This documentation describes a new ranking algorithm that has not yet been released in production

Search Ranking Algorithm

The search ranking algorithm is designed to return the most relevant profiles first, reward members that are selective in their choice of focus areas, and ensure that profiles that may not match many searches are given higher visibility when they do.

The algorithm does this by:

  • favouring specialists over generalists
  • favouring profiles that serve more specific markets for remote roles
  • favouring profiles that are closer to the office location for roles that are hybrid or onsite

As a result, profiles that have chosen fewer selections of focus area, and/or are more geographically focused, will appear in the results for fewer searches, but will rank higher in the searches they do appear in.

How it Works

Focus areas

For each result in a given search, we calculate a ranking factor for each of the following focus areas that the search filters on:

  • Discipline
  • Vertical
  • Model
  • Stage
  • Expertise

Note that:

  1. Role is not included, so selecting multiple roles will not impact your rank.
  2. Only focus areas that have been enabled as a filter for a search are considered in the rankings for that search.
  3. Discipline is a required filter, so is always a factor in ranking.
The ranking factor for a given profile and focus area is calculated as follows:
  • ((T - N) + 1) / T
  • where N is the number of options selected for the focus area on the profile concerned
  • where T is the total number of options available for that focus area

For example, the below table shows how the ranking factor for 'stage' varies depending on whether the profile has one, three, or all six of the available stages selected (pre-seed, seed, early, growth, exit, and mature):

Focus AreaOptions
available
Options
selected
Ranking
factor
Stages611.00
Stages630.67
Stages660.17
Location

A ranking factor is also calculated to reflect the relationship between the location of the role and the location of the profile. However, this factor is calculated in different ways depending on whether the search is for a remote or hybrid/onsite role.

For a search relating to a remote role, the ranking factor favours profiles that serve smaller areas, as per the below table.

Profile open to
work with teams
Role is with a
team distributed
Ranking factor
WorldwideWorldwide1.00
In a regionIn that region1.00
In a CountryIn that country1.00
WorldwideIn a region0.67
In a regionIn a country0.67
WorldwideIn a country0.33

For a search relating to a hybrid/onsite role, the ranking factor is based on the "as the crow flies" distance (as calculated using the Haversine formula) between the profile's location and the search location, and the number of locations on the profile. If a profile has multiple locations, we assume that the member concerned divides their time evenly between locations, and reduce the weight of each location accordingly.

The formula used is as follows:

  • (1000 - D)/(D x N)
  • where D is the distance in km between the closest location on the profile to the the role
  • where N is the number of locations on the profile
Hybrid role example
DistanceNo. of locations
on profile
Ranking
factor
30km10.97
100km20.45

Note that:

  1. Although the profile allows distances to be defined in miles, all distances are stored in kms, and ranking factor is therefore calculated in kms
  2. Adding additional locations to your profile will ensure that your profile is returned in searches for roles near all of those locations, but has a significant negative impact on the location ranking factor. You should therefore avoid adding more than one location unless you are equally open to hybrid / onsite work in multiple locations that are significantly separated in distance (eg. London and New York).
Rank calculation

Once all ranking factors have been calculated, the final rank of a profile is calculated by multiplying all the ranking factors that apply to that search together, and then by 100.

Sorting

All matching profiles are initially sorted by rank. However, because focus areas only impact ranking when a filter is applied to them, multiple profiles may have the same rank for a given search, particularly for searches for remote roles, or for hybrid / onsite searches in major cities.

In this situation, profiles that do have the same rank are sorted amongst themselves by two additional factors:

  1. The geographical distance between the profile and the approximate location of the end user performing the search, based on their IP address.
  2. As a final tie-breaker, a hash of the end user's IP address and the profile's ID. This ensures that although an individual user will see consistent ordering within a session, different users will see results of the same rank in different orders, to avoid favouritism towards certain profiles.

Examples

Consider a profile that has the following number of focus areas selected:

  • Disciplines: 1 out of 3
  • Verticals: 7 out of 61
  • Models: 3 out of 4
  • Stages: 4 out of 6
  • Expertise: 18 out of 28

The profile has a single location in Oxford, UK, and is open to remote work across EMEA.

  1. A query that filters on every focus area, for a remote role with a team in Germany.

    DisciplineVerticalModelStageExpertiseLocationRank
    100x1.0x0.9x0.5x0.5x0.39x0.67=5.88
  2. A query that filters only on Discipline and Stage, for a hybrid/onsite role 40km from Oxford.

    DisciplineVerticalModelStageExpertiseLocationRank
    100x1.0x0.5x0.96=48

Recommendations

  • Be selective: Choose only the focus areas where you have significant experience and expertise.
  • Consider your reach: If you prefer to work with local clients, or those with significant timezone overlap, specify your country or region for remote roles, rather than selecting worldwide availability.
  • Set just one location: Only add multiple locations if you are equally available for hybrid / onsite work in multiple locations that are significantly separated in distance (eg. London and New York).
  • Update regularly: Keep your profile focused on your current specializations and remove focus areas that are no longer central to your work.

You can try out the new ranking algorithm on our preview site. Update your profile selections and see how they affect your search ranking.