(Classic) Searching and Comparing Profiles

Overview of the Scoring Calculation

This topic discusses how the system calculates scores for target profiles to determine the profiles that best match the search criteria. Search and compare profiles uses the search engine to determine which profiles most closely meet the specified criteria. The search query is constructed of search elements generated from each profile item property in the search criteria. A full text search is performed to return matching profiles with a score which represents the match percentage. The scores are displayed in descending order with the profile having the best fit listed first.

To calculate the score for a target profile, the system calculates:

• Property scores for each property associated with an item.

  The search criteria comprises a set of profile items that you want to search for in the target profile. However, items in the search criteria can have multiple properties. For example, if you add a competency to the search criteria you can also specify the desired rating for the competency, or if you add a language skill you can specify the desired writing proficiency for that language. The competency rating and the writing proficiency are properties of the competency and language skill profile items. The system calculates a property score for each property that you specify in the search criteria and these contribute to the item scores.

• Item scores for each item in the search criteria.

  The item score is an average of the property scores for the properties associated with an item.

• Weighted item scores based on the Importance field.

  The Importance field is used to denote the relative importance of each item in the search criteria. The system applies a weighting to item scores that reflect the Importance values assigned to the items. The default for the Importance field is 3 - Average, but this can be changed.

• The sum of the weighted item scores.

  The overall profile score is the sum of the weighted item scores. This is the value that you see on the Search Results page.

Profile scores are also determined by the setting of the Mandatory check box. This is discussed later in this topic under the Calculating Item Scores for Mandatory Items section.

Calculating Property and Item Scores

The property score varies according to the property. For most properties, profiles with matching properties score 100 and those that don't score 0. However, properties with a rating scale such as the rating property and the interest level property have scores that are in the range 0–100 according to how closely the source and the target match. This is discussed later in this topic in the Calculating the Rating Property section.

To calculate the item score the system calculates the average of the property scores as follows:

In this example, the source profile contains three profile items for these content types:

For the degree item BA (Bachelor of Arts), the system searches for two properties, the item ID and the country. Therefore, the score for the degree item is calculated as follows:

This means that only profiles with the degree BA and country USA have an item score of 100.

To show how the system calculates the property scores, this example compares the source profile with three person profiles, for Nicola Edwards, Rachel Norris, and Helen Scott.

This table lists the items in the person profile for Nicola Edwards:

This table lists the items in the person profile for Rachel Norris:

This table lists the items in the person profile for Helen Scott:

This table lists the item scores for each person profile:

Calculating Item Weighting Based on the Importance Field

The system automatically includes the Importance field for each item in the search criteria and assigns the default value of 3-Average, which means that each item is of equal importance. Depending on the search type, you can override the default importance to specify the relative importance of each profile item. When you run the search, the system applies a weighting to each item score according to the importance that you assigned to that item. The result is that profiles that include the most important items are assigned a higher score than profiles that only include the least important items.

The system calculates the weighting for each item as follows:

To show how this works, suppose you specify the following Importance values to items in the search criteria:

In this example, the total importance points is 8 = 4 + 3 + 1

This table shows how the weighting is calculated for each item in the search criteria:

The system uses the item weighting value to calculate the weighted item score using this formula:

Using the person profiles from the earlier examples you can see how the weighting affects the overall score. This table lists the item scores and the weighted scores for Nicola Edwards:

Nicola's overall score is the sum of the weighted item scores:

Total score = 50 + 12.5 = 62.5 (63 rounded)

This table lists the item scores and the weighted scores for Rachel Norris:

Total score = 37.5 + 12.5 = 50

This table lists the item scores and the weighted scores for Helen Scott:

Total score = 25 + 12.5 = 37.5 = 38 rounded

Calculating the Rating Property Score

If the search criteria includes items with the rating or interest level property (JPM_RATING1, JPM_RATING2, JPM_RATING3, or JPM_INTEREST_LEVEL), the system calculates this property score differently than other properties. Instead of scoring 100 or 0, the system assigns a score that depends on how closely the rating or interest level in the target profile matches the rating or interest level specified in the search criteria.

The following use cases illustrate the scoring methodology which is tied to the profile usage (PERSON and JOB) of the search source and target profile types:

1. Person looking for a job match:

   • If target (job) profile has a matching item with a higher rating, then score < 100%.

   • If target (job) profile has a matching item with the same or lower rating, then score 100%.

   • If the target (job) profile has a matching item with a lower rating, then the “person overqualified” symbol will be displayed on the Compare page for that target profile item.

