200 conversations to full performance
An online school. 7,679 conversations reviewed by artificial intelligence against a fourteen-step sales script. We measured how many conversations a rep goes through before matching an experienced colleague, and what that means for the team as a whole.
- 7,679 calls
- 99.9% coverage
- four months of continuous review
How it started
“This person is in their third month. Are they still ramping up — or is this their level?”
The question comes up in every sales team, and very concrete decisions hang on the answer: whether to keep investing in this person, whether to move them to another area, how to plan hiring for the next quarter. The problem is that a manager has almost nothing to lean on.
If the person is still ramping up, their current figures say nothing about their future ones, and comparing them with the team is premature. The months invested in them will pay off later.
If, on the other hand, they have reached their level, then waiting longer means salary, leads and a seat in the team are being spent on a result that will not change.
In practice the call is made on the manager’s experience and on the unwritten “three-month rule”, which as far as we know has never been tested against data.
To answer with a number, you need to know what a rep’s learning curve looks like. Neither the CRM nor the plan-versus-actual report contains it.
The client and the task
A sales team that doubled during the observation window
The client is an online school working with both inbound and outbound traffic. In November 2025 the company ran a mass hiring round: seventeen trainees joined sixteen full-time reps, doubling the team in a single month. By January the core team had grown to twenty-two people, and twenty-nine employees made calls in total.
Throughout that time, artificial intelligence reviewed practically every conversation.
That combination — a fast-growing team plus a full rather than sampled review of calls — is rare. It is exactly what made the rest of this possible.
The November 2025 hiring round
Continuous observation window
Share of conversations included in the analysis
The blind spot
Three familiar tools that do not show this
Plan versus actual
records the result but says nothing about how the conversation went. It cannot tell you at which step of the script the rep loses the customer.
Spot-checking calls
covers a couple of percent of calls, and those calls are usually picked by a known outcome — either especially good or especially problematic ones. A learning curve simply cannot be seen in such a sample.
The calendar view
a monthly report mixes two different things: the progress of individual people and changes in the team roster. When a team grows and turns over, monthly averages mostly describe hiring.
None of the three tools is bad in itself. None of them is simply built to answer this particular question.
Explainer · why we dropped the calendar
Why monthly averages show nothing
Picture two employees. Anna has been with the team since September and in January is having her five hundred and twelfth conversation. Igor started in January, and this is his thirty-first conversation. In the January report they land on the same line.
Their combined average comes to 14.6% — while Anna’s figure is 20.8% and Igor’s is 8.4%. The resulting number describes neither of them.
The faster a team grows, the more a monthly average tells you about the hiring mix rather than about learning.
We tested that claim against the client’s data. We took the twenty-four employees who were present both at the start and at the end of the period — the same people — and looked at how their figures changed by calendar:
| The same cohort, 24 people | Oct + Nov | Dec + Jan |
|---|---|---|
| Conversations brought to a customer decision | 20.1% | 20.5% |
That is why the axis in this case study is each employee’s personal call counter rather than the calendar. From here on, “21–50” means a specific person’s twenty-first to fiftieth conversations, regardless of which month they fell in.
How we measured it
Four steps from a conversation to a curve
Review
Every call is scored against a fourteen-step sales script checklist.
Markup
Recorded separately: whether the customer reached a decision, how many sales attempts there were, whether the objection was pinned down, whether it was handled, whether the conversation moved to payment, whether the close was performed.
Numbering
Each employee’s calls are sorted by time and given a sequence number.
Grouping
Figures are computed by number groups: 1 to 20, 21 to 50, 51 to 100, 101 to 200, and 201 and up.
Worth noting is how this body of data was assembled. All four steps run automatically, with no human involvement, and cover the entire flow of conversations. Listening to the same volume by hand — 7,679 calls over four months — would take roughly fifteen hundred hours of work.
