricing, but a lot of it was timing. By the time we got the estimate out, the client had already gotten two other quotes and made a decision."

There were also consistency issues. Different estimators would scope similar jobs differently. A redecking job might include edge metal on one estimate and not another. Line items were inconsistent. When a client called to ask questions, pulling up the original estimate and its supporting photos wasn't always easy.

---

The Solution: AI Estimating from Job Site Photos

DFC moved to Ontrakt in January 2026. The workflow changed immediately: Before: Site visit → photos on phone → back to office → manual estimate in Excel → format PDF → email (2-3 hours) After: Site visit → upload photos to Ontrakt from truck → AI generates estimate in under 60 seconds → review and adjust on phone → send directly to client (15-20 minutes)

The AI analyzes each photo for material type (shingle brand, gauge of siding, gutter profile), estimates dimensions from visual cues, identifies scope items (flashing replacement, decking damage, fascia condition), and builds a complete line-item estimate with labor and materials.

"The AI is right maybe 85-90% of the time on the line items," Ezra says. "I still review everything. But I'm reviewing and adjusting, not building from scratch. That's a completely different amount of work."

---

The Results: Before and After

Estimate turnaround time

| Metric | Before | After |

|---|---|---|

| Time per estimate (build + format + send) | 2.5 hours avg | 25 minutes avg |

| Estimates submitted same day as site visit | ~40% | ~95% |

| Estimates submitted within 1 hour of site visit | 0% | ~60% |

The single biggest change was same-day submittals. Before, estimates often went out the next morning or the day after. Now, DFC typically sends the estimate from the parking lot before driving to the next appointment.

Close rate

| Month | Estimates sent | Jobs won | Close rate |

|---|---|---|---|

| Oct 2025 (pre-Ontrakt) | 31 | 9 | 29% |

| Nov 2025 (pre-Ontrakt) | 28 | 8 | 28.5% |

| Jan 2026 (Ontrakt, month 1) | 34 | 13 | 38% |

| Feb 2026 (Ontrakt, month 2) | 38 | 17 | 44.7% |

Close rate went from 29% to 44.7% over the first two months. Some of this is speed (first estimate in wins more often). Some is follow-up — Ontrakt automatically sends follow-up messages on Day 2 and Day 5 after an estimate goes out. Before, that follow-up happened when there was time, which often meant it didn't happen.

Revenue impact

DFC's average ticket is roughly $8,500 for a residential roofing job. Going from 29% to 44.7% close rate on 35+ estimates per month means approximately 5-6 additional jobs per month closed.

At $8,500 per job: +$42,500-51,000 additional monthly revenue from the same lead volume.

That's without increasing marketing spend, without hiring another salesperson, without more leads. Just closing a higher percentage of existing quotes — because the estimate went out faster and follow-up happened automatically.

---

The Features That Made the Difference

1. AI photo estimating

The core feature. Photos go in, line items come out. DFC uses it on every job now — roofing, siding, gutters, and storm damage restoration. The AI handles Xactimate-style categorization for insurance jobs automatically.

2. Automated follow-up sequences

Every sent estimate automatically queues a two-message follow-up: one at 48 hours, one at 5 days. DFC doesn't have to think about it. If the client responds, the sequence stops. If they don't, it keeps going.

"Before, I'd tell myself I'd follow up and then forget for a week," Ezra says. "Now it just happens."

3. Client portal and e-signature

Clients review and accept the estimate through a portal — no phone tag required. When they accept, the signed acceptance is logged and a construction contract is auto-generated and sent for e-signature. The entire paper trail — estimate, acceptance, contract — lives in one place.

4. Lead scoring and inbox

Thumbtack and web form leads go directly into the Ontrakt inbox. Each lead is scored by urgency, job size, and description detail, so DFC prioritizes "emergency" jobs (hail damage, active leak) above routine quote requests. That triage used to happen mentally, inconsistently. Now it's automatic.

---

What Took Getting Used To

The AI doesn't know everything. On complex custom jobs — unusual materials, complex multi-plane roofs, combination projects — the initial estimate requires more editing. It's still faster than starting from zero, but DFC's team learned to flag those jobs for a more thorough review rather than relying on the 90% accuracy threshold.

The first week was also a learning curve. Learning which photos give the AI better results (overhead shots for square footage, close-ups for material identification, wide shots for scope), and how to add annotations when something isn't obvious from the photos.

By week three, it was second nature.

---

The Bottom Line

For DFC, switching from manual Excel estimating to Ontrakt's AI workflow recovered roughly 60 hours per month of estimating time and added $42,000+ in monthly revenue from improved close rates and faster follow-up.

The math is simple: if you're submitting 20+ estimates per month and closing less than 40% of them, faster response time and automated follow-up will move the needle. The technology is available now, and it's not complicated to use.

DFC is now applying the same workflow to insurance restoration claims — using Ontrakt's Xactimate-format output for storm damage estimates, which gets submitted directly to adjusters without reformatting.

--- Want to see if Ontrakt would have a similar impact on your business? Apply for beta access → DFC Home Improvement operates in Maryland, Virginia, and DC. Roofing, siding, gutters, and storm damage restoration.