Coaching Field Reps at Scale: What the Evidence Says, and What It Doesn't

If you are about to put a few hundred field reps in front of customers on a deadline, the honest starting point is that ramp is slower now than at any point in the last twenty years. The second honest point is that the research supporting AI roleplay, the thing every vendor will pitch you this quarter, does not exist yet in peer-reviewed form.

Both of those are inconvenient. Both are better to know before you sign than after.

This post is about what the evidence actually supports when you are onboarding or re-skilling a large field team fast, and where you are making a bet rather than a decision.

How long does it really take to ramp a rep?

The Bridge Group's 2026 study of 158 B2B companies puts AE ramp to full productivity at 6.2 months, the highest figure in that study's twenty-year history. Companies are also hiring for more experience than they used to: 3.7 years on average, up from 2.7 in 2022. SDR ramp is faster at 3.0 months, from a separate 2025 study of 351 companies.

Sit with the direction of travel. Reps are arriving more experienced and still taking longer to get productive. Whatever is making the job harder is not something you fix by hiring better.

There is a turnover number attached that changes the arithmetic. Median annual SDR attrition ran 40% in 2024, with an interquartile range of 21% to 57%. Of that, 13% was involuntary, 11% voluntary and 16% promotions. The promotion share is the one to watch: it was 34% in 2020 and has halved. Average SDR tenure is 1.9 years, the highest since the early 2010s.

So if you onboard 200 reps and the fastest of them takes months to reach full productivity, and a meaningful share leave inside two years, treat onboarding as a permanent function rather than a launch project. Build it that way or you will rebuild it every year.

Both figures come from observational, self-selected samples of engaged sales leadership, which the Bridge Group discloses. Treat them as good benchmarks, not laws.

What actually predicts whether a field rep performs?

The largest meta-analysis in the field coded 2,105 correlations across 79,747 salespeople in 4,317 organizations (Verbeke, Dietz and Verwaal, Journal of the Academy of Marketing Science, 2011). The significant predictors were selling-related knowledge, adaptiveness, role clarity, cognitive aptitude and work engagement. Personality traits did not make the significant set at all.

What follows from that table for a large field team, including one item nobody budgets for.

Knowledge is the top predictor and it is trainable. For a product launch this is the easiest win available. Reps need to know the product, the competitive alternatives, the objections and the data cold.

Adaptiveness is nearly as strong and it is not a classroom skill. Adaptive selling is what a rep does when the conversation goes somewhere the deck did not cover. You cannot lecture it into someone. It gets built through reps doing the thing and getting feedback close to the moment.

Role ambiguity is a drag almost as large as knowledge is a lift. A coefficient of -.25 against knowledge's .28. Unclear territory rules, shifting call plans, vague success criteria and a compensation plan nobody can explain will quietly eat most of what your training program adds. Fixing that costs no software.

Worth keeping in perspective: the whole field explains only about 10% to 20% of performance variance, and 5% to 15% when the outcome measured is objective sales. Anyone promising you certainty here is selling past the evidence. We wrote more about which sales frameworks survive that scrutiny in types of salespeople.

Does AI roleplay actually improve sales performance?

Nobody knows yet, and any vendor telling you otherwise is quoting their own marketing. There is no published peer-reviewed evidence that AI roleplay specifically improves sales outcomes, and the one field study on the question (Habel, Ahearne, Tirunillai and Vandaveer Novak, 2025) sits on SSRN as an unreviewed working paper whose sample, method and effect size we could not retrieve. None of that means skip AI roleplay. It means buying it as a bet with a measurement plan attached.

What should a launch readiness assessment actually measure?

Measure whether a rep can perform the conversation, not whether they can recall the content. A certification quiz on the clinical data proves knowledge, which is the top predictor, so it is worth having. It proves nothing about adaptiveness, which is nearly as strong. Score both, separately, and do it before the rep is in front of a customer.

A readiness bar we would defend for a large launch:

1. A knowledge check with a real pass mark. Not a completion checkbox. Someone should be able to fail it and get another attempt.
2. A recorded conversation scored against a rubric, with at least one moment where the scenario deliberately goes off-script. That is the only way to observe adaptiveness rather than assume it.
3. A named reviewer for every rep. Automated scores are useful for triage and terrible as the final word, especially in the first cycle when nobody has calibrated the rubric yet.
4. A clear statement of what "ready" means, in writing, before the assessment runs. This is the role-ambiguity fix hiding inside the readiness program, and it is free.
5. A re-check at 60 and 90 days. Ramp is 6.2 months. A readiness gate at week two tells you about week two.

Which field enablement platforms have AI coaching and readiness assessments built in?

Four vendors treat AI roleplay and readiness as core product rather than a feature: Mindtickle, Second Nature, Quantified.ai and Hyperbound. Gong also entered this space in 2026 with AI Trainer inside Gong Enable, launched February 25, 2026, and Dry Run, announced June 24, 2026. None of them publish list pricing.

Quantified.ai is the one built for a regulated field force. It runs a six-agent platform aimed at life sciences and regulated industries, including AI Roleplay, a Readiness Coach, a Field Coach and an MLR and on-label Compliance Agent. It names Novartis, Johnson & Johnson, Sanofi, Takeda and Bayer on its site. If your launch involves medical, legal and regulatory review and on-label constraints, the compliance agent is a bigger deal than the roleplay engine. We say that knowing it points at a competitor.

Mindtickle is the broad enablement platform: content, training, call scoring, AI Sales Role Play and an AI Role Play Simulator, plus an agentic layer called ElevateOS. It cites Cisco at a 31% increase in deal size, which is a vendor claim with no published method.

