Special filter can be used while defining both- profile & query.

WITH PROFILE :
This gives a powerful capability to limit responses from active work seekers who pass the filter criteria you have defined (these responses are sent to you free). 
There are just a few attributes of requirements in your work position profile which are taken up for mandatory matching with the work seeker's profile and aspirations - these are : 
a) total work experience (the requirement could be nil also, for e.g. when freshers are sought to be recruited)), 
b) specific work experience in specified Function + Knowledge Domain areas with the given level of proficiency, 
c) Work locations specified (upto 4 places/regions may be offered in a work position, atleast one must match with candidate's choice to relocate to a region/city for work - this condition is ignored if option to work from home or distance is offered), 
d) Age range, 
e) Education level or degree in specified Knowledge Domains (atleast one KD out of three should match), 
f) Language proficiency and 
g) The work content in the offered position must satisfy candidate's interest, if specified ( in terms of function + knowledge domains and minimum desired concentration of time). 
Req.Service does not mandatorily match other attributes because it is a better strategy to give you the flexibility. Thus for satisfactory matching, you should set up your own additional filters (simple or special filters). Special filters are designed for combination attributes (more complex) as opposed to single attribute criteria (called simple filters). E.g. Sex, height or weight may be specified as simple filters, Combination of experience and functional areas, management levels and line of business etc. can be stipulated through special filters.

WITH QUERY:
Special filters are very useful while constructing a query. Here they can be used in three ways: 
For filter action only (i.e. include only if condition is satisfied - in this case score is nil because it is irrelevant), 
For filter + ranking purpose (i.e. include only if condition is satisfied but assign score depending upon values matched),
For ranking purpose only (i.e. assign score depending upon values matched but do not exclude if condition is not satisfied)

You can give score from 0 to 9. Since there may be multiple criteria used by you for ranking preferred values, it is necessary to give weights to each criterion (to reflect importance).

GENERAL TIP: The use of special filters reduces matched candidates more sharply than simple filters because of combination attributes, especially when AND conditions are stipulated or where choices from a large set of possible values are made for e.g. Knowledge Domain (KD). KD like some other data types is a hierarchical data set - in such cases the higher levels represent broader scope (e.g. Management) and lower levels narrower/restrictive scope (e.g. Human Resource Management or Compensation Administration). It is advisable to put broad (non-restrictive) choices in filters and / or use multiple occurrences with atleast one which is lax (without vitiating your expectations). In queries, therefore, if you are choosing a special filter for filter action, then use one occurrence which is lax ( this represents your rock bottom condition of acceptance).

You can mandate upto 3 sets* of degrees/certificates + knowledge domains (*AND). Any one out of five KDs* per degree/certificate is required (*OR). Each special filter occurrence gets linked as OR criteria. This means the candidate should fulfil any one of the occurrences but within each occurrence the set (max 3) of degrees and knowledge domains (any one out of five) become mandatory. Specify equivalent degrees allowed (e.g. B.Tech and B.E.) or not, also degree expected/finished flag values.

Upto 4 languages proficiencies can be mandated in this filter (AND). By repeating the criterion, you can specify other sets of language proficiencies which are connected by OR, i.e. any one should get satisfied in order not to be rejected. Thus if English is compulsory and either one of two optional languages proficiencies are required, then specify two occurrences in which English is repeated with one each out of the two optional languages. For ranking in query, give scores for higher levels..

A provider can filter or favour (by assigning higher scores in ranking) candidates who have expressed interest in given combination of functions & knowledge domains, the premise being candidates engaged in functional areas of their higher interest, perform better. CAUTION: Candidates are required to state their areas of interest, however, this data is not mandatory. Therefore, those candidates will get excluded by use of this filter who have not indicated any particular interest areas, so it is better to limit this criterion's usage for ranking purposes only.

Upto 5 visa sets could be mandated in this filter. You can repeat this filter with OR sense. When used with query, you can give higher score to criteria values you wish to favour and use multiple occurrences.

The candidates without experience should be filtered and ranked using academic performance criteria and this is a special filter to do that. Multiple occurrences should be used for filter/ranking especially if degrees are used instead of just education level. Remember that you need not specify all attributes - it is wise to omit the unimportant attributes - so they donot discriminate. Also try and use Knowledge Domain name at a high level in hierarchy, atleast in the instance of a filter.

Tight filter or ranking criteria. You can discriminate based on the education level or a specific degree that is obtained in a certain period from a given institute (identified by the organisation - principal and unit identifiers). Period is yymm i.e. two digits for year followed by two digits of month, e.g. 0003 to 3006 means 01Mar2000 to 30Jun2001 (period refers to degree obtained or due to be obtained in the time window). Be careful not to give too tight constraints in the filter usage.

CAUTION: Occupation titles (generic) very often donot tally with those in use in different organisations, moreover, they donot accurately reflect various various functions and knowledge domains engaged in ones work. Use this special filter only if you are pretty certain that the occupation title (or group) reflects the work content you expect the candidate to recognise/identify correctly. It is advisable to use titles as high in the hierarchy as possible.

CAUTION: Use function and KD names that are as HIGH in hierarchy as possible. When different values of Fn+KD are acceptable, give repeat instances - these are with OR sense. Multiple combinations (with AND) within each special filter put severe constraints on successful matching. Also remember discounted experience based on declared concentrations, typically is much smaller than normal experience. Always in filter usage, Fn+KD should be chosen hierarchically higher than those in ranking criteria.

CAUTION: Limit the AND usage within each special filter (max 4). Where multiple values acceptable, use multiple occurrences - these are with OR sense. Choose hierarchically higher objects in the filter criteria than in the ranking criteria.

This special filter discriminates candidates at the same management level based on the experience at that level & max. number of subordinates at any time (at that level), e.g. the nos. of subordinates can vary across the types of organisations and also within the same organisation for a given management level. This is an indirect measure of the responsibility of leadership, co-ordination and manpower management - use as filter(s) with minimum acceptable values and ranking with higher values.

Interpretation: The experience is supposed to be compared at the specified management level and at that level, the candidate should have worked in the specified Line of Business - the candidate's LoB should be same as or within the hierarchical scope of the specified LoB. If multiple values are acceptable, use multiple occurrences. Choose hierarchically higher LoBs and lower levels of management with minimum experience in the filter usage of the criteria, in ranking reverse the direction - higher scores for pinpointed choices.

Interpretation: The self-rated proficiency and sum total of experience in specified Function + Knowledge Domain should be more than a certain minimum. Also in the specified function, the candidate should have worked at/above a given management level. Be careful in specifying liberal constraints in filter criteria - i.e. higher hierarchical F& KD and lower management level choices, however, reverse direction for ranking - with higher scores for pinpointed choices.

The candidate should have attained a certain management level in atleast one of the specified sector (types of organisations). With filter usage, give lower management level and multiple sector choices, however, with filter + ranking usage give higher scores for pinpointed choices of management levels & sector types.

You can filter or rank Individual seekers who support a worthy cause with a specified level of involvement/commitment.

Age rule relaxation for special categories or handicapped categories of job applicants can be declared through this filter. The standard age requirement is declared in the position profile - the data here is to incorporate exceptions.

Application fees can be relaxed for specified special categories and physically handicapped categories.