Colour3 min read

Your customer searches for red. Why is your webshop showing green?

Appir Search reads colors from product photos. Search for a 'red dress' and you'll see true reds first, followed by burgundy and increasingly purple shades.

Remco Coppens · Founder, Appir

Search bar with the query “red dress” above six different dresses that shift gradually from bright red through burgundy to plum, each with a small color swatch.

Type 'red dress' into the search bar of a fashion webshop and you expect to see red dresses. The really red ones first, perhaps followed by burgundy or coral. You wouldn't expect a green dress.

Color plays a part in many webshop searches. Between 5 and 12% of all searches include a color (Baymard, 2016). Yet 39% of webshops fail to return good results for a search like 'red dresses' (Baymard, 2026).

What text doesn't tell you about color

When you rely on text to identify colors, you mainly look at which words tend to appear together. That works well for 'pullover' and 'sweater', because those words are often used in similar contexts.

Colors are different. In product descriptions, 'red' appears next to 'green' just as often as it does next to 'purple', as in 'available in red, green and navy'. A system that only analyzes text sees three related color terms. But the actual colors can be quite far apart.

Purple is closer to red than green is. Burgundy is closer to red than purple. You can't work out those distances from the product description.

Red first, then gradually moving toward purple

Someone searching for red naturally wants to see red items at the top. But the results don't have to stop at an exact match. Further down the page, you can include burgundy, raspberry and eventually shades that lean more toward purple. The selection gradually expands without the results becoming completely unrelated.

To make this work, you need to measure how close a color is to red, and how much red there is in a particular shade. That means assigning a score, from very red to only slightly red.

Recognizing color in the product photo

Since the beginning, when Appir was a consumer app where people swiped through clothes, we’ve used product photos to understand what a garment actually looks like. A color label doesn’t tell the whole story.

We analyze the colors in the product photo, garment by garment. If a model is wearing a blue jacket over a white T-shirt, we associate the blue with the jacket and the white with the shirt. When someone searches for a blue jacket, the T-shirt doesn't count.

We also account for the conditions in the photo. A shadow in a fold shouldn’t turn blue fabric into navy, and bright studio lighting shouldn’t make beige fabric look white. The important thing is to identify the actual color of the fabric.

We still take the webshop's own color label into account, but we don't rely on it. A sweater with the product name 'Oat' can therefore show up in a search for 'beige sweater'.

A blue and white striped sweater

Imagine someone searching for a blue and white striped sweater. At the top of the results are sweaters that clearly feature both colors. A plain blue sweater, or a white one with just a blue trim, will appear further down.

We also distinguish between different shades of blue. A general search for 'blue' might start with mid-blue items. But if someone searches for 'navy with white stripes', navy comes first, followed by cobalt blue options.

Color is just one of the factors that determine the order of the results. The rest of the search query matters too, as does what someone has looked at earlier during their visit. A navy sweater with white polka dots has the right colors, for example, but it's a less relevant match than a striped sweater.

Appir Search is now live on Shopify. Using a different platform? Let us know and we'll look into building a connection.

Curious to see how your collection performs when people search by color? Send your webshop URL to remco@appir.nl. We'll build a free demo store for you, with no obligations.

  • color detection
  • product catalog
  • search by color
  • site search

Remco Coppens

Founder, Appir

About Appir

Related articles

See what Appir finds in your own catalogue

A free demo store built from your products, every plan, and the dashboard that shows what search earns.