Rankings decide the shortlist: how mybest, Kakaku.com, @cosme and comparison sites shape buying in Japan

A shopper in Osaka wants a new hair dryer. She does not start on a brand website. She types ドライヤー おすすめ ("recommended hair dryers") into Google, opens a mybest article that tested dozens of models side by side, checks the top three on Kakaku.com and buys one on Rakuten. An IT manager in Nagoya choosing an expense tool does the same on other sites: a category page on BOXIL or ITトレンド, a few document requests, the scores on ITreview, then calls with two vendors.

In both cases the shortlist exists before any brand has made its case. If you are not on the list, you are rarely in the conversation.

In the outside reviews we do of international brands, this comes up constantly. Brands that lead their category in the US or Europe are missing from the おすすめ rankings that fill page one of Japanese search. Software vendors with hundreds of G2 reviews have no presence on the comparison sites Japanese IT teams actually use. It is rarely a budget problem. Nobody has looked at the category from the buyer's side, in Japanese.

Three kinds of ranking, three different jobs

Most plans go wrong by treating all of these as "review sites". They rank in three different ways, and each rewards different work.

Tested rankings. Editors buy products and compare them under the same conditions. mybest (my-best.com) is the largest: it reports 30 million monthly users and says it tested more than 16,000 products and services between October 2024 and November 2025. You cannot buy a position. You can make sure the editors know your product exists.

Review-driven rankings. Position comes from what users write. @cosme for beauty, the satisfaction rankings on Kakaku.com and ITreview for business software work this way. The job is collecting enough genuine reviews from real customers, in Japanese.

Activity-driven rankings. Position comes from sales, clicks or enquiries. Amazon.co.jp and Rakuten rank by sales. Kakaku.com's 人気売れ筋 ("popular and selling") ranking counts one week of clicks through to listed shops. BOXIL and ITトレンド rank software by document requests (資料請求). Here, distribution and paid listings matter more than reviews.

Search behaviour splits along similar lines. A 2023 analysis by the Japanese data company VALUES found that mybest drew its search traffic mostly from queries containing おすすめ and ランキング, while Kakaku.com led on 比較 ("comparison") and 口コミ ("reviews"). The first group is exploring. The second is checking a shortlist.

The main sites by category

Beauty. @cosme's Best Cosmetics Awards are the closest thing Japan has to a national beauty chart. The 2025 edition counted 1,487,167 reviews of 52,852 items posted between November 2024 and September 2025, with results by price tier, plus a half-year edition each June. Every review on @cosme has to say whether the item was bought or came from a monitor programme or gift (購入品 or モニター・プレゼント), and the ranking gives more weight to reviewers the site trusts. A sampling programme shows up as a sampling programme. That is fine. It just is not a shortcut.

Electronics and durables. Kakaku.com pages combine specs, shop prices, user reviews and a discussion board. The 人気売れ筋 ranking is calculated from a week of clicks to designated shop sites, so a product sold only on a brand's own site, or through retailers that are not listed on Kakaku.com, can have a page and still struggle to rank. Shops pay per click, from ¥10 a click plus a base fee of up to ¥10,000 a month, after a review that takes about a month. Manufacturers can also add a features section (製品特長情報) to their product pages.

Marketplaces. Amazon and Rakuten rankings are driven by sales within short periods. A launch spread thinly over three months ranks worse than the same volume in a few weeks, and the ranking badge then helps sell to the next buyer.

Business software. ITreview launched in 2018 in partnership with G2; its Grid plots products on satisfaction and recognition. A product needs at least 10 reviews to qualify for a Leader or High Performer badge; about 1,270 products received one in the Fall 2025 edition. BOXIL states that its document-request ranking only counts services from companies with a paid listing. So "No.1 in document requests" means first among paying advertisers, which head office should know before reading it as market leadership. Your G2 profile does not carry over: Japanese buyers want Japanese reviews from companies like theirs.

Getting on the lists legitimately

For tested rankings, make the editors' work easy. mybest opened a form in July 2024 where manufacturers can send catalogues, spec sheets and press releases, and states plainly that this guarantees neither inclusion nor better treatment. What helps is practical: a Japanese spec sheet, a price in yen and a clear answer to where the product can be bought in Japan. Editors test what their readers can buy. mybest also sells sponsored articles, labelled PR; treat those as advertising and judge them as advertising.

