How Does Dr. Jonah Tebaa Audit the Enquiries a Sales AI Turned Away?
Dr. Jonah Tebaa audits rejected enquiries using a five-step framework called the Declined-Pile Audit. Across a 60-day period, a salesperson blindly reviews 40 evenly sampled declined conversations, classifying each as a correct decline, an undecided case, or an erroneous refusal categorized across four specific causes. In an illustrative composite case, wrong refusals cost roughly $400 each versus $10 for a wrong approval, prompting targeted adjustments such as adding a single question, introducing a second-look lane, and scheduling monthly re-audits.
Most sales dashboards describe the people who got through the door. Dr. Jonah Tebaa argues that the more important group is the one that did not: the enquiries a website assistant turned away before any salesperson saw them. His recent article proposes a one-afternoon audit of that group, built around a labelled composite, and a way to weigh what a wrong refusal costs against what a wrong approval costs.
Why does a wrong refusal go unnoticed?
According to Dr. Tebaa, the two kinds of error behave differently. A wrong approval is visible. A salesperson gives an hour to someone who was never going to buy, and that hour appears on a calendar. A wrong refusal appears nowhere. The person never becomes a lead record, so no lost-deal reason is written, and nobody complains, because they simply contact a competitor.
He compares this to survivorship: reports study the enquiries that passed, so the filter always looks sound. He does not argue against filtering. A polite refusal of an unsuitable job is useful, in his view. His concern is that a decision no one reviews receives no feedback, and a filter without feedback drifts.
What is the composite?
Dr. Tebaa opens with a fit-out contractor, an illustrative composite and not a real client, with every figure invented. In one month 410 enquiries reach the assistant, 292 are passed to sales and 118 are declined with a line about focusing on larger projects. The owner reads thirty of the declined conversations on a Thursday afternoon and judges 21 correct and 9 wrong. He is explicit that none of these numbers is a statistic.
How does the Declined-Pile Audit work?
He calls the method the Declined-Pile Audit and sets it out in five steps.
- Define "declined" broadly, to include any enquiry that no human answered, and pull the last 60 days.
- Read everything if the pile is under 40. Otherwise read 40 chosen evenly across the period, avoiding both the first few and the interesting ones.
- Ask someone who sells to read blind and mark each as a right decline, a wrong decline, or one that cannot be told, with a line of reasoning for each wrong one.
- Sort the wrong ones by cause and put a price on them.
- Change a single rule or add a single question, schedule a re-audit in 30 days, and name the number that should move.
He cautions that thirty conversations is a thin sample. The resulting share should be read as a rough size and tightened by later audits.
What causes a wrong refusal?
Dr. Tebaa groups the causes into four.
- A rule too blunt, where a threshold is applied to the headline and ignores context in the same message.
- A field missing, where the assistant never asked the question that mattered.
- A phase hidden, where the first job is small and the account is large.
- A source unseen, where a referral or repeat client is handled like a stranger.
In his composite, the nine wrong refusals split four, two, two and one across hidden phase, unseen source, missing field and blunt rule. The grouping matters because each cause points to a different repair, and none requires the filter to be loosened.
How does he weigh the two mistakes?
The calculation, which he labels as illustrative, begins with the sample. Nine wrong out of thirty is 30%. Applied to 118 declines, that gives about 35 wrong refusals a month. With an assumed 10% close rate that is about 3.5 jobs, and with an assumed margin of $4,000 each, roughly $14,000 of monthly margin.
He then prices the opposite error. Suppose 100 of the 292 passed enquiries went nowhere, each costing a salesperson 20 minutes at $30 an hour. That is about 33 hours, or $1,000 a month. Per enquiry the contrast is sharper: a wrong refusal is worth about $400 on average, a wrong approval about $10, so roughly 40 wrong approvals equal one wrong refusal. Dr. Tebaa also tests the result by halving the error rate to 15%, which still leaves about $7,000 a month against $1,000. He stresses that readers should replace each number with their own, and never borrow a benchmark.
What does he recommend doing about it?
In the composite the owner leaves the filter in place. She adds one question about whether the job is the first phase of something larger, and a second-look lane in which refusals carrying a referral or repeat-client flag are glanced at by a person before the conversation closes. Dr. Tebaa adds an honest caveat: some refusals are wrong only in hindsight, so the reader should judge on what the conversation showed, and treat the rest as undecided.
The owner now reads thirty declined conversations on the first Thursday of each month. His closing line is that a filter nobody audits is a policy nobody has read. Readers who want the full arithmetic and his own wording can find it in Dr. Tebaa's article on the sales pile nobody reads, in his own words.
Related evidence: A qualifier tuned for precision (fewer wasted sales hours) can quietly lose recall (good buyers declined); reading a sample of the declines is how you measure recall. (recall, the share of real buyers the assistant actually let through)
The weekly report only shows the enquiries that passed the filter, so the declined ones never enter anyone's judgement of whether the rule works. (survivorship bias, judging a filter only by what passed it)