Limit one decision set to three to five candidates that serve the same use. Each row should state the source, selected variant, strongest evidence, largest unknown and next action. If two rows cannot be compared on the same criteria, separate them into different lists.
Write down what you actually need
A clear shortlist begins with a sentence that describes the job, not the item name. For example: “Choose one lightweight everyday hoodie that fits over a T-shirt, has documented garment measurements and stays reasonable under a cautious parcel estimate.”
This sentence gives you a category, purpose, fit requirement, evidence requirement and cost boundary. Without it, every new link can look relevant and the sheet becomes storage rather than a decision tool. A useful DogeGoo price comparison therefore compares like-for-like candidates and states which costs are confirmed, estimated or still absent.
One small travel bag that fits a 13-inch laptop and has visible internal measurements.
Best bag from the spreadsheet.
Keep six columns you will actually fill in
| Column | Record | Avoid |
|---|---|---|
| 1. Candidate | Short neutral name plus selected color/size. | Full promotional title copied from a listing. |
| 2. Source | Final URL, page role, item ID and capture date. | “Found online” or an image without a link. |
| 3. Evidence | Strongest visible fact: dimensions, QC view, material label or included part. | “Looks premium” without a visible basis. |
| 4. Fit for purpose | How it meets the decision sentence. | A generic score that hides why it matters. |
| 5. Unknown / cost confidence | Largest unresolved issue and whether cost inputs are confirmed, estimated or absent. | Filling gaps with averages from unrelated items. |
| 6. Next action | Keep, ask one question, compare, or remove. | Leaving every row as “maybe.” |
Add extra columns only when they can change the decision. A beautiful spreadsheet with twenty mostly empty columns creates work without adding confidence.
Build the shortlist in three passes
Do not research every row equally. Pass 1 should be fast. Spend time on measurements, QC images and cost assumptions only after a candidate survives the relevance screen.
Treat look-alikes and duplicates carefully
Two rows may share the same image while pointing to different listings. One listing may appear under several names or agent routes. Use stable clues—item ID, destination URL, variant set and distinctive images—to group possible duplicates, but preserve separate source records until the match is direct.
When two candidates are truly equivalent for your decision, keep the better-documented row. Documentation quality is itself useful because it reduces the number of assumptions you need to make.
Keep “I like it” separate from “I can check it”
You may prefer one design even though another has clearer information. Keep those ideas apart. A plain three-level note is enough:
- Well documented
The source, selected variant and the measurements or photos that matter are clearly shown. - Partly documented
The candidate is still comparable, but one important detail is missing or only implied. - Poorly documented
You cannot confirm the source, variant or an important product detail.
Clearer product information does not prove seller reliability, authenticity or transaction safety. Check those separately.
Ask one question: why is this row still here?
Every candidate should have a unique reason to survive: best measurement match, clearest photo evidence, lower-bulk design, preferred feature, or a useful control for comparison. If two rows have the same reason, keep the one with better evidence unless another trade-off matters.
Archive removed candidates instead of deleting their history if you may need to explain the choice later. The active view should stay small; the archive can preserve the research trail.
Finish with a written trade-off, not a winner badge
A useful conclusion names what you gain, what you give up and what remains unknown. Example: “Candidate B best matches the reference measurements and has clearer QC views; Candidate A appears lighter but lacks package data. Keep B, ask for one weight photo, and do not treat the current estimate as final.”
This statement is more useful than “B wins” because it tells you what to do next and what would change the choice.