Second Nature is roleplay as the whole product. Video-avatar and multi-persona simulations, more than 30 languages, and SCORM or LTI packaging so it slots into an existing LMS. Its claims of 34% ramp reduction across 2,000-plus teams are self-reported.

Hyperbound splits practice from real-call scoring, covers 25-plus languages, and holds SOC 2 Type II and ISO 27001.

Funding figures for all four are unverifiable from their own sites, so discount any number you see quoted secondhand. The fuller category breakdown, including the tools that only review calls after the fact, is in WingRep vs. Gong.

Do leaderboards and missions actually help?

Honestly, we do not have good evidence either way, and we are not going to pretend otherwise. None of the sales-performance meta-analyses we trust measure gamification as a variable. What the research does say is that role clarity matters a lot, and a leaderboard is a statement about what counts. That is where its value or damage comes from.

So the useful question is not "should we gamify." It is what the board ranks.

Rank activity volume and you will get activity volume, including the calls nobody should have made. Rank something a rep controls and that maps to the actual job (discovery meetings that produced a documented next step, say) and the board becomes a clarity tool rather than a pressure tool. Missions and streaks are the same story. They direct attention. Point them at the wrong behavior and you have industrialized it.

One thing we would avoid: a public leaderboard on closed revenue during a launch, when territories have not stabilized. You are mostly ranking territory quality and telling half the team they are bad at their jobs for it.

What about coaching support reps rather than sales reps?

The performance research above is sales-specific, so do not import the coefficients wholesale. What transfers is the shape of the finding: knowledge and in-the-moment adaptiveness beat personality profiling, and unclear role expectations cost real performance. Support work has the same two variables under different names.

The practical difference is volume and channel. A support rep may handle dozens of interactions a day across chat, email and phone, which makes after-the-fact call review impractical as the main coaching loop. Sampling and live assistance fit that shape better than review does.

What this means if you are onboarding 200 reps next quarter

Put the money where the evidence is, and treat the rest as an experiment you are measuring.

Fund knowledge first. It is the strongest predictor, it is the most trainable, and for a launch it is also the cheapest thing to get right.

Fix role ambiguity before you buy software. Written territory rules, a clear definition of ready, a compensation plan someone can explain in a sentence. That is a negative coefficient of similar size to your biggest positive one, and it is fixed with a document.

Buy for adaptiveness, not for content delivery. Most enablement platforms are very good at distributing material. Fewer help a rep in the moment the conversation goes sideways, which is where the second-largest predictor lives.

Measure your own outcome. Since the AI roleplay evidence is not in yet, run it as a pilot with a holdout group and look at ramp time and early-cycle conversion rather than platform engagement metrics. Completion rates are not evidence.

Where does WingRep fit?

We built WingRep around the two predictors the research supports rather than around a training catalog. It prepares a rep before every call with what they need to know about that account and that stakeholder, which is the knowledge variable. It assists live during the call, which is the adaptiveness variable, and the reason we work in the moment instead of a week later in a review session.

Our per-seat price is published on our pricing page, and CRM sync and Gong integration are included. For a 200-rep field team, that is a published number you can compare against a quote-only enablement platform without running a procurement cycle to find out.

We are not an LMS and we do not run certification tracking or content management. If your launch needs formal MLR-reviewed content workflows, a platform built for that will serve you better and we would rather say so. See how it works for sales leaders, or look at the price.

Common questions

How long does it take to onboard a field sales rep?

Around 6.2 months to full productivity for an AE, according to the Bridge Group's 2026 study of 158 B2B companies, which is the highest figure in that study's twenty-year history. SDRs ramp faster at 3.0 months, from a 2025 study of 351 companies. Plan your launch timeline against months, not weeks.

Does AI roleplay improve sales performance?

There is no published peer-reviewed evidence that it does. A University of Houston working paper on the question exists on SSRN (Habel, Ahearne, Tirunillai and Vandaveer Novak, 2025) but has not been peer reviewed and its findings are not publicly verifiable.

What is a launch readiness assessment?

A check, before a rep talks to customers, that they can perform the conversation rather than just recall the content. A defensible version scores product knowledge with a real pass mark, scores a recorded conversation against a rubric including an off-script moment, and repeats at 60 and 90 days rather than gating once at week two.

Which enablement platforms are built for pharma field teams?

Quantified.ai targets life sciences and regulated industries directly, with an MLR and on-label compliance agent alongside its AI roleplay, readiness and field coaching agents. It names Novartis, Johnson & Johnson, Sanofi, Takeda and Bayer on its site. Mindtickle is the broader enablement platform to compare it against.

Do sales leaderboards work?

There is no meta-analytic evidence in sales research either way, so treat claims in both directions with suspicion. What the evidence does support is that role clarity predicts performance, and a leaderboard is a public statement about what counts. It helps when it ranks a behavior reps control and hurts when it ranks territory luck.

Sources

• The Bridge Group, 2026 AE Models, Motions & Metrics, 10th edition, published June 22, 2026, n = 158: bridgegroupinc.com
• The Bridge Group, 2025 SDR Models, Motions & Metrics, published February 6, 2025, n = 351: bridgegroupinc.com
• Verbeke, Dietz & Verwaal (2011), "Drivers of sales performance: a contemporary meta-analysis," Journal of the Academy of Marketing Science: doi.org
• Habel, Ahearne, Tirunillai & Vandaveer Novak (2025), Do AI Role Plays Improve Sales Performance?, SSRN working paper, not peer reviewed: doi.org
• Gong, Mission Andromeda press release (AI Trainer, February 25, 2026): gong.io
• Gong, Mission Big Dipper press release (Dry Run, June 24, 2026): gong.io
• Quantified.ai: quantified.ai
• Mindtickle: mindtickle.com
• Second Nature: secondnature.ai
• Hyperbound: hyperbound.ai