Optimising location filters for candidate curation

I was responsible for end-to-end design in a mission to help ops team members work more efficiently.

Context

OfferZen provides companies with a curated marketplace where they can find quality software developers. After onboarding candidate profiles go through a manual curation process. As OfferZen expanded into new supply markets, curation tooling needed to be updated to enable Marketplace Specialists to curate more efficiently.

The location filter highlighted in the curation tool

Problem

A critical function of the curation tool is to filter candidates by location. The filter didn't serve Marketplace Specialists well anymore and in the rapidly expanding marketplace they were at risk of being overwhelmed. I invited the Product Manager and Engineering Manager on my team along to chat with Dina and Reece, our Marketplace Specialists. Here is what we found:

Some candidates from unsupported locations need to be rejected manually

Although there are auto-reject rules in place, they aren't foolproof. The result is some pain in the curation process.

Current location filters don't support new location based curation heuristics

European candidates are all grouped together in the EU bucket. Candidates from different European countries require different curation heuristics.

Candidates are sometimes sent to the wrong location "bucket"

Some UK candidates are landing in the “No region” bucket, causing friction in the curation process.

Technical requirements

The old region polygon data model is being replaced with a new comprehensive location data model. All components in the product that use location data need to be updated to accommodate the new model.

Gaps in the region polygon is what was causing some UK candidates to be sent to the “No region” bucket.

The location filter highlighted in the curation tool

Business needs

OfferZen wants to provide candidates with a zen job search experience. Quick and efficient curation enables the business to deliver this brand promise.

Candidates on OfferZen are released in batches. If candidates aren’t curated on time they are at risk of missing the batch which degrades their experience and increases the likelihood that they’ll focus their job search efforts outside of OfferZen.

Goal

Enable more efficient candidate curation and prepare the location filter to use a new comprehensive location data model.

Success metrics

  • Current efficiency is maintained: all candidates in the curation tool are curated before 12PM every day.

  • Marketplace Specialists are able to reach a flow state while curating.

User research

Dina and Reece are curation experts. Collaborating with them in the design process was vital to creating location filters that would meet their needs.

Card sorting exercise

I asked Dina and Reece to arrange countries they would curate together in groups. To support faster sorting, sticky notes were colour coded according to the primary regions as generally defined at OfferZen.

The card sorting exercise Miro board

Findings

Allowing Dina and Reece to physically move countries around on Miro lead to great conversations about their different curation heuristics. The final artefact was much more valuable than a verbal or written response would have been.

  • Dina and Reece created black sticky notes to symbolise filter groups they wanted to see.



  • Countries were grouped according to their curation heuristics.



  • Countries that were in an experimental phase were singled out in black. Dina and Reece wanted the ability to curate these countries individually or as a group just in case heuristics changed.



  • Countries that are not currently supported on OfferZen were purposefully excluded from filter groups.

Allowing Dina and Reece to physically move countries around on Miro lead to great conversations about their different curation heuristics. The final artefact was much more valuable than a verbal or written response would have been.

  • Dina and Reece created black sticky notes to symbolise filter groups they wanted to see.



  • Countries were grouped according to their curation heuristics.



  • Countries that were in an experimental phase were singled out in black. Dina and Reece wanted the ability to curate these countries individually or as a group just in case heuristics changed.



  • Countries that are not currently supported on OfferZen were purposefully excluded from filter groups.

Allowing Dina and Reece to physically move countries around on Miro lead to great conversations about their different curation heuristics. The final artefact was much more valuable than a verbal or written response would have been.

  • Dina and Reece created black sticky notes to symbolise filter groups they wanted to see.



  • Countries were grouped according to their curation heuristics.



  • Countries that were in an experimental phase were singled out in black. Dina and Reece wanted the ability to curate these countries individually or as a group just in case heuristics changed.



  • Countries that are not currently supported on OfferZen were purposefully excluded from filter groups.

The completed card sorting exercise

Wireframes

The insights I gathered in the card sorting exercise allowed me to iterate on wireframes quickly. I selected the version that would best meet their needs for testing.

Version 1

✅ Dina and Reece would be able to filter countries by group or select individual countries to experiment with heuristics.

❌ Including all experimental countries as individual options adds a lot of noise.
Consider: Dina and Reece need to select experimental countries individually sometimes, but it would not be their primary workflow.

❌ Technical limitation: OfferZen uses a custom component library and at this time it did not include a searchable multi-select dropdown component.

