Methodology

Where specs come from

We collect each spec from one or more sources and keep every candidate value, not only the one we show:

• Manufacturer: the brand’s product page or spec sheet.
• Retailer: listings from stores that sell the product.
• Review: published measurements from independent reviewers.
• Measured: our own measurements, marked “Measured by us.”

Every value on a product page links to its source and records when we retrieved it.

We read brand sites politely: our requests identify themselves, follow each site’s robots.txt, are spaced out, and happen about once a week. If a site blocks automated requests, we stop and wait for an official feed rather than work around it.

Some brands’ sites refuse automated requests or publish no readable product data (Field Notes, Casio, Patagonia and others). For those we read an open retailer that sells the brand instead, under the same rules: its public product feed and pages, its robots.txt, once a week. Specs from a retailer’s page go through the same quote check, are labeled as retailer-sourced, and the product page says which retailer they came from. Prices, links and photos for those products are the retailer’s. We only use retailers that price in US dollars and have never sued over scraping or AI.

How we normalize

We store weights in grams, volumes in liters and dimensions in millimeters, then show both imperial and metric. Water resistance uses five fixed levels, from “none” to “waterproof (rated),” so a brand’s marketing language maps to the same scale as everyone else’s.

For laptops, we record the brand’s stated maximum size and the sleeve’s internal dimensions when available. Stated sizes vary between brands; dimensions don’t.

Colors keep the brand’s own names, like “Periwinkle” or “Basecamp Orange.” For filters, search and swatches we group each into one of 16 color families, including “multicolor” for patchwork and color-blocked designs, so a periwinkle pack shows up when you look for purple. A basic color word in the name decides first; then the brand’s own swatch color, when it publishes one; and when neither can tell (a name like “Eclipse” on a dark swatch), an AI model looks at that colorway’s photo and picks the family. Swatches show the family’s tone, not the exact product color; the photo shows the real one. A two-color name counts as both colors. A product whose color the brand doesn’t publish at all gets a plain swatch labeled “Color not published” and is left out of color filters.

Battery banks are compared on capacity in mAh, the figure brands advertise, rated at the cells’ nominal voltage (about 3.7 V). A few brands state mAh for the whole pack at a higher voltage instead (for example, 6,400 mAh at 14.8 V); we convert those to the 3.7 V equivalent of their published watt-hours, so every bank compares on the same basis, and show the brand’s own figure next to it. Watt-hours are only shown when the brand publishes them. “Under 100 Wh” and the airline notes come from that published rating, never from a conversion, because 100 Wh is the limit for carrying a spare battery on a plane without the airline’s approval. Output is shown per port and in total, as stated; when the total is capped below one port’s maximum, the headline uses the faster figure. Colors aren’t compared for battery banks.

Multi-tools are compared on the tool count each brand states, since brands count differently (some count both ends of a driver, others count one tool per function). So we also list which tools each one includes, and filters use those. “No knife blade” is only shown when the brand says so; “carry-on friendly” additionally needs a published closed length under TSA’s 7-inch limit for tools. TSA officers make the final call.

Flashlights are compared on the maximum output each brand states, usually a short turbo burst, not what the light sustains. Beam distance and peak intensity (candela) are shown when published; “high CRI” needs a stated CRI of 90 or more. Weight is the brand’s figure, with the battery when the page gives both.

Pens are compared on closed length, notebooks on page count, watches on case diameter and wallets on the number of cards the brand says they hold. “Pocket size” notebooks have a published page size up to 10 × 15 cm. Watch water resistance is the stated rating; “ISO 6425” only appears when the brand states that diver’s standard. A pen “writes upside down” and a wallet “blocks RFID” only when the brand says so.

Headphones are compared on the frequency response each brand states, the one figure nearly every brand publishes for wired and wireless models alike; it says nothing about how they sound. Driver size, impedance and battery life are shown when published; “noise cancelling” only when the brand says so. Smart watches and fitness trackers are compared on case size: the diameter for round cases, the longest side for rectangular cases and bands. Battery life is the brand’s figure for typical smartwatch use, with GPS hours separately. Sling bags, hip packs and messenger bags are compared on stated capacity in liters, like backpacks (cubic inches are converted), never estimated from dimensions. Sunglasses are compared on lens width, the first number of the size printed inside the arm (for example 55-18-145: lens, bridge, temple); “polarized” and “ANSI Z87.1” only appear when the brand states them for that lens.

Charging cables are compared on length, converted to meters and shown in feet too; when one listing sells several lengths we show the range. Power is the wattage the brand rates the cable for, which is a ceiling: the charger and device decide what actually flows. Data speed is shown as stated; when a brand names only a standard with a fixed rate (USB 2.0 is 480 Mbps, Thunderbolt 4 is 40 Gbps) we show that rate, and plain “USB4” (20 or 40 Gbps) stays unpublished. The cable type (USB-C to USB-C, USB-C to Lightning …) is read from both published ends, and video support, certifications and bend ratings appear only when the brand states them.

Water bottles are compared on capacity, converted to milliliters and shown in fluid ounces too; when one listing sells several sizes we show the range, and its weight, diameter and height are the smallest size’s. Whether a bottle fits a backpack side pocket is judged only from the diameter the brand states: 75 mm or less (about a 1 L smartwater bottle) counts as slim. “Ultralight” means 100 g or less per liter held, computed from the brand’s weight for single-size listings. Insulation, stove safety, filter compatibility and dishwasher safety appear only when the brand states them.

When sources disagree

1. Our own measurement wins.
2. Otherwise, the manufacturer’s figure wins.
3. We flag the disagreement on the page and list the other values with their sources.

How we measure

When we weigh a pack, we weigh it empty, with all included accessories attached, on a scale accurate to 1 g.

Prices

We check prices about once a week on brand stores (and, later, retailer and affiliate feeds) and show the date of the last check. After four weekly checks, product pages show the typical price range (the middle 80% of weekly prices) and the most common price. Products we've seen on sale are tagged “Likely to go on sale”, always with how often, how deep and when. When a store marks a price down from its usual price, we show the sale. Prices change between updates, so always check the retailer for the current price. We don’t track in-store stock.

Review scores

Where independent review sites publish a score for a product, we show a combined review score on a 1–10 scale. Each site's score is converted from its own scale (its lowest possible score becomes 1, its highest 10) and averaged, weighted by the site's popularity: its rank in the Tranco list of the most-visited websites, so a widely read site counts more than a small one. You can leave any site out by selecting its logo. Every score links to the review it came from. We don't write reviews ourselves.

How we use AI

AI helps extract specs from source pages and drafts short summaries. Summaries may only state facts that come from a stored spec or a calculation on our data, like “lighter than 80% of 20–30 L packs.” They never include hands-on impressions we didn’t have. Every extracted spec must come with an exact quote from the source page, and we check that quote really appears there. Specs that can’t be matched to a quote are dropped, not guessed. When a brand doesn’t publish a spec, we say “Not published.”

Affiliate disclosure

Some retailer links earn us a commission. Commissions never affect which products we list, how filters work or how lists are sorted. Details: affiliate disclosure.