For review-driven rankings, put product in the hands of real users through labelled monitor programmes. Ask every buyer for a review in the same way after delivery, whatever their experience was. For software, ask your Japanese customers to review you on ITreview; the site runs its own gift card campaigns for reviewers, so you do not need to offer anything yourself.

For activity-driven rankings, paid listings are a normal part of the market. A Kakaku.com shop listing or a BOXIL plan is an ad with a disclosed model. Budget it like search and judge it on sales or pipeline, not clicks or leads.

What the stealth marketing rules changed

Since 1 October 2023, advertising that consumers cannot recognise as advertising counts as a misleading representation under Japan's Act against Unjustifiable Premiums and Misleading Representations (景品表示法). The Consumer Affairs Agency holds the advertiser responsible, not the influencer or reviewer. Whoever ran the campaign, the liability sits with your company.

The first order came in June 2024: a medical corporation had offered patients a discount on flu vaccinations if they posted high star ratings on Google Maps. In November 2024 Taisho Pharmaceutical was ordered to act after reusing excerpts of influencer posts marked #PR on its own website as customer voices, without the disclosure. The lesson for rankings: disclosure has to travel with the content. In September 2024 the agency also reported on "No.1" claims such as 顧客満足度No.1 (No.1 in customer satisfaction): many advertisers did not know how the surveys behind them had been run. If you quote a ranking, quote a real one, with source and date.

• Fine: gifting product with no conditions on what reviewers say, disclosed with words such as 広告, PR or プロモーション.
• Fine: sponsored articles and paid placements carrying a clear PR label.
• Fine: asking all customers for a review after delivery.
• Not fine: rewards that depend on the rating or content. Rakuten's shop rules have banned this since 2015, and Amazon.co.jp allows free product for reviews only through Vine.
• Not fine: reusing paid content in your own site or ads without the disclosure.
• Not fine: a "No.1" badge based on a survey nobody on your side has checked.

Connecting rankings to search and measurement

A ranking sends people to search for you by name. After reading a mybest list, the next search is often the brand name plus 口コミ or 評判 ("reputation"). In our reviews, those searches are often answered by resellers, competitors or old blog posts. That is where search ads earn their keep: brand terms plus 口コミ and 評判, on Google and Yahoo! JAPAN, landing on a Japanese page with price in yen, where to buy and reviews from Japanese customers.

Measure the chain in four places. Referral traffic from my-best.com, kakaku.com, cosme.net and itreview.jp. Branded search volume, week by week, before and after a listing or award. Daily marketplace sales. For software, document-request leads followed through the CRM to meetings and deals, because a cheap lead is often early research. Switch paid listings on by category or in planned windows, so the effect shows in the data instead of being assumed.

Mistakes we see most often

• Translating the English product page and assuming the rankings will find it.
• Treating a G2 or US Amazon rating as proof that carries over to Japan.
• Reading a document-request ranking as market share.
• Copying a global seeding brief without checking how the disclosure reads to a Japanese consumer.
• Sending samples to editors with no Japanese spec sheet or yen price.
• Judging ranking work on impressions instead of branded search and sales.

Two examples

These are illustrations, not client cases.

Take a European skincare brand selling in Japan through its own site and a Rakuten store. At home it is a top-three name. In Japan, the search for its category plus おすすめ returns mybest and @cosme pages where it does not appear. A sensible quarter: send mybest a Japanese spec sheet for its two strongest products, run a labelled @cosme monitor programme on the same two, ask every Rakuten buyer for a review, and run brand plus 口コミ search ads to a Japanese page with yen prices. Then compare referrals, branded search and Rakuten sales with the previous quarter.

Now a US expense software company with two people in Tokyo, hundreds of G2 reviews and none on ITreview. First, ask its Japanese customers for ITreview reviews until it passes the 10 needed for a Grid badge. Second, test a paid BOXIL listing in one category for a quarter and judge it on meetings and deals, not lead count. Third, publish two Japanese customer stories, because Japanese buyers look for companies like their own.

A short checklist

• Search your category with おすすめ, 比較 and ランキング in Japanese and note who is on page one.
• Note which kind of ranking each page uses: tested, review-driven or activity-driven.
• Prepare a Japanese spec sheet, yen pricing and where-to-buy information for editors.
• Plan reviews from real customers, disclosed wherever anything was given.
• Treat paid listings as ads: label them, budget them, judge them on sales or pipeline.
• Track referrals, branded search and sales before and after each change.

Meridian's outside review starts with exactly these searches, done the way a Japanese buyer does them, and shows which gaps are quickest to close.