CRM setup for a trade business: the five things worth building

Why most setups fail

Not from missing features. From having too many, configured once, never tested, and then quietly wrong. A CRM sold on a list of two hundred capabilities gets adopted as a contact list with an expensive subscription, because the part that would have earned the money was the twenty minutes of follow-up nobody wired up.

The other common failure is inheriting somebody else's template. A prebuilt set of automations looks like a head start and behaves like a liability: nobody on your team can say what any of it does, so nobody dares turn it on, and the ones that are on cannot be reasoned about. We have audited exactly this, and the answer was to delete it and build five things we understood.

The five, in build order

Order matters. Each one is independently useful, so a stall halfway still leaves you better off than when you started.

What a good follow-up ladder looks like

Five touches over a week, then stop. Immediately, an hour later, the next day, three days, seven days. Text where you have permission to text, email otherwise. Any reply or booking ends the sequence.

The last message should say it is the last message. Telling someone you will stop following up, and then stopping, is worth more than a sixth touch: it reads as confidence rather than desperation, and it is the message people most often reply to.

Three traps that are not obvious

These cost us real time to find, and none of them appear in vendor documentation.

Trigger timing against your own website

If your site creates the contact and then adds tags in a separate step, an automation triggered on contact-created can fire before the tags exist. Any branch reading those tags then reads them as absent. Trigger on the tag itself, or on the last thing your site writes, not the first.

Consent stored somewhere the CRM does not check

If you record text consent as a tag, the platform's own do-not-contact flag is still empty, so it will happily text someone who declined. Every message step needs an explicit check against wherever consent actually lives. This is the mistake with real consequences.

Tag writes that overwrite instead of append

Some contact-update endpoints replace the entire tag list. A returning customer resubmitting a form can wipe every tag your team added by hand. Check whether your integration appends or replaces before you rely on tags for anything.

What not to build

• Review gating. Routing unhappy customers to a private form and happy ones to your public profile violates Google's policies and risks the listing that brings you most of your work. Ask everyone.
• Nurture that never ends. A sequence with no exit eventually messages someone who became a customer two years ago about becoming a customer.
• Anything triggered by a stage nobody moves. If the automation depends on someone dragging a card, and nobody drags cards, it will never run. Trigger on something that happens by itself.
• Lead scoring, at this size. It is a solution to having more leads than you can call, which is a nice problem you do not have yet.

How to tell whether it is working

Not by opens, clicks, or contacts created. Those move whether or not the business does better, which is what makes them comfortable to report and useless to act on.

Measure booked work. Jobs on the calendar, and how many came from a lead that had gone quiet before an automation reached them. That second number is the one that tells you the follow-up is doing something a person was not already doing.

Test every automation with a real submission through your own website before you trust it. Not a test contact created by hand in the CRM, which skips the exact step where timing bugs live. Fill in your own form like a customer would.