How I make the assessment: källor, prisdata och betygsskalan 0–10

The basis behind each assessment

An assessment is only worth something when it can be verified afterward. That’s why every review is based on information that is published, dated, and linked: the manufacturer’s own documentation, the stores’ product data, and a comparison with named competitors. The same basis is collected in the same order for each product, and the same template applies to all 395 reviews that are published. That’s the prerequisite for two ratings set six months apart to be comparable.

The manufacturer’s own information

The foundation is the manufacturer’s official product page, the specification sheet, and the PDF manual or quick guide that comes with the article. The manual is usually the best source—it regularly includes battery voltage, capacity in mAh, charging current, and the connection type that the product page’s sales text skips. Before any information is used, it’s checked that the page really applies to the correct article: the EAN number should be possible to find in the page’s own source code, preferably in the meta tag gtin13, since the same model name often covers several variants with different dimensions and different batteries.

Information from a retailer is used only as a supplement, and then it is stated in the text that the information does not come from the manufacturer. Standards and authority text are used where they belong: IEC 60529 for IPX classes, konsumentköplagen and distansavtalslagen for the right to make a complaint and the right of withdrawal, Kemikalieinspektionen for questions about softeners in plastic.

Price data and EAN matching across 19 stores

Article and price data is pulled from the product feeds from 19 Swedish stores via the affiliate networks Adtraction and Addrevenue. The feeds are matched to each other using EAN/GTIN, which is global and unambiguous, so the same physical product can be tracked throughout the entire retail chain. The numbers are normalized first: the same item can appear as 0848518046116 in one feed and as 848518046116 in another. The stores that only state the EAN for parts of the assortment are instead matched using the normalized brand and model name—a weaker connection, which is therefore flagged in the underlying data.

Nineteen feeds are thus imported, but not all 19 stores carry all products. Fourteen of them appear as a price in a published review; the other five have so far had no article that also exists on the site. That’s why the price rows in a review are never 19 items.

It’s this matching that makes the price comparison meaningful. The price in a review is not a list price, but the lowest amount charged by any of the stores for the same article number at the most recent price check, with the date shown. The variation is consistently larger than most people expect: Satisfyer Pro 2 Next Generation ranges from 219 to 599 SEK at the eleven stores that sell exactly the same EAN—173 percent difference on the same item, in the same package.

Comparison with competitors

A single value says little by itself. That’s why each product is compared against named competitors in the same price range, in the fields that can be substantiated for both: material declaration, dimensions and weight, number of modes and power levels, battery capacity and charging time, IPX rating according to IEC 60529, warranty period, and price. If two models differ only in price, that’s shown. If the more expensive one is better on just one point, it’s made clear which point it is, so you can decide for yourself whether it’s worth the price difference.

When the information is missing or contradicted

If the manufacturer does not publish a value, it’s shown in plain text instead of being replaced with an estimate: “Satisfyer does not publish either battery life or sound level.” Sound level is the piece of information that’s most often missing, and the one that’s most often filled in with a made-up number elsewhere. No decibel figure, no operating time, and no measurement are stated in a text here—only a source that can be opened.

If the manufacturer contradicts itself, both pieces of information are shown, each with its own source, and the contradiction is spelled out. We-Vibe states a 90-minute charging time in the specifications tab and 150 minutes in the function graphic on the same page. Satisfyer states eleven vibration programs on the product page and “70 Programs” on the packaging. Quietly choosing one of the values would be more convenient—and worse for you, since you’re the one who has to buy.

Source links and link verification

Every numerical item in a review has an entry in the source list at the bottom of the page, with a link and the date it was checked. The list is processed automatically with scripts/kontrollera-kallor.mjs, which collects all source URLs across all reviews, removes duplicates, and checks that each one still responds. Links that stop working are fixed before anything is published. Some manufacturer sites block automated requests and end up in a separate list that is reviewed manually in the browser— a source is never removed just because it refuses to respond to a robot.

The five criteria

• Material and hygiene — declared material in the contact area, information about phthalates, BPA, and latex, and what the manual says about cleaning and storage. A declaration that only says “body-friendly material” is considered missing information because the claim can’t be verified.
• Water resistance — IPX rating according to IEC 60529 and what the rating actually allows. A manufacturer that writes “water-resistant” without a rating can’t count it toward anything.
• Battery and charging — published capacity, charging time, runtime, and charging solution. A proprietary magnetic charging cable lowers the rating; a standard connection like USB-C does not.
• Function and controls — dimensions, weight, number of levels, and how the controls are built according to the manual: number of buttons, whether you can step back, and whether the model is available in multiple sizes.
• Price vs performance — the lowest price at the stores that sell the same EAN, set against what the specification provides compared with cheaper models in the same category.