Explainer · what exactly is measured
We measure behaviour in the conversation, not only its outcome
Whether a sale happens depends on many circumstances at once: lead quality, price, season, how well known the product is. The rep controls only one of them — how they run the conversation. That is why measuring learning by sales is incorrect: you are measuring a blend of the employee’s work and market conditions, and you cannot separate the two.
Behaviour in the conversation is what the employee chooses, and it is what changes as they learn. So the measurement is built on three levels.
What the rep did
Fourteen script steps. Did they uncover the need, pin down the objection, handle it, move to payment, perform the close.
How the conversation ended
Whether the customer’s decision was recorded, or the dialogue was left undefined, with neither a clear yes nor a clear no.
How persistently the rep ran the conversation
How many times in a single conversation they came back to the offer to buy. Among experienced employees this figure is noticeably higher.
One clarification: six of the fourteen steps are counted as shares — the ones for which the system returns a strict yes or no. For the rest it produces a written recommendation to the employee. That recommendation is useful in day-to-day work, but it cannot be added up into percentages, and we do not do so.
Explainer · the headline metric
“The customer reached a decision”: what it means and what it is not
Definition. The share of conversations in which the customer arrived at an explicit decision. The rep drove the dialogue to certainty instead of leaving it in “call me back in a week” limbo. A firm no counts as a decision here too.
This is not conversion to payment, not the share of sales and not revenue. Payment also depends on price, product and lead quality, whereas this metric describes only the conversation itself.
Why this metric. A conversation that never reached a decision is the least useful outcome a call can have. The lead is spent, the time is gone, and no certainty was gained: it is unknown whether to keep working with this customer or close the deal out. Such enquiries sit in the CRM without movement and create the illusion of a full pipeline. Bringing the other person to a clear answer is one of the first skills a rep acquires, and it barely depends on price or product. That is why we consider it the cleanest indicator of the employee’s own growth.
How to read the change
8.4% → 20.8%
The figure grows from 8.4% over the first twenty calls to 20.8% after the two hundredth. It can be written two ways: as +12.4 percentage points or as a two-and-a-half-fold increase.
Percentage points are the arithmetic difference between two shares. A multiple is the ratio of one share to the other. It is the same change, and we give both forms side by side so the number cannot be read more favourably than it is.
calls reviewed
over four months of observation. This is not a sample but 99.9% of all substantive conversations in the team, processed against a single set of criteria.
The finding
The learning curve exists, and it can be measured
- New hires — twenty-four people who joined the company during the observation window. For them, the first reviewed call really was their first.
- Experienced — twenty people who were in the team before the review started. They already had tenure behind them.
Among new hires the figure grows steadily across all five marks, without a single setback, and ends up roughly where their experienced colleagues had been all along. The convergence point is around 21%.
To be clear: this is the same period, one product, one script and one review algorithm. The only thing separating the two groups is accumulated experience.
Explainer · validity
Why this is learning and not a composition effect
Any chart that goes up has a simple alternative explanation: only those who stayed longer remain on the right-hand side of the sample, so the picture may reflect a change in who is being measured rather than growth in people. In statistics this is called survivorship bias, and it is better checked in advance than at the moment the question comes from the audience.
In this case that explanation does not hold, for the following reasons.
If the growth were explained by group composition, it would show up just as much among the experienced employees — their roster changed as well. Yet their line stays practically flat throughout.
If the growth were explained by the season, a product change or a quirk of the review algorithm, it would show up equally in both groups. It appears in only one.
Finally, if the growth were explained by the calendar, a cohort of the same employees would show it in a month-by-month view. As we saw, it shows no change at all.
So the flat line of experienced employees is not a backdrop but a proper control group. It is what turns the new-hire line into evidence.
