How We Score

Last updated: August 2026

Most fragrance sites tell you a perfume is good. Very few tell you how they know. This page explains where our data comes from, how the scores are built, and where they fall short.

Where the data comes from

We aggregate what real buyers say about fragrances in the places they actually talk to each other. As of August 2026 our corpus holds 875,512 pieces of community content:

  • Reddit: 747,432
  • Facebook groups: 83,418
  • TikTok: 30,446
  • YouTube: 14,216

From that corpus we have extracted and matched 82,782 individual product mentions to fragrances in our catalogue. We combine this with structured data from Fragrantica, including note pyramids and community vote distributions.

The distinction matters: Fragrantica reviews are written by enthusiasts reviewing a fragrance. Our corpus is ordinary buyers talking to each other, unprompted, in their own words. Those are different populations answering different questions, and the second one is better at predicting whether you will actually like something.

How the scores work

Each fragrance carries scores out of 10 for longevity, projection, blind-buy safety, and how polarising it is. They are derived from community discussion and vote distributions, not from brand marketing and not from our own opinion.

Two rules govern them:

  • An evidence floor. We need at least three independent mentions before we treat community signal as meaningful for a fragrance. Below that, we say nothing rather than guess. About a third of our catalogue currently clears this bar.
  • Smoothing before ranking. Scores are shrunk toward the catalogue average based on how much evidence sits behind them. A fragrance with one enthusiastic comment cannot outrank one with two thousand. This is the difference between a raw ratio and an honest estimate, and it is where most competitor scoring falls down.

What we will not do

  • No paid placement. No brand can pay to rank higher, be recommended by Dira, or have a negative note removed. We are an affiliate business and that is disclosed, but commission has no influence on scoring.
  • No invented specifications. Every note, accord, and score traces to a source. Where we do not have data, the page says less rather than filling the gap.
  • No hiding the downside. If a fragrance is polarising, wears close to the skin, or needs weeks to settle before it is worth judging, we say so, including on fragrances we earn commission from.
  • No treating absent data as bad data. A fragrance with no votes is not a fragrance with a score of zero. If we do not know, we do not display a number.

Where this falls short

Being honest about the method means being honest about its limits.

  • Coverage is uneven. Popular fragrances have thousands of mentions. Newer or more obscure releases may have none, and their pages will be thinner as a result.
  • Community discussion skews. People post more about fragrances that are cheap, hyped, or divisive. A quiet, well-made fragrance can be under-represented.
  • Skin chemistry is real. Longevity and projection vary by person more than any score can capture. Our numbers describe the average reported experience, not yours.
  • Data ages. Reformulations happen, and older discussion may describe a version of a fragrance that no longer ships. We refresh the corpus regularly but cannot catch every batch change.

Questions

If you think a score is wrong, tell us. Get in touch and point us at the fragrance. Corrections that hold up get made.