How AI Turned 388 Phone Calls Into a Map of Real Demand
A home-remodeling company was generating hundreds of phone calls a month, but had no idea which were real leads, which sources drove them, or how many never got answered. TrolleyShield's AI listened to every one, and surfaced a costly blind spot no spreadsheet had caught.
The Problem
For a home-remodeling business, the phone is the funnel. Most leads call, they don't fill out forms. But a ringing phone is a black box: the owner could see that calls were coming in, not what they were worth.
Were they qualified buyers or price shoppers? Vendors and job seekers? Which marketing channels actually drove real projects versus noise? And critically, how many calls were slipping through unanswered? None of it was measured. Marketing decisions were being made on gut feel and a call log that showed numbers, not intent.
The Solution
TrolleyShield's Call Shield was connected to the company's call tracking. Every inbound call is transcribed and analyzed by AI in real time, classified by intent, scored for confidence, and tied back to the marketing source that drove it.
Over 10 weeks, the AI analyzed 388 calls. For the first time, the business could see its phone demand the way it sees its web analytics: by quality, by source, and by outcome.
What the Callers Actually Wanted
AI intent classification across every transcribed call separated real buyers from the noise, automatically.
38% of analyzed calls were genuinely qualified leads.
Just as importantly, the AI flagged the 18% that weren't customers at all — vendors, job seekers, wrong numbers, and spam — so the team could focus on the calls that could become revenue.
Which Sources Drove Real Leads
By tying intent back to attribution, TrolleyShield showed not just where calls came from, but which sources produced qualified ones. The share below is the qualified-lead rate within each channel.
The data also confirmed a tightly local market: 75% of callers were in-state and 67% were first-time callers, proof the marketing was reaching new, geographically relevant demand, not the same handful of repeat numbers.
The discovery
56% of calls were going unanswered.
The most valuable finding wasn't in the leads that were analyzed, it was in the ones that weren't. By scoring every call, the AI surfaced something the call log had buried: 219 of 388 calls — more than half of every lead the business paid to generate — were missed.
With 67% first-time callers, that's new demand ringing into the void. The owner couldn't fix a leak nobody had measured. TrolleyShield measured it, and turned a marketing analytics tool into a revenue-recovery roadmap.
Calls analyzed
388
over 10 weeks
Qualified leads found
38%
of analyzed calls
Non-customer noise flagged
18%
vendors, spam, wrong numbers
Missed-call leak surfaced
56%
How It Works
Every Call Captured
Inbound calls hit a tracking number that records the call and captures the marketing source, attribution, and caller details in real time.
Transcribed & Analyzed
Each call is transcribed and run through TrolleyShield's AI, which classifies intent across eight categories with a confidence score and reasoning.
Tied to the Source
Intent is linked back to the channel that drove the call, revealing which campaigns produce qualified leads versus vendors and noise.
Surfaced as Insight
Lead quality, source performance, and answer rates roll up into a clear picture, turning a black-box phone line into measurable marketing data.
You can't improve what you can't hear.
Now they can hear all of it.
Before TrolleyShield, the phone was the biggest, least-understood part of this company's marketing. After, every call is a data point: which are real leads, where they come from, and which ones are being missed. The same AI that scores leads exposed a 56% revenue leak, the kind of insight that pays for itself on the first recovered project.
Know what your phone calls are really worth.
TrolleyShield listens to every call and form submission with AI, so you know which leads are real, where they come from, and what to do next.
Get Started