Breakdown by script step
Every measured step grows, not one single technique
| Metric | calls 1–20 | 201 and up | Change |
|---|---|---|---|
| Customer reached a decision | 8.4% | 20.8% | +12.4 pp · ×2.5 |
| Closing | 7.8% | 18.8% | +11.0 pp · ×2.4 |
| Move to payment | 6.4% | 16.7% | +10.3 pp · ×2.6 |
| Objection pinned down | 19.3% | 28.0% | +8.7 pp · +45% |
| Objection handled | 14.3% | 18.0% | +3.7 pp · +26% |
| Sales attempts per conversation | 1.07 | 1.69 | +58% |
What changes is not one technique but behaviour in the conversation as a whole. The biggest growth is in the block responsible for finishing the deal: customer decision, closing and the move to payment — all three more than double.
This table is the most useful one here. It shows not just that “the employee got better”, but which skills form and in what order. A newcomer pins down objections almost as well as an experienced colleague — hearing an objection and probing it is learned quickly. Driving a conversation to payment and closing it properly, on the other hand, are the skills that take longest to build. That is where the onboarding programme is worth pointing.
is what a rep goes through before matching the results of experienced colleagues. At the workload this team runs, that is two to three and a half months of work.
In sales management this stretch is sometimes called the ramp, by analogy with the on-ramp a car uses to reach the speed of the main traffic. It means the interval between an employee’s start date and the moment their figures stop differing from those of experienced colleagues. Importantly, this is not a training period in the usual sense: training and mentoring may end in the first week, while reaching full performance continues for several more months. Throughout that time the employee works with real customers and real leads, and their result naturally differs from the team average. From here on we call this stretch the ramp-up period.
Explainer · where the number 200 comes from
Two hundred is not the moment learning stops
A new hire’s total climb is 12.4 percentage points: from 8.4% to 20.8%. Below is how much of that path is covered by each mark.
| By which call | Conversations with a decision | Share of the path |
|---|---|---|
| First 20 | 8.4% | start |
| By the 50th | 13.1% | 38% |
| By the 100th | 14.3% | 48% |
| By the 200th | 16.1% | 62% |
| After the 200th | 20.8% | 100% |
The fastest growth comes in the first fifty conversations: over a third of the whole climb happens there. After that the gains come in noticeably smaller steps.
The two-hundred mark is the point at which a new hire practically catches up with an experienced one: 20.8% against 21.7%.
What happens after the four hundredth or six hundredth conversation we cannot say: only a handful of people made that many calls within the observation window. The last group for new hires rests on 106 conversations — the thinnest point in the calculations, and we are pointing at it ourselves.
What it means for the team
The team’s figure depends on how many conversations new hires are running
So far this has been about an individual. Let us look at the same thing at the level of the whole team. In January 2026 the team ran 2,428 reviewed conversations. They were distributed as follows:
Almost two thirds of January’s conversations fell to employees who had not finished the ramp-up period. The monthly figure — 17.5% — came out of exactly that proportion: it sits closer to 15.2% than to 21.1% simply because there were more of the former conversations.
What it means for the team
The practical takeaway for a manager
A team’s monthly average reflects not only how well people work but also how the team is composed. In a month of active hiring it naturally drops, even if every individual is growing at that time. And the other way round: it can rise on its own if there was no hiring and everyone gained experience.
This explains a situation many managers know: the team report stands still while nobody is working worse than before. As long as the share of conversations run by new hires stays high, the team average will sit below the level the team is capable of.
The difference between 17.5% and 21.1% is 3.6 percentage points. That is exactly what the team leaves on the table in January purely because of how the workload is spread between employees with different tenure. This is not about recruiting or the quality of people: the gap appears in any growing team and disappears as employees gain experience.
There are two ways to manage this figure: shorten the ramp-up period, or plan hiring so that the share of new hires in the team stays below a certain level. Both are solvable — but only if the figure is measured.
What is discussed here is the share of conversations brought to a customer decision, not revenue. We will not translate this gap into money: that would require the company’s average order value and margin.
