A ranking you paid for is an advertisement, whatever the page calls it

For four years I ran the insurance business of a comparison site, and one of the products I built was a ranking that insurers paid to enter. I believed then that a sentence at the foot of the page, stating that the site received fees from some insurers, was enough. I no longer believe it. My position, now from the insurer's side, is that an insurer should not buy any ranking position, recommendation or selection that an ordinary reader would not recognise as paid at the moment of reading it. It should also judge the comparison partners it keeps by how long their policies stay in force, not by what an application costs.

How I sold an editor's selection

I joined Hakari Navi in 2013, when insurance comparison in Japan still largely meant requesting brochures from several companies at once. We built tables comparing premiums and main benefits for medical, cancer and term life policies, and readers used them. Traffic to our insurance pages grew from about 300,000 visits a month to 2.1 million by 2017. The site was an advertising and referral business; it held no agency registration and sold no policies. Its revenue came from insurers paying for referrals and, increasingly, for prominence.

The editor's selection began honestly. A small editorial team chose three policies per category that they considered good value for common situations. Within a year insurers were asking how to be included, and our commercial team proposed a fee for consideration. I agreed, on condition that the editors kept the final choice. In practice, policies whose insurers did not pay were seldom considered, because nobody prepared the material on them. By 2016 every policy in the selection came from a paying insurer. The disclosure was accurate. Our own reader research found that fewer than one reader in ten noticed it.

I resigned at the end of 2017, after the company decided to sell positions in more rankings, including a list of popular policies that would have blended paid placement with application counts. I am not proud that it took four years and a larger plan before I objected. I describe it because insurers sometimes speak of comparison sites as neutral ground, and I helped make one that was not.

What the 2023 change did not settle

Since October 2023, advertising that conceals the advertiser's involvement has been treated as a misleading representation under Japan's law on premiums and representations. The change pushed sites and advertisers to label paid content more clearly, and it was overdue. It did not settle what an insurer ought to buy.

The rule asks whether a reader can tell that a representation is the advertiser's own. A comparison ranking raises a subtler question. A list can carry a label saying that it includes sponsored listings and still invite the reader to treat its order as a judgement. A page can disclose fees in general without saying which positions were bought. Such arrangements may comply, and they still trade on the reader's belief that someone disinterested has weighed the options.

At Tomoshibi Life we reviewed every comparison placement in the summer of 2023. Three sites carried most of our comparison business. One agreed to mark each paid position individually where it appeared. Two declined, on the reasonable commercial ground that marked positions would earn less, and we left both that October. At the same time we withdrew a claim we had used for two years, that we were first in satisfaction among online medical insurers. It rested on a survey in which people who were not our customers rated insurers' websites. That measured impressions, not satisfaction, and our name should never have been attached to it.

The cost was substantial. Comparison sites had produced 29% of our new policies at one of our lowest costs per application, about ¥39,000. In the 2024 financial year they produced 18%. Our chief financial officer and our head of digital both argued for a slower exit, and they were right that the volume would be hard to replace. Total new policies fell 6% that year.

Buying comparison as advertising and judging it as distribution

The second half of my position concerns measurement. Most insurers pay comparison sites per application or per issued policy and compare channels on that cost. The measure rewards sites that send many applicants, including those who apply because a list told them to and lapse once they read their documents.

We now judge each comparison partner by thirteen-month persistency, and we pay part of the fee only when a policy reaches its fourth monthly premium. On that basis our remaining partner, which labels every paid position, looks better than its cost per application suggested. Persistency at thirteen months for policies from comparison sites rose from 84% to 90% after we left the two unlabelled rankings. Customers who chose us from a table they could read stayed longer than customers who chose us because we had been placed first.

There are fair objections. Comparison sites help customers, and many smaller insurers depend on them; a reader comparing premiums in a table is better informed than one responding to a television spot. I agree, and I am not arguing against comparison. I am arguing against buying the appearance of a judgement. Another objection holds that when one insurer withdraws, a competitor takes the position and the reader gains nothing. In the short run that is true. It is also the argument every participant in a practice uses to justify continuing it.

There is a commercial point for sites as well. A comparison business that sells the order of its lists is spending the trust that brought readers to it, and readers eventually notice. The sites I respect most now publish how their default order is decided and let a reader sort by premium, by benefit or by the insurer's complaint ratio. Insurers that want such sites to exist should pay them for traffic and leave the order alone.

I would put three questions to any insurer's comparison programme. Which positions do we pay for, and would a reader know it at the moment of reading? Which of our public claims rest on surveys of people who are not our customers? And what share of policies from each partner is still in force after thirteen months? The first two questions concern honesty. The third concerns money, and in my experience it is the one that changes budgets.