Short answer
Ignore the row label at first. Confirm that each candidate points to the same item and variant, then compare photo coverage, measurements, source-page stability, evidence date, and shipping context. Choose the row with fewer unresolved identity gaps—not the row with the most confident wording.
Step 1
Align the comparison before scoring it
A fair comparison needs a common unit. “Black sneakers” is too broad; “the same model, black colorway, size range, material version, and source listing” is closer. Open each destination and record what the page actually shows today, because thumbnails and spreadsheet notes can outlive the linked listing.
If those fields do not align, keep the candidates in separate groups. A lower price for a different version is not evidence that one source is better.
Step 2
Use an evidence matrix, not an overall impression
Create one column per candidate and one row per question. Mark each cell as confirmed, partial, missing, or contradictory. The matrix makes uncertainty visible and prevents a strong thumbnail or detailed description from hiding a weak source chain.
| Evidence area | What good evidence looks like | Weak signal |
|---|---|---|
| Identity | Destination images and identifiers match the spreadsheet row. | A generic title or cropped thumbnail. |
| Photos | Useful angles cover the category-specific risk points. | Many repeated or promotional images. |
| Sizing or specs | Measurements explain how and where they were taken. | A size label with no unit or method. |
| Source stability | The current page resolves to the intended listing or catalog. | A redirect to a homepage, login screen, or unrelated item. |
| Parcel context | Weight status is labeled measured, listed, estimated, or unknown. | An exact-looking number with no source. |
Step 3
Diagnose the type of conflict
Same link, different notes
The notes may have been written at different times or for different variants. Treat the live destination as current page evidence, then check whether either note has a review date, photo reference, or measurement record. Do not merge unsupported claims.
Same thumbnail, different links
A reused image cannot prove that the sources are identical. Compare the seller or catalog identity, product ID, variant menu, and additional images. If only one destination can be traced, the other remains unresolved.
Same title, different construction
Broad titles often collapse material or version differences. Look for category-specific identifiers: outsole shape and panel layout for shoes, pocket and lining arrangement for bags, or port layout and model code for electronics.
Step 4
Apply tie-breakers in the right order
- Identity match: can you show that the evidence belongs to this exact item and variant?
- Category evidence: do the photos or specifications address the features that matter for this category?
- Measurement quality: are units, method, and size or model context clear?
- Current source: does the destination still resolve to a usable and relevant page?
- Price and parcel context: are cost and weight clues comparable rather than merely present?
Price comes late in this sequence because a precise price attached to the wrong variant is not a useful comparison.
Practical test
Run a ten-minute, three-row exercise
Choose no more than three candidates. Spend two minutes confirming identity, three minutes on photos and measurements, two minutes on the link destination, and three minutes recording gaps. Stop when the time expires. If none survives, change the category or missing detail instead of opening ten more weak rows.
Stop conditions
Know when comparison is no longer productive
Pause when the same product image appears across unrelated identities, when size or version cannot be established, when every destination is dead or gated, or when exact claims have no visible evidence. More browsing does not repair a broken comparison unit.
Next useful read: once a row survives source comparison, inspect whether its photos cover the right questions.
Read the QC photo guide