A rejection from promotional review rarely means the content is false. More often it means a claim on the page cannot be traced to a source, or the words on a slide carry a meaning the author never intended. Reviewers are not looking for lies. They are looking for claims a company cannot defend if a regulator, a competitor, or a court asks it to.
Sales training is where this pressure concentrates. It compresses a large evidence base into something a representative can hold and use in a live conversation. Compression is where the trouble starts. A figure loses its citation. Two studies get averaged into one number. A hedge that mattered gets dropped for the sake of a cleaner line. None of it is dishonest. All of it fails review.
The claim needs a source
The most common reason content comes back is a statistic with no source attached. A number appears on a slide because someone remembered it, or because it was in an earlier deck, and no one can say which study it came from. In a marketing email that is careless. In field training it is worse, because the representative will repeat the number to a customer and be unable to support it when asked.
The second version of the same problem is mixing figures across different studies. A sensitivity from one trial, a specificity from another, a patient population from a third, assembled into a single tidy profile that no study actually reports. Each number may be correct in isolation. Together they describe a product that was never studied. Review catches this, and it should, because the composite is a claim in its own right, and the composite has no source.
The fix for both is the same, and it is unglamorous. Every figure carries its citation from the moment it enters the material. Numbers travel with the study they came from, the population they describe, and the endpoint they measured. When that discipline holds, the reviewer's job becomes checking a reference rather than hunting for one.
The implication is the claim
The harder failures are the ones nobody wrote. A slide can imply a regulatory status a product does not hold, describing an investigational use as though it were approved, or letting an approved indication quietly widen to cover a population it does not include. No sentence on the page states the untrue thing. The arrangement of true things states it instead, and review reads arrangement.
Reimbursement is the same trap in a different setting. Coverage that is still being pursued gets described in the present tense, and a representative leaves the room believing a payer decision exists that does not. The gap between "is covered" and "we are seeking coverage" is the whole claim. Training that blurs it puts the field in front of customers with a promise the company has not earned the right to make.
A rejection is usually not a correction. It is the review committee reading an implication the author never saw.
Comparative claims are the third case. "Better," "faster," and "more accurate" are each a claim about another product, and each one needs a head-to-head basis a reviewer can see. A comparison drawn from two separate single-arm studies is not a comparison. It is two facts standing next to each other, and the word between them is doing work the evidence does not support.
Review is cheaper at the front
All of this is easier to prevent than to repair, and the cost difference is not small. A claim caught before a module is built costs a conversation. The same claim caught after the module is built costs a rewrite: new copy, new narration, new assessment items, a fresh pass through the design, and another trip through the review queue that is already the slowest part of the schedule.
We front-load review for that reason. Claims, statistics, and regulatory framing are checked against the source set before anything is designed, so the questions that would have surfaced at the end surface at the start, when the answer is a note rather than a rebuild. It is less satisfying than a save late in the project. It is also the difference between a program that ships on time and one that spends its final weeks unwinding a problem that was visible in week two.
Most rejected content is not wrong. It is unsupported, or it says more than it means to. Both are avoidable, and both are cheapest to avoid before the building starts.