GoHighLevel CRM: How the Contact and Pipeline System Actually Works

GoHighLevel's CRM is contact-centric and deliberately simple. A contact is a person, an opportunity is a potential piece of business attached to them, and a pipeline is the sequence of states that opportunity moves through. Everything else in the platform hangs off that.

It is capable enough for most service businesses and agencies. It is also easy to structure badly, and a badly structured CRM produces reporting nobody trusts, which is worse than no reporting at all.

The data model

The relationship that matters most: one contact can hold many opportunities. A homeowner who books a repair in March and a replacement in September is one contact with two opportunities. Creating a second contact loses their history and breaks every piece of reporting that depends on customer lifetime value.

Designing pipelines that report honestly

This is where most accounts go wrong, and the fix is a single rule.

A pipeline built correctly looks like this:

Two further rules worth following. Keep it to five or six stages, because a twelve-stage pipeline is a process diagram, not a sales pipeline, and nobody maintains it. And use separate pipelines for genuinely different processes: a residential repair and a commercial contract do not share stages, and forcing them into one produces meaningless averages.

Tags versus custom fields

The distinction is simple once stated, and getting it wrong makes segmentation painful later.

Custom fields hold facts with one value. System age. Property type. Budget band. Preferred technician. If a contact could only sensibly have one answer, it is a field.

Tags hold states and memberships. Lead source. Campaign entered. Lifecycle stage. Interested in a service. A contact can carry many at once, and that is the point.

The practical test: could this contact reasonably have two of these at the same time? Two lead sources, yes, because they found you on Google and were also referred. Two property types for one address, no. So source is a tag and property type is a field.

On tag hygiene: use a prefix convention from day one. src-google, src-referral, svc-repair, svc-install, life-customer. Without it you end up with four hundred tags, three of which mean the same thing with different capitalisation, and nobody willing to clean it up.

The conversations inbox

The unified inbox is one of the genuinely strong parts of the platform, and it is underused.

SMS, email, Facebook Messenger, Instagram DMs, web chat and WhatsApp all appear in one thread per contact, in chronological order. Someone who texted in March and emailed in June is one conversation, not two systems to cross-reference.

For a business whose customers call and text, this is more valuable than most of the automation features. Being able to see the whole relationship in one place before picking up the phone changes how the call goes.

What the CRM does not do well

• No true company object. The model is contact-centric. B2B relationships between people and organisations are handled with custom fields or custom objects rather than proper associations. For a business selling to companies with multiple stakeholders, this is a genuine limitation.
• Reporting is thin. Pipeline value and conversion by stage are available. Multi-touch attribution and custom report building are not. Agencies that need real analytics export elsewhere.
• Deduplication is basic. It matches on email and phone. It will not catch the same person using two email addresses.
• No forecasting. Weighted pipeline value by stage probability is not natively available.
• Bulk editing is limited. Restructuring a messy account is more manual than it should be, which is an argument for structuring it properly at the start.

A structure that holds up

If you are setting up from scratch, this order avoids most of the rework:

1. List the facts you need about a contact. Those become custom fields. Create them before importing anything, because mapping on import is far easier than backfilling.
2. Decide the tag convention. Prefixes, lowercase, hyphens. Write it down somewhere the whole team can see.
3. Map the real sales process. Ask what states a deal genuinely passes through, then cut it to five or six.
4. Build a separate pipeline per distinct process. Not per service, per process. Repair and replacement may share one; retail and commercial usually do not.
5. Import contacts last, once fields and tags exist to receive them.
6. Attach automation to stage changes, not to tasks. A stage change is a real event you can trigger from reliably.

Every snapshot we build ships with this already done: fields, tags, pipelines and stage automation named for the industry rather than left as platform defaults.