Duplicate Detection
Automatically catch duplicate tasks with configurable rules, then compare, link, or dismiss candidates before anyone works the same request twice.
Two tasks can describe the same real-world request — an invoice submitted twice, a customer issue logged through two different channels, or the same order entered by two team members. MergePoint's Duplicate Detection & Resolution catches these automatically, so your team never processes (or pays) the same request twice.
1. Set up duplicate rules for a task type
Duplicate rules live in Form Studio, on the task type you want to protect:
Open the task type in Form Studio.
Go to the Duplicate Detection panel.
Add one or more rules. Each rule can be:
Field-based — compare specific fields using Exact match, Loose match (ignores case and extra whitespace), or Fuzzy match (a configurable similarity threshold, default 70%).
AI similarity [comingsoon] — compares the overall content of the task using AI, for requests that don't have one clean field to match on (like free-text descriptions). Choose a sensitivity: Lenient, Balanced, or Strict.
Combine multiple rules with AND, OR, or a custom combination (e.g. "Rule 1 AND (Rule 2 OR Rule 3)").
Optionally add a guard condition — an "applies only when" filter, such as "created within the last 90 days" — so the rule only runs where it makes sense.
For field-based rules where every condition is an exact match, you can turn on Auto-link. New tasks that match are linked to the oldest open match automatically, without blocking creation or requiring anyone to review them.
2. Detection happens automatically
Once rules are configured, every new task is checked against your open queue the moment it's created — no manual search required. If a task matches an existing open task under your rules, it's flagged as a possible duplicate.
3. Review flagged duplicates
When a task has possible duplicates, a Similar tasks banner appears on the task. It shows the most relevant match by task number and name, plus a "+ N others" count if there's more than one candidate.
From the banner, you can:
Open the comparison view to review the candidates side by side.
Dismiss all if you've determined the flagged tasks are unrelated — dismissed candidates won't resurface for that task.
4. Compare and resolve
Opening the banner brings up the duplicate comparison dialog, showing up to three candidate tasks side by side:
Fields relevant to your duplicate rules (like invoice number, amount, or vendor) are automatically lined up for comparison.
Click a card to mark it as the master — the record that should stay open. The other cards are marked "will be marked duplicate."
Click Link to confirm: the non-master tasks are closed and linked as duplicates of the master, so anyone opening them later can see exactly which task they were merged into.
Click the unlink icon on any card to remove it from the set without linking it — useful for a false positive. If more than three candidates exist, the next one in line takes its place.
Made a mistake? Use Undo immediately after linking or dismissing.
Linking never deletes or overwrites data — it closes the duplicate and cross-references it to the record you kept, so the full history stays intact and reversible.
Best practices
Start with field-based rules on your most reliable identifier (invoice number, order ID, ticket reference) before reaching for AI similarity.
Use AI similarity for free-text requests — support tickets, general inquiries — where there's no single field to match on.
Reserve Auto-link for exact matches you fully trust. Anything with judgment involved (fuzzy or AI matches) should go through the compare-and-link review step instead.
Add a guard condition to avoid flagging tasks that are naturally similar but not real duplicates, like recurring monthly requests from the same customer.
Was this helpful?