In evaluating opportunities, it’s common to designate one option as the primary and another as the backup, as though the designation itself confers reliability. The primary is where the real work goes — the close examination, the honest accounting of weaknesses, the multi-bar test. The backup sits quietly in reserve, examined just enough to be named but not enough to be verified. It feels like insurance. It usually isn’t.

The problem is that a backup that hasn’t been held to the same scrutiny as the primary isn’t a real option — it’s an assumption. You know it exists. You know it has at least one promising attribute (that’s what got it named as backup). But you don’t know whether it clears the full bar, because you didn’t hold it there. What you actually have is a hypothesis that a fallback exists, not a fallback. The comfort it provides is real; the security it provides is not.

The pattern tends to surface slowly. The primary gets examined in depth. Several rounds of research surface its strengths, confirm its weaknesses, and leave you with a clear picture of whether it works. The backup accumulates no comparable evidence. When you finally do examine it closely — out of necessity, or because you’ve run out of other leads — it fails the same bars the primary clears. The problem is real, but the market is crowded. The margin is good, but there’s no path to buyers. The space is open, but it requires capabilities you don’t have. The attributes that made it attractive as a backup were real; the attributes that make it viable were absent and assumed.

This is structurally predictable, not bad luck. The backup tends to be named at the point where its attractive attribute becomes visible, which is early in evaluation, before the full bar is applied. At that point the primary is still being examined and the backup seems promising enough to reserve. But “promising on one dimension early” is exactly the standard that generates false positives. Every candidate has at least one attractive attribute — that’s the minimum to get noticed. The rigorous part is the rest of the evaluation, and backups typically never get that far.

The fix is straightforward: don’t name something a backup until you’ve held it to the primary bar. If you haven’t run it through the full multi-bar test, it’s a hypothesis, not a fallback. A hypothesis is fine — generating candidates is a legitimate research step — but it shouldn’t be named and filed under “insurance” as though it’s real. The naming creates false security that reduces the urgency to examine it, which is how you end up discovering, when you need it, that the backup was always a label attached to an idea that hadn’t been properly tested.

A real backup is rare for the same reason a real primary is rare: most candidates fail at least one bar, and the ones that don’t are uncommon. When a backup clears the full test, that’s a genuine asset. When it hasn’t been tested, it’s a piece of mental furniture that takes up space and provides comfort without providing cover.