Picture a rep who just lost a deal. Their manager listens to the call and says, “you should have handled that price objection better.” The rep nods. Next week, on a different call, the same thing happens. Nobody lied. Nobody was lazy. The advice was even correct. And still nothing changed, because one sentence of feedback isn't proof of anything. It's a fragment, floating with nothing attached to it.
What a fragment actually is
A fragment is any piece of training handed to a rep with nothing connected before it or after it. A scorecard is a fragment. It tells you what happened, not why, and definitely not what to do about it. A roleplay session is a fragment. You practiced something, but nobody checked whether it was the right thing to practice, and nobody checked afterward whether it actually showed up on a real call. A one-off coaching call is a fragment for the same reason: real and useful in the room, gone the moment it ends.
None of these are bad on their own. The problem is that most sales training stops at exactly one of them and calls it a system. A rep gets a score. A rep gets a roleplay tool. A rep gets a coaching call once a quarter. Each piece is true and useful in the moment, and none of them talk to each other, so nothing compounds. You get the same conversation about the same weakness, dressed up as different advice each time.
The loop REPS actually runs
REPS doesn't start with a lesson plan. It starts with evidence: real uploaded calls, baseline assessments, roleplays, Skill Ranking sessions, drills. The actual record of how a rep performs, not a guess about what they probably need. From there, the loop runs in five stages, and skipping any one of them is how you end up back at a fragment.
- Establish Your Real Performance: your real calls come in and REPS pulls the evidence out of them. What was actually yours gets separated from what was never in your hands, so you are not training somebody else's problem.
- Find Your Biggest Bottleneck: ten things could be better. That does not mean all ten deserve Tuesday morning. REPS narrows it to the one costing you the most right now, and everything else waits its turn.
- Train Until You Can Execute: you see it performed first, then attempt it with support, then the support is withdrawn and it is just you against a buyer who keeps moving the goalposts.
- Prove It On Real Calls: you return to selling. REPS lines the trained skill up against the calls you ran afterward and settles the only question that pays you: did it hold under pressure?
- Fix The Next Bottleneck: it stuck, so you move on to what is costing you most now. It came back, so you train it again. Either way the loop turns, because performance never stays still for long.
“Here's your score.”
Here's the score, the specific moment that caused it, and a tested theory of why.
“Practice more roleplays.”
Train the one weakness that's actually costing you deals, until you can perform it alone.
“Great job, 94/100.”
Did it hold up on your next real call? That's the only question that counts.
“You passed the roleplay.”
Prove it against a harder buyer, then a different one, before it counts as learned.
Why proving it on real calls is the stage almost everyone skips
Evidence is easy to collect, call recording software has done that for years. A rough diagnosis isn't much harder, plenty of tools will tell you a rep is weak at discovery. Training is where most products stop, because a roleplay demo looks impressive in a sales meeting. The stage almost nobody does is the fourth one: going back to the rep's real calls and checking whether the trained skill actually survives a real buyer, real pressure, days or weeks later. That's the REPS Standard: training completed is not skill proven. A skill that only works in practice was never actually learned. It was memorized for the test.
This is also why the loop doesn't stop once a skill is proven. The moment one weakness is proven fixed, it stops being the priority and something else becomes the bottleneck. There's always a next thing standing between a rep and a deal they should have won. A system that closes the loop once and stops is still just a longer fragment. Fixing the next bottleneck only means anything if the loop keeps running.
None of this requires a new theory of selling. It requires refusing to stop at the fragment: refusing to hand someone a score and call it done, refusing to let a roleplay pass for proof. That's the whole difference, and it's bigger than it sounds from the outside.