thoughtfulcandles

Metrics with Etsy search analytics.

April 18, 2020 | 5 minute read

Etsy provides a beta tool page that helps sellers understand the performance of their listings. This tool is the Search Analytics tool.


It outputs a series of data that includes:

  • Impressions: Total appearance in search
  • Position: Average position of your listing for that keyword search
  • Visits: Total visits to your shop and listings
  • Revenue: Gross revenue resulting from this search
  • Listings: Number of listings that appeared for this search.


From their seller-handbook post, this tool was meant to:

give you in-depth data about the queries shoppers are using to find your items, which tags on your listings match shoppers searches, and how well your listings are converting from search. You can also see your average position in search results so you can track changes in your ranking as a result of your optimization efforts.



  • give you in-depth data about the queries shoppers are using to find your items - The table of data definitely does this. You can filter the columns to identify top-performing queries.
  • which tags on your listings match shoppers searches - This is a little bit harder to decipher, but if you read deeper into how they calculate position, it incorporates tag relevancy. The higher the position, the more likely that the tags you’ve selected match your query.
  • how well your listings are converting from search - The conversions column pretty much does this. You can even roll this into a ratio. Maybe a visits/conversion that can then help forecast your performance in the coming months if you were to up visits by X amount.


Overall I think it’s a great tool! It does what it intends to do, and provides an avenue for you to use data to devise strategies to improve your store.

How I’ve been using this data is performing stupid naive calculations on it to spit out an aggregated value representing organic shop-performance.

The calculations take in impressions, shopVisits, and rank for all the queries I ranked for within a 30-day window for that date. I’m no statistician, but I believe this practice of using windows is pretty common when working with time-series data. I think it also helps smooth the curve out. I also ended up using windows because I forgot that I was pulling data that was in the 30-day range. 😅

Edit: As of 05/17/2020, I’ve changed up the logic of how I calculate the end-aggregated value. I’ve added conversions into the list of parameters (existing being: impressions, shopVisits, and rank) I use to calculate the end aggregated number. And I’ve removed this idea of using a window. Instead, I query specifically for all the relevant stats for that day.

Reason for the first change was because I started seeing some conversions come through, and this parameter now became applicable to my study. Reason for the second change was that I realized that by using a window range, it was impossible to tell whether the movement of that day’s value was due to the changes I made or due to some backdated fluctuations. By switching to only focus on that day’s figures, this makes it clearer to discern the impact of my listing’s changes.

The motivation to do this was because I lacked a good number to tell me how my store was performing over time. Originally I was measuring shop-performance by checking the number of queries I was ranking for, but this wasn’t accurate enough as I could be ranking for tons of keywords at positions 100+, meaningless keywords, but my resulting performance-score would be wrongly high.



If you are curious about results of my experiments, head here to read more. I’ll be updating that post instead of this one as this one is already pretty cluttered.


Anyways, below is the graph. Hopefully the graph will continue to go up and to the right. I don’t know if any listings can actually lose keywords if the content for those listings don’t change, but I guess we’ll find out.

Feel free to click into any of the points on the graph.

When you do click into a point, a table will pop up below that’ll show all the queries that were applied to calculate that datapoint.







TODOs:

  • Add in favorite actions into this figure. As for visually, maybe graph this as a separate line here.
  • Add in ad-related movement.


Additional notes: