I am trying to dedup the below data using an LLM prompt

    Product Line:DTVs, Sales Channel:INDIRECT , gross adds =51  
    Product Line:BYOD, Sales Channel:ONLINE2, gross adds =100 
    Product Line:BYOD, Sales Channel:ONLINE1 , gross adds =200  
    Product Line:BYOD, Sales Channel:ONLINE3, gross adds =400 
    Product Line:BYOD, Sales Channel:ONLINE4, gross adds =500 
    Product Line:BYOD, Sales Channel:null, gross adds =300

dedup criteria:
1.identify all the Sales Channel=null records and
see if they are duplicate
2.they are duplicate if the same product
line and sum of gross adds of any other sales channels equals the
null records gross adds.

========================================================================

example:

for Sales Channel:null, gross adds =300,

Sales Channel:null, gross adds =

Product Line:BYOD, Sales Channel:ONLINE2
+
Product Line:BYOD, Sales Channel:ONLINE1

, where 300==300

So I need to identify Sales Channel:null, gross adds =300 as a duplicate

how to effectively create LLM prompts for such problems ?
I also need to identify trends from similar data sets, whats the effective approach ?
I am planning on using Mistral/DBRX models.