The two external criteria, material and hygiene as well as price vs performance, are included in every review. The three intermediate ones are replaced with what actually differentiates the products in the category: in a category without a battery, instead, construction, care, or how well the manufacturer presents its measurements is assessed. For a small number of app-controlled products, the controls are assessed as a sixth criterion, with the same weight as the others.

How a “best in test” ranking is done

An individual review and a ranking are two different things. The reviews are written separately, and a rating set without seeing the rest of the fields can’t be used to rank—because the fact that one product got 8.3 and another got 8.0 can just as easily reflect two different strict evaluations as a real difference. Before a guide is written, all ratings in the category are therefore calibrated against each other, in three stages.

1. 1

   One: reading the values

   Each review is read through by someone who only extracts hard numbers—runtime in minutes, charging time, IPX class, warranty, material, price, number of stores—and notes what the manufacturer doesn’t publish. No assessment is done at that stage. The result is a table where all products are described with the same fields.

2. 2

   Two: the whole field is scored in one sweep

   The whole field is then scored in one sweep against that table, which is exactly what the individual reviews couldn’t do. The points are set 0–10 on the five criteria, one criterion at a time across all products in the category—never one product at a time. Otherwise the scale drives the outcome: you give battery points to the first product without knowing that the seventeenth lasts four times as long.

   Each criterion is reviewed three times, with different things in the foreground: what the product costs, how it works in everyday life, and how it holds up after two years. The same specifications are the basis each time—the difference is which attribute weighs heaviest when two products are evenly matched. If the three rounds differ on whether a product is better, that’s a signal that the placement is uncertain, and it’s carried forward to the next stage.

   The rating is the average of the five criteria, and the placement follows from the rating. So no one sets a placement manually. It’s a construction rather than a formality: it makes it impossible for number four to have a higher rating than number three, which is exactly the mistake the first attempt made when placement and rating were treated as two separate evaluations.

3. 3

   Three: the rebuttal

   The top five and the three rounds where the results differed the most are then tested separately, with a single task: to knock over the placement. The starting point is that the rating is wrong until there is evidence in the underlying material. The reviewer must answer whether any sub-score lacks support, whether there is a product that is better on the same criteria but ended up lower, and whether any reasoning is based on something that can’t be verified. If the objection is accepted, the points are changed, and the change is written out in the guide.

   In the clitoral stimulator guide, eight objections were accepted and the ratings were changed. One of them concerned the winner, whose battery points were lowered because the charging cable is product-specific—something this page explicitly says should lower the rating. Another concerned a product that had been scored on a price at a store that was sold out; with the actual lowest price, it fell one step. Both errors would have gone unnoticed without the third stage.

Why sound level is not a criterion

Sound level would be a reasonable criterion if it were possible to compare. It isn’t. In the clitoral stimulator category, twenty-nine of the thirty-two manufacturers publish no sound level at all, and the three that do provide a ceiling — “under 60 dB” — without specifying the measuring distance or measurement method. You can’t set two such figures against each other, and ranking based on quietness would have required measurements that weren’t done. The criterion therefore is completely left out instead of being filled with guesswork.

Thin documentation doesn’t lower the rating

A product doesn’t become worse just because the manufacturer’s website is poor. If a value is missing, the product can’t score points in that area, but the lack of information itself must never reduce the rating — it is a limitation for the person writing, not a characteristic of the product. The same applies to contradictions in the manufacturer’s own material: they are presented in the review, but they don’t affect the placement.

Independence and funding

The site is funded by an affiliate commission from store links. The commission is the same regardless of which product is recommended, and it is paid by the store, not by the manufacturer — that is the whole reason it can’t influence a placement. The commission rate isn’t included in any of the five criteria; no store can buy a spot in a ranking, and no texts are sponsored. No manufacturer may read a text in advance or see a rating before it’s published. The details are listed in affiliatepolicyn.

Updates and corrections

The prices are pulled when an article is reviewed, and the date of the price check is shown next to the price information. A text is reviewed when the manufacturer changes a specification, when a successor makes the comparison outdated, or when a price changes persistently. The date at the top of the article shows when it was most recently reviewed.

If you find information that doesn’t match the source, I want to know about it. The error is corrected, the rating is recalculated if it’s affected, and a note at the bottom of the article explains what was changed and when — see rättelsepolicyn. Correction notes aren’t removed afterward, because they are the only part of a correction promise that can be verified.