Choosing between hoobuy spreadsheet lists: a side-by-side frame

Most comparisons of this material compare entry counts, and entry count is the least informative column available. A list with four thousand entries and a review date from eleven months ago tells you less about your next order than a list with four hundred entries reviewed last week. The column that looks like size is a column about accumulation.

What we compare instead is six fields that change a purchasing decision: entry count, most recent review date, distribution of link shapes, whether QC references are included, how many facets the list is organised by, and the duplication rate measured by seller identifier. Each of those is checkable by hand, and each of them is more predictive than size.

What we refuse to compare is anything we cannot verify: claimed number of curated items, claimed number of buyers served, and any statement about the maintainer experience. Those numbers appear on every list and are verifiable on none.

There are also situations where the comparison itself is the wrong move, and naming them prevents wasted effort. If you already know the exact item you want, a cross-list comparison adds nothing, because the only question left is whether that item pointer resolves. If your haul is a single parcel below the consolidation threshold, differences in facet count and duplication rate rarely change what you pay or how fast it moves. And if both lists in front of you were last touched in the same quarter, their remaining differences are mostly presentational, which means the honest answer is that either will do.

The table below records our Q3 2026 review of eight publicly visible lists. Fields marked Not verified are ones we could not establish by hand, and we are leaving them empty rather than estimating.

Entry count is measured by counting rendered entries, not by reading a stated figure. Review date is the most recent date we could find anywhere on the list. Link shape distribution is classified into marketplace address, short address, or a mix. QC reference is a binary: does the list link to or display inspection material? Facet count is the number of distinct grouping dimensions the list offers. Duplication rate is measured on the first fifty entries by seller identifier.

Each column has a rule attached, and the rule matters more than the reading. Entry counting starts from what renders: we scroll to the end, count distinct rows or cards as they appear, and apply the duplication rule at the point of measurement rather than at the point where a repeat is first noticed. Where a list states a total, that figure is recorded separately and never used as the count, because a stated figure describes an intention while a rendered count describes what is present.

Review date needs a decision rule, since lists carry several dates. We take the most recent date appearing anywhere on the page as rendered, whether it sits in a header, a footer, a changelog line or a note beside one entry. A date inside a filename or an address is not counted, because those usually record when a file was created rather than when it was last touched. A list carrying no date at all is recorded as undated, which is a different value from old.

Link shape is classified per address, in three passes. A full marketplace address naming the marketplace and carrying an item identifier counts as a marketplace address. An address that routes through a redirect service or a shortener, or that fails to name the marketplace, counts as a short address. Anything resolving to a search page, a storefront root or a category rather than to a single item is recorded as unclassified and excluded from the share, because those pointers cannot decay in the way this column is trying to measure.

Duplication is measured on the first fifty rendered entries, keyed on the seller identifier rather than the title. One seller frequently lists the same product under several names, and a title-based measure would score those as distinct. Fifty is a sample size rather than a threshold: it is enough to catch a list that repeats inside its own opening screen, and small enough to do by hand for every list in the sample. A list opening with fifty distinct sellers can still repeat further down, and we say so rather than implying otherwise.

Facet count is the number of distinct regrouping dimensions a list offers, not the number of groups inside one dimension. A list sorted into twelve categories offers one facet with twelve values. A list that can be regrouped by category, by provider and by destination region offers three. The distinction earns its place because facets answer a question the maintainer did not anticipate, while values inside a single dimension only answer the question they were built for.

The sample itself needs describing, because a table of eight rows invites the question of how those eight were chosen. They are the lists that surfaced repeatedly in public discussion of this category during the quarter, taken in the order we encountered them, with no filtering for quality and no list excluded for looking poor. Two families appear more than once, which is why some rows carry the same review date to the day. We did not expand the sample to thirty, because each extra row costs about an hour of hand measurement, and a shallow sample of thirty would tell a reader less than a careful sample of eight.

Entry counts in our sample ranged from roughly two hundred to over four thousand. The four-thousand-entry list carried a review date nine months old; the two-hundred-entry list was reviewed within the fortnight. If the purpose of a list is to give you a live pointer, the smaller and fresher list is the more useful artefact.

A useful habit is to divide entry count by months since the last review. The result is not a quality score, but it does tell you how much accumulation has happened without verification, and accumulation without verification is where dead pointers come from.

Counting is fiddlier than it sounds, and three conventions hold our numbers comparable across lists. Visible rows only: entries hidden behind a fold, an accordion or a load-more control count only after we open them, which sometimes moves a total by a wide margin. One row equals one entry, even where a row carries two addresses, because the column measures pointers as presented rather than pointers as usable. And a row struck through or annotated as retired still counts, with the annotation noted separately, since retired rows are evidence about the maintainer and are frequently the most informative rows on a page.

Review dates clustered into three groups in our sample, separated by roughly two months. That clustering matched the version markers rather than the entry counts, which is the strongest evidence we have that the markers track rebuilds rather than quality.

Where a list carries no date at all, we treat it as unverified rather than old. The absence of a date is not evidence of age; it is evidence that the maintainer does not date their work, which is a different property.

Two dates that look like review dates usually are not. A copyright line updates on its own and says nothing about content. A brand new listing date on the newest item shows the marketplace clock rather than the maintainer clock, since a pointer can be added to a list long after the listing itself was created. The date we record is the one describing work done on the list, and where no such date exists we record undated and move on rather than inferring one from context.

Every list hides something, and the thing hidden is usually structural rather than deliberate. Lists organised by a single facet hide items that do not fit the facet. Lists organised by brand hide the unbranded items that make up most of a haul by volume. Lists that present only marketplace addresses hide the short-link problem until you paste one into an order form and it fails.

The hidden thing that costs the most, in our experience, is the duplication rate. A list with a high duplication rate looks larger than it is, and if you are working through it manually you will spend time on entries you have already seen.

The unhidden thing that costs the second most is link shape, because it is visible and easy to ignore. A list of short addresses looks tidy, shortens well in a chat message and copies cleanly, which is exactly why the problem stays hidden until a paste fails. Where a list mixes shapes, the mixes are rarely random: the older material carries full addresses and the newer material carries short ones, so the failures cluster in the part of the list the maintainer touched last.

Match the list to the shape of what you are buying rather than to its size. A single high-value item wants a list with recent review dates and inspection references. A bulk order of unbranded basics wants a list with a high facet count, because facets are how you filter. A mixed haul wants a list that includes link shape information, because mixed hauls are where address problems surface.

None of that requires the biggest list. It requires the list whose columns match your decision, which is why we publish the columns rather than a ranking.

A short worked example makes the logic concrete. Suppose two lists sit in front of you: one with four facets, four hundred entries and a duplicate share under a fifth, and another with one facet, three thousand entries and a duplicate share above a third. For a single considered purchase, neither of those column sets is decisive, and the review date settles it in favour of the smaller list. For a bulk order across two categories, the four-facet list lets you filter twice while the larger list makes you scroll, so the smaller list wins again. The larger list only wins when your question is breadth with no filter, which is the least common shape of an order.