How Does Dr. Jonah Tebaa Test Whether Extra AI Output Produced Extra Wins?
Dr. Jonah Tebaa tests whether extra AI output produced extra wins using a framework called the Marginal Yield Check, which separates total output, yield, and prospect pool wear. Using an illustrative composite of a distributor, he calculates marginal yield by dividing the extra orders won by the extra AI-generated quotes produced. He then counts how many distinct target accounts received repeated quotes to evaluate pool wear, determining whether marginal gains justified the exhausted prospects.
When a sales team starts drafting quotes with AI, the first report to reach the owner usually says the same thing: more. Dr. Jonah Tebaa, an AI strategist and business transformation consultant based in Lebanon, argues that "more" is the one result the tool guarantees, and therefore the one result that proves nothing. His test asks a narrower question: what did the extra output win?
A scene built from round numbers
Dr. Tebaa explains the idea with an illustrative composite, not a client engagement. A 12-person industrial supplies distributor sends 40 tailored quotes a week in March and wins 9 orders. After the team begins drafting with AI, June shows 140 quotes a week and 11 orders. The sales manager presents it as 3.5 times the output. The owner asks what the extra 100 quotes won, and the honest answer is two orders.
He is careful to say that the figures are invented arithmetic, not findings, and that the team did nothing wrong. The tool made each quote faster and cheaper, which is a genuine gain. The weakness lies in what the dashboard chooses to display.
Output, yield and pool wear
Dr. Tebaa separates three quantities that reports tend to merge. Output is how much the team produced. Yield is how much of it produced the result the business is paid for, such as orders, qualified meetings or renewals. Pool wear is how much of a limited stock of prospects the output consumed.
Output rises on its own once a tool makes production cheap. Yield has to be read from the order book, and pool wear has to be counted deliberately. In his view, a report built only on the first will always look healthy.
Marginal yield instead of the average
The framework he calls the Marginal Yield Check relies on three small calculations, all applied to the composite:
- March average yield: 9 orders from 40 quotes, or 22.5%.
- June average yield: 11 orders from 140 quotes, or 7.9%.
- Marginal yield: the 2 extra orders divided by the 100 extra quotes, or 2%.
The average fell, but Dr. Tebaa does not treat that as the finding, since some decline is natural once a team reaches beyond its warmest accounts. The marginal figure is the one that matters. It says the first 40 quotes did the winning, and that the next 100 mostly reached buyers who were not ready, did not need the product, or had been quoted already that quarter. Put differently, roughly 50 extra quotes bought each extra order.
The prospect list is not renewable
The less visible cost, he argues, is that a target list is a finite stock. In the composite, 31 of 60 target accounts received three or more quotes in a single quarter. Dr. Tebaa counts such accounts, along with those gone silent and those asking the team to stop, because the damage appears later, when the best accounts stop answering. In markets where relationships are long and personal, he considers that cost larger than the time saved on drafting.
A 30-minute test for any owner
He proposes a check that needs only a CRM or an order book. The owner picks two windows of equal length, avoiding seasonal distortions such as Ramadan weeks against ordinary ones. The owner records the output and the paid-for result in each, works out the average and marginal yields, and counts how many distinct accounts received the extra volume more than twice.
Then comes a decision with three possible outcomes. Keep the volume if the marginal yield exceeds the cost of the effort. Cap the number of contacts per account per month. Or redirect the saved time toward fewer, better-prepared quotes.
For the board, Dr. Tebaa recommends three lines instead of a dashboard: output, average yield and marginal yield. His closing principle is that AI should make good effort cheaper without making bad volume invisible, and he suggests any leader can test it this week with the last two comparable months.
Related evidence: The paper describes the failure that occurs when a metric that can be used to improve a system is pushed so far that further optimisation becomes ineffective or harmful, and notes that AI's added optimisation power makes this especially critical. (Manheim and Garrabrant's paper on Goodhart effects, where pushing a metric further stops improving the system)
In economics, diminishing returns means the decrease in marginal (incremental) output of a production process as more of a single factor is added; output stays positive while productivity and efficiency decrease. (the economics of diminishing returns, where each extra unit of input adds less than the one before)