How FreyoMetalGoods tested one title change at a time

FreyoMetalGoods makes shelf brackets, furniture legs, towel racks and bathroom hardware in a workshop in Latvia. Irbe hand-welds and powder coats every piece, and custom sizes and finishes are one message away. The shop opened on Etsy in 2023 and has close to 700 reviews at an average of 4.8 stars, with listings available in English and ten other languages.

The listings read like they were written by someone who has held the steel. Standard legs are cut from 50 by 5 mm flat bar. The tall 71 cm desk legs move up to 50 by 8 mm, because a taller leg carries more leverage. Legs are sold one at a time as well as in pairs, for anyone replacing a single broken leg or building something that needs three.

The newest part of the catalog is bathroom hardware: wall-mounted washbasin brackets, freestanding vanity legs and frames for floating vanities that hold up to 140 kg. There's no single name for these products. The same kind of steel frame can be a vanity unit base, a basin stand, a sink stand or a washbasin frame, and the bathroom range already uses several of those names across its listings.

An Etsy title only has room for a few of those words before the buyer stops reading. Picking by gut feel is easy. Finding out whether the pick cost sales is much harder. Rewrite a title on a hunch and sales move a little, and there's no way to tell whether the words did it, the season did it, or nothing did.

Irbe took these questions to A/B tests in Alura. A test pairs Variant A, the listing exactly as it is, with Variant B, the same listing with one thing changed. Alura switches the live Etsy listing between the two on a schedule, every 8 hours by default, so both versions spend time in front of real shoppers. Every shopper sees the same version at any given moment. At the end, Alura compares the sales and revenue each version brought in, along with how many hours each one was live.

Setting one up is short work: start from the current listing, change the one part in question for Variant B, and choose whether the test should optimize for sales or revenue. Irbe kept each test narrow. Nearly all of them changed only the title, so any difference could be traced back to the words. The one exception was a photo test on a vanity frame, which left the title alone.

The tests ended automatically. In that mode a test runs for at least seven days, then Alura applies the version that did better on the goal picked at the start, sales or revenue, so nobody has to remember to switch anything back. A test that keeps the original is still useful. It says the words already in the title are doing their job, and that the time is better spent elsewhere.

The pairs below are real tests from the shop: the original title, then the version tested against it. Two ask whether a longer title that names more uses beats a short one. The third keeps nearly the same words and only moves the product name, washbasin bracket, to the front.

The kitchen island bracket's test version spent its characters on words like steel and hanging shelves in place of the note that it is sold individually. That note matters to buyers, since one shelf needs at least two brackets, and the live title still carries it at the end. The catalog's current titles share a pattern too: the product name first, then where it goes, then sizes and colours, then "Handmade EU".

Test one thing at a time, and give it long enough to mean something. A title change and a new main photo in the same week tell you that something happened, but not what.

And put the reasoning in the listing. Tall legs use thicker steel, so the description says so and explains why. Buyers who are building their own furniture read that kind of detail closely, and it answers the question before they have to send a message. The same goes for the title. If buyers keep asking whether a bracket comes as one or a pair, that answer belongs in the title.