The conversation log from your AI agent is the most useful document you are probably not reviewing before you approve or cut a broker portal lead source. Your CRM shows stage transitions. Your booking reports show contact rate and booked meetings. What neither of them shows is what actually happened in the thread — and that is where a new portal either earns its budget or burns it.
Here is what to look for, and what each signal tells you.
Response Rate Is the First Filter — and Most Portals Fail It
Before you analyze anything else, pull the total response rate for the lead source in question. Out of every lead the agent texted from this portal, how many replied at all?
A response rate below 20% on a new portal means one of three things: the portal is sending leads who submitted a vague inquiry and moved on, the candidates were not expecting to be contacted over SMS, or the portal is recycling aged leads under a new submission timestamp. All three are common. None of them improve with more budget.
If leads are not responding to a text sent in under 60 seconds, the problem is not your follow-up speed. The problem is lead quality. The logs prove it before you've committed to a quarter of spend.
What Candidates Are Actually Asking Tells You Who They Are
A response rate tells you how many people engaged. The content of those responses tells you who they are.
Open a random sample of threads from the new portal — ten to fifteen is enough for a directional read. Look at what the candidates are asking. Serious candidates ask about territory, unit economics, the FDD, Discovery Day logistics, and what existing franchisees say about the system. They ask follow-up questions. They come back when the agent follows up.
Candidates who submitted a form and moved on ask nothing, or ask questions that have nothing to do with the franchise — questions about jobs, customer inquiries, or responses that suggest they did not understand they were inquiring about a franchise opportunity.
The question content is the single fastest signal for whether a portal is sending franchise-ready buyers or filling a form quota.
Where Leads Drop Off in the Thread Pinpoints the Friction
Most conversation logs from a low-quality portal show the same pattern: the lead responds once, sometimes twice, and then goes silent before a meeting is booked. The agent follows up. No response. The lead does not advance in the CRM.
Map where the drop-off happens relative to the calendar offer. If leads are going dark before the agent ever offers times, the issue is engagement — these candidates were not serious enough to stay in a conversation. If leads are going dark after the agent offers times, the issue is either the times themselves (a minimum notice window set too tight, a buffer that leaves nothing available for a week) or the candidate's intent.
Those are two different problems with two different fixes. The log tells you which one you have.
"44% of franchise brands never send a personal reply to a new lead — just an automated acknowledgment, or nothing at all."
— FranFunnel Franchise Lead Response Time Study · 530 brands · fielded March–April 2025
Booking Rate by Source Is the Verdict — Not the Starting Point
Your custom reporting gives you booking rate by lead source. That number is the verdict. The conversation logs are the trial.
FranFunnel books 14.6% of the leads it texts when times are offered in the text thread — 1,535 booked appointments from 10,506 leads texted by the assistant, across select client accounts in 2025-2026. That figure includes every lead the assistant texted, not a filtered subset of engaged leads. If a specific broker portal is producing leads that book at a fraction of that rate, the logs will show you exactly why — before you spend another month of portal fees to confirm it.
When a portal rep pushes back on your decision to cut or reduce spend, the booking-rate-by-source report and the transcript sample are your documentation. A portal that books at 4% with consistent silence in the threads has a documented lead-quality problem. That is a different conversation than "we felt like the leads were not great."
Escalation Flags Signal When the Agent Could Not Do Its Job
Review any thread where the AI agent was unable to progress the lead and note the reason. Did the candidate ask a question the agent could not answer — detailed FDD questions, territory-specific legal questions, a question that required a rep's direct knowledge of the brand? That is not a failure. That is the agent doing its job and surfacing a lead worth a rep's attention.
The pattern to flag is something different: leads who engaged but whose conversations went sideways in ways that suggest they had no frame of reference for what they were inquiring about. Candidates who asked about prices as a consumer, candidates who expressed surprise that this was a franchise inquiry at all, candidates who responded with frustration about being contacted.
A handful of those in any batch is noise. A consistent pattern across a portal's leads is signal. It means the portal's intake form is not doing the qualification work it should, and the leads arriving in your pipeline are not leads — they are form submissions.
What to Do With What You Find
If the logs show solid response rates, engaged questions, and consistent progression toward a booked intro call, the portal earns a second quarter and a place in your approved lead source stack. If the logs show low response rates, surface-level or off-topic questions, and drop-off before the calendar offer, you have a documented case to cut or limit the source at renewal — without waiting for a full quarter of bookings data to prove it.
The reporting tells you the outcome. The logs tell you why. Review both before you approve or renew any new broker portal source, and make sure your stage-specific agents are configured to tag lead source so the data stays clean as you add sources over time.
See how FranFunnel tracks contact rate and booking rate by lead source, and pull conversation logs directly from your pipeline. Book a demo at franfunnel.com.