I have a large amount of duplicate records (25.000) in an even larger table (250.000 records). I've done the detection using the reporting function. The criteria I used to detect the duplicates were 3 fields (unique account identifier, unique user identifier and date). Based on a list of criteria, I would like to either merge or delete these records. I've got a handle on the logic associated with the merging and the deleting but what I'm having trouble with is defining the duplicate detection process.

I'm looking at either creating a custom object to take care of the detection or dumping the records into an SQL database and taking care of it offline. Any thoughts on which one of the two is the best approach?


After reading the first comment, I think the more appropriate question would be "How do I approach the problem in SOQL".

The object I'm working with is the Event object and I'm really only looking at the standard fields, AccountId, OwnerId and ActivityDate. From what I've found online, there's no easy way to find records that have the same value for these three fields. I think it could be accomplished by nesting SOQL queries but I'd like to use a different approach if at all possible.

  • If you have some Apex code to post or a schema of how you intend to do what you're asking about, I think you'll be more likely to get a useful response for your duplicate detection. Remember this isn't an SQL group, so you're more likely to get a response on how to help you do it in Apex. Asking for opinions on "which is better for this, SQL or Apex?" are often difficult to support. Questions of that type tend to get closed in this group for exactly those reasons.
    – crmprogdev
    Commented Dec 16, 2013 at 13:12
  • Thanks, I've modified my question. I'd love to provide some SOQL or Apex examples but I simply haven't written them. I wanted to get some pointers on where to start first as my experience with Salesforce is still limited.
    – len
    Commented Dec 16, 2013 at 13:40

1 Answer 1


This could be base of your query for duplicates:

SELECT COUNT(Id), AccountId, OwnerId, ActivityDate
FROM Event
GROUP BY AccountId, OwnerId, ActivityDate

Don't run it yet! It's very likely to timeout and I can imagine it having serious performance issues.

This version is bit better (and it returns the earliest Event that matches the conditions):

SELECT MIN(Id) oldest, COUNT(Id), AccountId accId, OwnerId userId, ActivityDate d
FROM Event
WHERE AccountId != null AND ActivityDate = THIS_MONTH
GROUP BY AccountId, OwnerId, ActivityDate

Keep tweaking it till you're happy it can run in reasonable time. Make the WHERE clause selective, maybe increase the HAVING... The MIN(Id) is an idea for the "winner" record. Maybe you'll decide that newest one should win. Maybe you'll build something that asks the user to decide...

Anyway. Let's say that "earliest" Id will be a winner (come to thing about it, we probably shouldn't count in Id's being sequential...).

So you have List<AggregateResult> which you can use to make the detailed query for all Events:

// Id oldestEvent = (Id) results[0].get('oldest');
Id accId = (Id) results[0].get('accId'), userId = (Id) results[0].get('userId');
DateTime actDate = (DateTime) results[0].get('d');

Use these to build SELECT ... FROM Event WHERE AccountId = :accId AND ... (optionally AND Id != oldestEvent if you want to skip my "winner". Or sort them by CreatedDate / ActivityDate DESC if you want to be sure the record you want is first in the result set.

The standard merge statement works only on Accounts, Contacts and Leads. I think you'll have to delete them with your own code, deciding which values to keep in the winner, which should just go... Luckily we can't make lookups to Events so at least you don't have to worry about relationships...

  • Thanks, this is a big help. The duplicates are the result of a 3rd party application and spot checks have proven that date is not a reliable criterion to determine which one to delete. So I'll be using a range of other fields to select the winner. Good to know that merge doesn't work, I was planning on using it but it seems I'll have to write my own merge routine.
    – len
    Commented Dec 16, 2013 at 14:21
  • O, good to know. If you'll figure out some clever idea please share :) For now your best option might be ORDER BY SomeField1__c NULLS LAST, SomeField2__c NULLS LAST etc so the Events with most fields filled in will bubble to top... Or play with a formula field that will apply some kind of scoring...
    – eyescream
    Commented Dec 16, 2013 at 14:24
  • This only works with groupable fields. Commented Feb 14, 2017 at 21:55
  • @StevenStewart-Gallus true that! But it solved the task at hand, the integration must have been writing to these fields. And this question is 3+ years old :) there are modifications and different approaches available (unique external id field populated from workflow ? duplicates rules? offload the main work to report + analytic snapshot? external DB?)... But to just get some quick&dirty results such query could be good enough to prototype the final solution.
    – eyescream
    Commented Feb 15, 2017 at 10:20
  • 1
    @eyescream Thanks Yuo a lot. it helped me in my situation. Commented Dec 7, 2021 at 10:45

You must log in to answer this question.

Not the answer you're looking for? Browse other questions tagged .