A common, quiet CRM problem is that the same person appears two or three times. One record came from the website form, one from an ad, and one was typed in by hand after a phone call. Each holds part of the story, so nobody has the full picture.
In two recent Upwork engagements I built n8n workflows that score, tag and map incoming leads into a CRM, with duplicate handling in one of them on Zoho CRM. Both clients rated the work 5.0. Here is the approach, because the same pattern applies to any practice that takes enquiries from more than one place.
Prevent duplicates, do not clean them up
The most important decision is where the duplicate check sits. It runs before a record is created, not as a clean-up job afterwards.
A clean-up job means duplicates exist for a while. Someone may already have called the "new" lead, or sent two welcome emails. Checking first means the duplicate never exists.
The pattern I follow for every incoming lead looks like this:
- Normalise the details that identify a person, such as a trimmed, lower-case email address and a phone number in one consistent format.
- Search the CRM for an existing contact that matches.
- If there is a match, update that record and add the new enquiry to it.
- Only if there is no match, create a new record.
Score with rules your team can read
The second decision is how leads are qualified. In these engagements, scoring and auto-tagging were built from simple rules the client can read and change themselves. Typical inputs for rules like these are where the lead came from, what they asked about and whether their contact details are complete.
I prefer visible rules to letting an AI model judge each lead freely, for two reasons:
- Every score can be explained. When a lead is marked low priority, anyone can see why.
- The client owns it. If the business changes what counts as a good lead, they change a rule without calling me.
AI still has a place. It is useful for summarising a long enquiry or drafting a first reply. But the decision about what happens next should be one you can read.
Map fields once, carefully
The third piece is field mapping: deciding exactly which CRM field each incoming detail goes into. It sounds boring, but it is where most "the CRM is a mess" complaints start. Getting it right once, in one workflow, means every source arrives in the same shape.
What this looks like for a practice
The engagements above were for businesses outside healthcare. A practice can use the same pattern for patient enquiries, with one extra step at the start: deciding what enquiry data you actually need, where it is stored and who can see it. Health information needs that review before anything goes live.
The result to aim for is simple: one clean, tagged record per person, with an owner, so every enquiry gets followed up exactly once.
The Lead Qualification & CRM Automation case study has the workflow and the client feedback. To see how it maps to a practice, read enquiry qualification and CRM entry for healthcare practices.