Ask people whether they would buy something and a reliable share say yes. Ask them to actually buy it and a much smaller share do. This is not because people lie. It is because a survey removes everything that makes buying hard.
In a survey there is no price comparison, no shelf, no competing product they already trust, no forgetting, no running out of time. Every one of those exists in the real purchase, and every one of them removes buyers.
The mistake is not running concept tests. They are useful for what they measure: whether the proposition is comprehensible and interesting. The mistake is treating the resulting percentage as a forecast, then building an inventory plan on top of it.
What to do instead is unglamorous. Prefer observed behavior to reported intent wherever you can get it, and where you cannot, label the number as what it is so the next person reading the deck does not mistake it for demand.