A CRM integration is not a CRM implementation

The most common CRM failure is not technical. It is a system that works perfectly, is used by nobody, and is bypassed within a quarter.

By the time that shows up in a dashboard, the investment is sunk and the conclusion drawn is that CRM does not work. The conclusion is correct; the diagnosis is wrong.

What bypass actually looks like

Store staff do not decide to reject your CRM. They encounter a friction the process has not accounted for, and they take the path of least resistance.

• The customer has no email address. The record requires one. Staff invent one, or skip the save.
• The customer already exists under a different spelling. Resolution takes four screens. Staff create a duplicate instead.
• A loyalty tier change requires a supervisor approval that takes two days. Staff note it on paper and reconcile later.
• The form is slow on the tablet in a busy shop. Staff batch it at close, from memory.

Each of these is a specification gap, not a compliance failure. And each one generates exactly the kind of incomplete, contradictory data that makes the CRM useless.

The integration was never the hard part

Connecting POS, e-commerce and loyalty systems is well-understood. Record linkage, conflict resolution and synchronisation all have known patterns.

The hard part is that integration produces a system capable of answering questions, and nothing obliges anyone to ask them.

Data quality improves when the daily workflow depends on it. If store managers do not open the CRM to do their morning trade summary, the data will drift regardless of how well the systems are integrated.

Design the workflow, not the record

The useful shift is asking what a store manager needs on a Tuesday morning, and making the CRM the fastest way to get it.

• Which loyalty tier changes need approving before the doors open
• Which customers have an unresolved delivery issue
• Which bundles are available in this store today
• Who has an open enquiry from last week

Build around those. The record model follows from the workflow, not the other way around.

Then measure the daily active rate among store staff. Not licence seats, not record counts — how many people used it, on an ordinary Tuesday. That number predicts data quality better than any integration metric.

Enablement is engineering work

Treat enablement as part of the delivery, with the same rigor:

• Observe the current process in the store before designing the replacement
• Remove fields that no one uses rather than waiting for adoption to fix them
• Make the common path one screen, and the rare path still reachable
• Pilot with two or three stores and change the design based on what they actually do
• Give store staff a route to say the workflow is wrong, and act on it

The pilot is the part teams skip. It is also the only mechanism that surfaces the friction before it is embedded across 240 stores.

Measure what changed commercially

Adoption is a proxy. The real tests:

• campaign conversion measured against a control group, not just against last month
• churn in segments defined by the new customer profile
• stock availability accuracy at order time
• reduction in duplicate customer records over time

Without a control group, every improvement gets attributed to the CRM, including the ones caused by the season.

The realistic timeline

Systems integration: weeks. Data migration and profile resolution: months. Adoption across a store network: two to four quarters, and it does not finish.

Budget and plan for that. A CRM implemented in three months and evaluated at six will look like a failure, and often is one — not of the technology, but of the timeline.