Why we publish recommendations instead of rankings

A ranking implies that the things being ranked are comparable on a single dimension, and that there is a correct order. Both implications are usually false. "The ten best productivity apps" is not really a list of ten best — it's a list of ten popular options, ordered by some combination of click-through-rate, advertiser pressure, and the writer's recent experience. The reader who treats it as a real ranking is usually misled, and the writer who produces it is usually compromised.

A recommendation is a smaller and more honest claim. It says: this thing was useful for this specific person in this specific situation, and here is the thinking that connects the two. The reader can transfer the thinking, judge the situation, and decide whether to apply the recommendation to their own. Nothing is hidden; nothing is ranked; the basis is visible.

The reason rankings still dominate online publishing is mostly economic. Rankings get clicked. They convert to affiliate revenue. They produce charts that scale. Recommendations are slower work and harder to monetize, because they depend on the reader trusting the writer's judgment, which takes time to build. A writer who relies on rankings for revenue is locked into producing more rankings; a writer who relies on recommendations is in a different relationship with the audience.

This shows up in the texture of the writing. Rankings are written in a confident, generic voice — "the best app for X" — that flattens the writer's actual experience. Recommendations are written in a smaller, more specific voice — "this app worked for me when I was trying to do X with these constraints." The smaller voice is harder to fake, and readers calibrate their trust accordingly over time.

There's a useful second-order effect of publishing recommendations instead of rankings, which is that the reader develops their own judgment faster. A reader who consumes ten years of rankings tends to outsource decisions to the rankings. A reader who consumes ten years of recommendations starts to recognize when a recommendation will or won't apply to their situation, and to ask better questions of any source. That's a more durable kind of value.

The hardest discipline in publishing recommendations is being willing to say "I don't know" or "this didn't work for me." A ranking can't admit this; the format requires confident verdicts. A recommendation, by contrast, gets stronger when the writer's uncertainty is visible — because the reader can now distinguish between the parts of the recommendation that are tested and the parts that are speculative. Writers who learn to do this well end up with the most loyal audiences in their categories, because the audience trusts the inventory of unknowns.

We publish recommendations here partly out of conviction and partly out of self-interest. The conviction is that rankings have hollowed out a lot of the internet's writing, and the world doesn't need more of them. The self-interest is that recommendations build a different kind of relationship with the reader — one where the reader keeps coming back not because the algorithm pushed our piece up, but because they remember our name and want to hear what we'd suggest.

The piece we'd point you to on this argues the case more thoroughly, with examples of categories where rankings have done particular damage to reader judgment. The retail-electronics section is the most pointed and the most useful.