The two things that hold projects up

Almost every delay comes from one of two places.

Business verification. Meta checks your business is real, using documents. Days to weeks, depending on your paperwork.

Template approval. Every proactive message needs approving before first use, and first submissions get rejected more often than people expect.

Both can run in parallel with other work. Start them first, and the rest of the project fits around them.

Leaving either to the end is the most common cause of a missed launch date.

The order

Weeks 1–2: paperwork and decisions

Start business verification. Day one.

Decide what you will actually send. Order updates? Reminders? Support only?

Choose a provider.

Find a phone number not already registered on WhatsApp.

Weeks 2–3: data

Clean your phone numbers. Country code, no spaces or punctuation. This decides whether anything works, and it is entirely within your control.

Add consent fields to your customer records — separate flags for utility and marketing.

Weeks 3–4: templates

Write them as utility messages, not marketing.

Submit early. Expect at least one rejection — it is normal.

Weeks 4–5: build

Connect to your business system. One message type first.

Log every send on the customer record.

Week 6: test and launch

Test with real orders, including awkward ones.

Go live with one message type.

Watch the quality rating from day one.

The data work nobody budgets for

Cleaning phone numbers is the single most underestimated task.

Most business systems hold them inconsistently — some with country codes, some without, some with spaces, brackets or dashes, some with a leading zero.

WhatsApp needs a specific format. Numbers that do not match fail to send, often quietly.

Three things to do:

Normalise what you have. Every number, country code included, no punctuation.

Validate at entry. Stop new bad ones being created.

Make send failures visible. A failed send because of a bad number should surface, not disappear.

Do this before building anything else. A perfect integration on top of messy phone data does not work.

Consent, built in

Not a policy document — a field on the record.

Separate flags for utility and marketing. Somebody who agreed to order updates has not agreed to promotions.

Checked at every individual send, not at the campaign level.

Opt-out handled automatically. A customer replying “stop” updates the record, and every system respects it immediately.

If opt-out is manual, it will be missed — and that is how quality ratings fall without anyone meaning it.

Start with one message type

Order confirmation is the usual choice. High value, clearly wanted, clearly utility, and easy to verify.

Why not launch everything at once:

You find out whether your phone data is clean before it affects five message types.

You build a quality rating on messages people want.

One thing going wrong is a small problem.

Add the next type when the first is stable. Two weeks apart is a reasonable rhythm.

What to measure from day one

Take a baseline before you send anything.

Current “where is my order” enquiry volume. Current no-show rate, if you send reminders. Current support response time.

Without these you cannot tell whether it worked.

Then watch weekly:

Quality rating. The number that warns you before you lose access.

Blocks and reports. Rising means you are sending things people do not want.

Delivery failures. Usually a phone data problem you can fix.

And monthly: support enquiries avoided, no-shows reduced, response times.

Who owns it

Three roles, and they can be the same person in a small company — but name them.

Who watches the quality rating. Weekly, five minutes.

Who answers messages. With defined hours, stated publicly.

Who owns the technical side. Templates, integration, failures.

Unowned messaging programmes drift. Somebody sends more than they should, nobody notices the rating falling, and access gets restricted.

Common mistakes

Messaging your existing customer list. Having somebody’s number is not consent. This is the fastest way to damage your rating.

Starting with promotion. Build the rating on utility first.

No route to a person. Especially if you deploy a bot.

Unclear hours. Customers expecting replies at midnight.

Volume without capacity. Opening the channel and not staffing it.

Nobody watching the rating. The warning goes unheard until access is restricted.

What it costs

Three layers, and companies routinely budget for one.

Meta’s charges, per conversation or message, varying by category and country.

Your provider’s fee, structured differently by provider.

Your integration work, if you connect it to your systems.

Get current pricing from your provider. It has changed more than once, and any article’s figures go stale.

And budget the data cleaning. It is real work and it is usually the largest hidden cost.

A realistic first eight weeks

Weeks 1–3. Verification, provider, number, phone data cleaned. Mostly waiting and data work.

Week 4. Templates written and submitted. Expect a rejection.

Week 5. Integration built for one message type. Tested with real orders.

Week 6. Live with order confirmations. Watching the rating.

Week 8. Add dispatch and delivery updates.

Week 10. Open support conversations.

Then run it for a couple of months before adding anything else.

The short version

Start verification on day one. Everything waits on it.

Clean your phone numbers before building. Nothing works without that, and no amount of good integration fixes it.

Write templates as utility, submit early, expect a rejection.

Launch one message type, watch the quality rating weekly, and expand when it is stable.