2. Job needs to be filled with the best person match:

   • If target (person) profile has a matching item with the same or higher rating, then score 100%.

   • If target (person) profile has a matching item with a lower rating, then score < 100%.

   • If the target (person) profile has a matching item with a higher rating, then the “person overqualified” symbol will be displayed on the Compare page for that target profile item.

3. Look for people with certain qualifications: (uses fit-gap scoring method)

   • If target (person) profile has a matching item with the same rating, then score 100%.

   • If target (person) profile has a matching item with a lower or higher rating, then score < 100%.

   • The Compare page will not display the “person overqualified” symbol in the legend since this does not apply.

4. Look for jobs with certain criteria: (uses fit-gap scoring method)

   • If target (job) profile has a matching item with the same rating, then score 100%.

   • If target (job) has a matching item with a lower or higher rating, then score < 100%.

   • The Compare page will not display the “person overqualified” symbol in the legend since this does not apply.

The interest level property is discussed in the Calculating the Interest Level Property Score section later in this topic.

The system calculates the rating property score based on the number of levels in the rating model defined for that item.

This table shows an example of a rating model and the rating levels that are assigned by the system for the search and compare:

When you assign a rating to a profile item, the system compares the level of the requested rating to the target rating level in the search and compare. For this example, the source profile has the following item and associated rating:

This table lists target profiles that include the same item and the ratings for that item:

The system calculates the difference between the desired rating level (in the source) and the actual rating level in the target profiles as follows:

ABS (absolute value) means that the system ignores the sign of the number that results from the calculation of Desired Rating Level – Actual Rating Level. By calculating the difference in this way, if the actual rating has a higher value than the desired rating the system doesn't use a negative number in the calculation of the rating score.

This example demonstrations a search query using the Rating property. The Rating Model contains the following entries in its rating scale:

1–Little

2–Fair

3–Good

4–Excellent

Incremental weighting = 100 / 4 scale entries = .25

Search criteria is looking for profiles having a matching item whose rating is 3-Good.

1. Use Case 1: Person looking for a Job match

   This use case measures how closely the job requirements match a person’s qualifications indicating overqualification for those items where the person’s skill level exceeds the job requirement.

   The default logic for assigning the weights is as follows:

   • Find entry number in scale (i.e. 3 is the 3^rd scale entry).

   • Calculate incremental weighting for each scale entry: 100 / total scale entries = n% (i.e. 100/4 = .25).

   • If the target rating is greater than (>) the desired rating, the score calculation is:

     1 – [(Target rating – desired rating) * incremental weight]

   • If Target rating less than or equal to (<=) the desired rating, a score of 100% is assigned.

   The scores generated based on the search criteria rating value 3-Good are:

   Target rating of 1-Little = 100%

   Target rating of 2-Fair = 100%

   Target rating of 3-Good = 100%

   Target rating of 4-Excellent = 1 – [(4 – 3) * .25] = 75%

2. Use Case 2: Job looking for a Person match

   This use case measures how closely the person’s qualifications match the job’s requirements indicating an overqualification for those items where the person’s skill level exceeds the job requirement.

   The scores generated based on the search criteria rating value 3-Good are:

   Target rating of 1-Little = 1 – [(3 – 1) * .25] = 50%

   Target rating of 2-Fair = 1 – [(3 – 2) * .25] = 75%

   Target rating of 3-Good = 100%

   Target rating of 4-Excellent = 100%

   • Calculate incremental weighting: 100 / total scale entries = n% (i.e. 100/4 = .25).

   • If target rating >= desired rating, a score of 100% is assigned.

   • If target rating < desired rating, the score calculation is:

     1 – [(desired rating – target rating) * incremental weight]

3. Use Case 3 and 4: Person compared to other people or Job compared to other jobs.

   This use case measures how closely the qualifications of several people match or the requirements of several jobs match (i.e. “fit-gap” scoring).

   The scoring algorithm can be calculated as follows:

   • Calculate incremental weight = 100 / total scale entries = n% (i.e. 100/4 = .25).