Version 2

✅ The filters Dina and Reece need most often are separated visually so that they are easy to find and select.

✅ All countries in “Middle Earth” are available to select below the divider line, enabling quick experimentation with little dependency on Product.

✅ Unsupported locations are included as a group to catch candidates that fall through the cracks in the auto-reject rules.

✅ The majority of candidates in curation are based in the South Africa. Keeping it as an individual option would ease distribution of work between Dina and Reece.

User testing

I created a prototype of my suggested filter configuration and tested it with Dina and Reece in separate online sessions.

Assumptions

  • Displaying primary filter groups above the divider line will reduce noise. Dina and Reece should find the groups they need easily.

  • Dina and Reece will understand that individual countries below the diver line can be selected in combination with country groups.

  • Trying to search for a country in a list will be a natural response when opening the filter.

Figma prototype created for user testing

Findings

Dina and Reece were able to:

  • Filter by a country group easily.

  • Add and remove countries from a group to conduct a curation experiment.Clear their filters.

Both Dina and Reece tried to search the list. Our technical limitations degrade the user experience in this case.

Dina and Reece were able to:

  • Filter by a country group easily.

  • Add and remove countries from a group to conduct a curation experiment.Clear their filters.

Both Dina and Reece tried to search the list. Our technical limitations degrade the user experience in this case.

Dina and Reece were able to:

  • Filter by a country group easily.

  • Add and remove countries from a group to conduct a curation experiment.Clear their filters.

Both Dina and Reece tried to search the list. Our technical limitations degrade the user experience in this case.

Solution

The final location filter

Results

Dina and Reece are able to complete their curation tasks on time, contributing to a zen job search experience for candidates on OfferZen.

"The new location filters have made it a lot easier to work through large volumes of candidates each day. The new groupings are have been meaningfully defined, so it helps us run through the curation process in less time with less mental load - easily grouping candidates into logical subsets which require different heuristics, and tackle them straight on.” - Dina, Marketplace Specialist

Dina and Reece are able to complete their curation tasks on time, contributing to a zen job search experience for candidates on OfferZen.

"The new location filters have made it a lot easier to work through large volumes of candidates each day. The new groupings are have been meaningfully defined, so it helps us run through the curation process in less time with less mental load - easily grouping candidates into logical subsets which require different heuristics, and tackle them straight on.” - Dina, Marketplace Specialist

Dina and Reece are able to complete their curation tasks on time, contributing to a zen job search experience for candidates on OfferZen.

"The new location filters have made it a lot easier to work through large volumes of candidates each day. The new groupings are have been meaningfully defined, so it helps us run through the curation process in less time with less mental load - easily grouping candidates into logical subsets which require different heuristics, and tackle them straight on.” - Dina, Marketplace Specialist

Future improvements

During this mission we identified a few future improvements that could improve curation efficiency even more:



  • Create a searchable multi-select dropdown component to reduce the effort of finding a specific country in the list. This will serve Dina and Reece well when they experiment with new locations.



  • Enable configurable filter groups. This would reduce dependency on Product by placing the power in the user’s hands. Dina and Reece could the filter groups they need as their heuristics and experiments change.

  • Improve auto-reject rule logic to prevent candidates from unsupported locations being sent to curation.

During this mission we identified a few future improvements that could improve curation efficiency even more:



  • Create a searchable multi-select dropdown component to reduce the effort of finding a specific country in the list. This will serve Dina and Reece well when they experiment with new locations.



  • Enable configurable filter groups. This would reduce dependency on Product by placing the power in the user’s hands. Dina and Reece could the filter groups they need as their heuristics and experiments change.

  • Improve auto-reject rule logic to prevent candidates from unsupported locations being sent to curation.

During this mission we identified a few future improvements that could improve curation efficiency even more:



  • Create a searchable multi-select dropdown component to reduce the effort of finding a specific country in the list. This will serve Dina and Reece well when they experiment with new locations.



  • Enable configurable filter groups. This would reduce dependency on Product by placing the power in the user’s hands. Dina and Reece could the filter groups they need as their heuristics and experiments change.

  • Improve auto-reject rule logic to prevent candidates from unsupported locations being sent to curation.

What I learned

The best way to create a valuable, effective solution is to include your users in the discovery and design process. Dina and Reece are the experts in their domain, by collaborating with them I was able to design a solution that truly met their needs.