Business figures for the same period
Paid orders grew faster than the inbound flow
The data comes from the company’s own reporting for the same window. It is given in index form: October and November 2025 are taken as 100. Absolute amounts are not disclosed.
Conversion from a created order to a paid one rose from 65.5% to 80.6%.
Explainer · how to read these lines
What matters is not the growth rates themselves but which one outran which
The 28.8% growth in leads is a condition, not a result of the sales team. Marketing delivered the larger inbound flow. If every line below had grown by exactly the same 28.8%, that would mean nothing changed in how enquiries were handled.
Orders created added 10.0%, noticeably lagging the flow. There were almost a third more enquiries, yet only a tenth more orders were created from them. At this stage the team simply processed the increased volume.
Paid orders grew by 35.4% — substantially faster than the inbound flow. This is where a number appears that traffic cannot explain: out of the same number of created orders, noticeably more made it to payment. The difference between 10% growth and 35% growth is exactly the conversion that rose from 65.5% to 80.6%.
Revenue added 21.7%, lagging the number of payments. There were 35% more paid orders but only 22% more money coming in: the average value of a paid order fell by roughly a tenth.
There is something here for you. Growing the inbound flow by almost a third while keeping enquiry quality is a result in its own right, and it shows up in the reporting as a separate line.
Two features of the period should be stated plainly. December was an outlier: conversion reached 82.8% against October’s 62.1% — in December the company ran a New Year marathon. On top of that, conversion on the flagship products did not change at all over the period, staying at 62.1%. The growth was therefore spread unevenly across segments.
What we claim credit for, and what we do not
We did not make these employees better. We showed for the first time in what exactly, by how much, and over how many conversations they get stronger
No controlled experiment was run, and we do not claim the revenue growth. Over the same period the company changed its compensation system, introduced a team-lead structure and regular practice sessions, built up outbound sales, launched upsells and a sales bot, ran a New Year marathon — and the lead flow grew by almost a third.
What is more, our own data shows that in the calendar view the employees’ conversation behaviour did not change. It follows that the growth in order conversion did not come from it — and we are saying so ourselves, without waiting to be asked.
What does belong to our work: not one of the numbers in the ramp-up sections would exist without a full review of every conversation. Spot-checking covers a couple of percent of calls and is assembled around a known outcome, so a learning curve is fundamentally invisible in such a sample.
The scope of this case study
Everything above came from a single measurement layer. Three tools were never switched on
Daily personal reviews
Right now an employee goes through their two hundred conversations alone, by trial and error. Feedback on every call is the direct way to shorten that path.
Automatic training built for a specific skill
A session aimed at the script step that actually fails for this particular person, instead of a general programme for everyone.
A mobile app for the rep
The employee sees their metrics and their earnings daily rather than a month later at a stand-up. In the early stages this is the fastest way to move the result.
If you have already tried speech analytics and given up on it, this case study describes exactly the level of usage that usually results in practice. We consider it a starting point rather than the end state of a rollout.
The answer to the original question
“Third month — still ramping up, or already the ceiling?” Now the answer has numbers
The ramp-up period takes about two hundred conversations. For this team, two to three and a half months. An employee in their third month has most likely not reached their level yet, and comparing them with the team is premature.
The difference between the start and full performance is 12.4 percentage points. At team scale it turns into a 3.6-point gap between the actual and the achievable monthly figure.
Both quantities are measurable. Both the length of the ramp-up and its effect on the team are computed across the body of conversations. They can therefore be managed rather than discussed in general terms at a stand-up.
All of the above comes from the measurement layer alone. The tools that shorten the ramp-up period — personal reviews, training for a specific skill, an app for the employee — were not used in this case.
Final step
Test it on your own calls
Send us 10 recordings — we will run them through the base system and return the review for free, with no contract and no obligations. You will see what it sees in your conversations before making any decisions.
The free review runs on the base system, without custom configuration for your company. Calibrating the parameters to your script and playbook happens during rollout.