@RaviK6943 , Whenever people ask questions about context dictionary, I try to get them to step back and consider if every element being considered to be added as a context is indeed one.  The main question to answer is, am I intending to have actions specifically targeted for this context.  Will the business have communications that are ONLY relevant if perceived by the customer if they can relate it to the context.  If the answer to that is no, then consider if this data is just associative data that is important for the purpose of making a decision and therefor should be added as an association instead.

With that out of the way, for your specific example, I'd question of policy is any different than an account?  Is a policy just a type of account and therefore should it be represented as 2 context only, with a type qualifier to differentiate various types.  For example, I have credit card accounts, mortgage accounts, insurance (policy) accounts, etc.  There are likely subtypes within each of these that is important for decisioning and action logic as well.  Secondly, if policy information is specialized and needs additional data, is that data really important to be represented as context data?  Often there is a lot of special information at policy level that is important associative data, but not really important as a context.  So consider having the portion that is account level to be represented at that level and additional data that will not be used for context in a child association data entity, that way continuing to have 2 levels for context.

It looks like you know about the supported context dictionary configurations, but for completeness for others, I'll point you at this document that highlights there more explicitly https://docs-previous.pega.com/pega-customer-decision-hub-user-guide/87/understanding-multilevel-decisioning-and-context-dictionary