Designing IVR Qualification That Does Not Hang Up On Good Callers
Qualification gates raise billable rate — until they get long enough that qualified callers abandon. Where the line sits.
Every qualification question you add to an IVR does two things: it filters out callers who cannot be served, and it loses some fraction of callers who could have been.
The design problem is finding the point where the first effect still outweighs the second. That point is earlier than most people building these trees assume.
The abandonment curve is not linear
Abandonment through an IVR tree is not evenly distributed across questions. The first question loses very few people — the caller has just dialled and is invested. Somewhere between the third and fifth question, abandonment rises sharply, and it rises fastest among callers who are most qualified, because they have the strongest alternatives.
That last part is the counterintuitive bit. The homeowner with a $340 monthly utility bill has three installers competing for them. They will hang up on question five. The renter who does not qualify anyway has nothing better to do and will answer all seven.
Long trees can therefore make your qualified mix worse while your billable rate looks fine, because the denominator shrank.
Order by disqualification power
Ask the question that eliminates the most people first.
In solar, that is homeownership — it disqualifies a large share and takes four seconds. In debt relief, it is the debt threshold. In mass tort, it is diagnosis. Getting the highest-yield question first means most disqualified callers exit after one question, and the callers who continue are already substantially more likely to qualify.
The instinct to "warm up" with an easy question first costs you the whole efficiency of the tree.
Three or four questions, not seven
As a working rule, keep pre-transfer qualification to three or four questions and push the rest to the agent.
Anything that is genuinely binary and genuinely disqualifying belongs in the IVR. Anything that is nuanced, that benefits from a human reading tone, or that is really discovery rather than qualification belongs with the agent — where it is also doing double duty as rapport building.
The mass tort firm I worked with cut a nine-question tree to four and saw signed cases rise, despite qualified-call volume being flat. The five removed questions were doing screening the agents were better at anyway.
Write questions people can answer instantly
"Do you own the home where the system would be installed?" is answerable in under a second.
"Is your average monthly electricity expenditure across the last twelve months in excess of two hundred dollars?" is not, and the pause while the caller works it out is where abandonment happens.
Say the number, keep the sentence short, and accept a rough answer. You are filtering, not underwriting.
Give the caller a reason
A single sentence before the first question measurably reduces abandonment: "Two quick questions so we can connect you with the right specialist."
It converts the experience from an obstacle into a stated process with a visible end. It costs three seconds and it is the highest-return change in most trees.
Measure abandonment per question, not overall
Overall IVR abandonment is a number that tells you something is wrong without telling you where. Per-question abandonment tells you which one to rewrite.
If one question loses three times what its neighbours do, it is almost always phrasing rather than the criterion. Rewrite it before you consider dropping the gate.
The check to run quarterly
Pull the callers who abandoned mid-tree and sample what they would have answered based on the questions they did complete. If a meaningful share of them look like they would have qualified, your tree is too long — regardless of what the billable rate says.