   • Score = 1 – (ABS(target rating level – desired rating level) * incremental weight)

   The scores generated based on the search criteria rating value 3-Good are:

   Target has rating = 1-Little: 1 – (ABS(1 – 3) * .25) = 50%

   Target has rating = 2-Fair: 1 – (ABS(2 – 3) * .25) = 75%

   Target has rating = 3-Good: 1 – (ABS(3 – 3) * .25) = 100%

   Target has rating = 4-Excellent: 1 – (ABS(4 – 3) * .25) = 75%

   This table lists the difference between the desired rating and the actual rating in the example profiles:

This difference value is used in the calculation of the rating score, which the system calculates using the following formula:

The Rating Weight is calculated as follows:

In the example rating model, the total number of rating levels is 5 and the rating weight is therefore 20 = 100/5.

This table lists the rating score for the example profiles:

The highest rating score is 100 and this is assigned only if the item has the same rating in both the source and the target. If the actual rating in the target is lower or higher than the desired rating, the rating score is less than 100. In the example, you can see that profile 02000 and 03000 have the same rating score even though the ratings are different. This is because the difference between the desired and actual rating is the same for these profiles.

The system applies weightings to items with the rating model property in the same way that it applies weightings to items without the rating model property.

To illustrate how this works, the next example shows how the system calculates scores for items with the rating model property. In this example, the job search criteria includes three competency items and desired rating, with importance values as listed in this table:

The target profiles for the search and compare are person profiles. This table summarizes the competency ratings for Nicola Richardson:

In this search and compare, the system calculates two property scores for each item:

• Item ID.

  The profiles that have the same competency item score 100. If the competency is missing, the item score is 0.

• Rating score for the Rating property.

The first step in the search and compare is to calculate the property scores for each competency item. This table lists the property scores for Nicola Richardson:

Next the system calculates these values:

• Total property scores.

  The total property score for each item is the sum of the property scores for item ID and rating.

• Average item score.

  This is calculated as follows:

  Total property score/Property Count

  In this example, the property count is 2 (item ID and rating).

This table lists the item scores for Nicola:

Next the system applies the weighting to the average item scores. To do this, the system calculates the weighting for each item based on the Importance assigned to the item using this formula:

In this example, the total importance is 9 = 2 + 4 + 3.

This table lists the weightings for the competency items:

The system calculates the weighted item score as follows:

This table lists the weighted item scores for Nicola:

Finally, the system calculates the overall profile score by summing all the weighted item scores:

Total Profile Score = 91 = 20+ 44 + 27

Calculating the Interest Level Property Score

If the search criteria include items with the interest level property (JPM_INTEREST_LEVEL), the system calculates the property score in the same way as the rating property. The interest level property score depends on how closely the interest level in the target profile matches the interest level specified in the search criteria:

The Rating Weight is calculated as follows:

The delivered interest level model (which is delivered as a rating model) has 5 levels and the rating weight is therefore 20 (100/5).

Calculating Item Scores for Mandatory Items

The system automatically includes the Mandatory check box for each item in the search criteria. If you select this check box for an item, the system searches for profiles that include the item and all the properties specified for that item. Profiles must include the item and an exact match for the requested item properties for the profile to be included in the search results. Profiles that have different item properties are not included.

For this example, the search criteria includes two competency items and a degree as shown in the table:

This table lists the items in the target profile and the item scores:

Basic Search Criteria

Profile Search Results

This scroll area lists the profiles that match the search criteria you selected. Click the profile name that you want to use as the source for the search and compare. Depending on the search type, the system either displays the Search Criteria page or runs the search using the selected profile. If the criteria rule for the search is Default, Display the system displays the Search Criteria page and you can modify the search criteria before running the search. If the criteria rule for the search is Default, No Display the system runs the search and displays the results on the Search Results page.

Search Filters

Filters may include profile groups, current location, and travel preferences, depending upon how the profile was set up. Use the Profile Groups group box to limit the search to selected profiles. The Profile Groups group box appears only if the target profile type belongs to searchable profile groups. Filter group boxes for current location and travel preferences may also appear if the target profile type includes this content section in the Profile Type configuration.

Filter data is used to filter the target profile candidates, but does not contribute to the overall search score. For example, you may want to find those person profiles that best match the requirements of the job Manager-Information Systems and are currently located in the Paris office. The Current Location fields filter out those candidates that are not in the Paris office and the score of the candidates is based on how well their profiles match the Manager-Information Systems job profile. If no candidates are found, then the Travel Preferences properties, such as Willing to Relocate and Take International Assignment can be used as filters to find those candidates outside the Paris office. The PERSON profile type is delivered with the Mobility content section where you can enter these preferences for a person.

See Profile Group Types Page.

Search Criteria

Fields in this section depend upon how the profile type is set up.

Search Results

Selected Profiles