One customer, three records

A customer phones with a question about an order. To answer it properly, someone opens the shop’s admin panel for the online orders, the ERP for invoices and payments, the inbox for last month’s complaint, and the marketplace panel, because the same company sometimes buys there too.

Each system holds a fragment. None holds the customer. And the person on the phone has about a minute to assemble the fragments into a relationship before the customer notices that they are being looked up.

Why the record falls apart

Identity is not shared. The same company exists as three records under three spellings, with a VAT number in one and not in the others, and an email address belonging to someone who left two years ago. No system knows that the three are one.

Companies are not people. A business customer has buyers, an accounts department, a warehouse with its own delivery address and a manager who approves larger orders. Systems built around a consumer model one person with one email, and everything else ends up in a notes field.

The awkward history lives elsewhere. Returns, complaints, late payments, the order that was shipped wrong twice. It is the part of the relationship that most affects the next conversation, and it usually sits in an inbox or a spreadsheet rather than on the record.

Channels hide connections. An order on a marketplace, under a personal name, may come from a buyer at your largest wholesale customer who needed something by tomorrow. Nothing links the two, so the customer looks smaller than they are.

What it costs

The customer repeats themselves. Every call starts from nothing, which is the clearest possible signal to a customer that you do not know them.

Decisions are made blind. A discount for someone who returns most of what they buy. Trade credit for a company with overdue invoices in the other system. A long-standing customer treated as a new one because they came in through a different door.

Reports describe someone else. Revenue per customer, lifetime value, segments: all calculated on records that split one customer into several and merge nobody. The numbers are precise and describe customers you do not have.

Simple requests become projects. Under the GDPR, a person can ask what data you hold about them. Answering honestly means finding them in every system, under every spelling.

What a real customer record looks like

One identity, with duplicates resolved. Merged deliberately, with a record of what was merged, and with the VAT number as an anchor wherever there is one.

The company and its people, separately. Who buys, who pays, where the goods go, who approves. Each a real field rather than a sentence in a note.

Every order from every channel. Shop, marketplace, phone, counter and wholesale on one timeline.

Returns and complaints on the same timeline. Not as a judgement about the customer but as an accurate description of the relationship, so that the person answering the phone knows what the customer remembers.

What was promised. A special price, a delivery date, a goodwill gesture. The things a customer will hold you to, written down where the next colleague will see them.

Why it matters for what comes next

Recommendations worth anything start from what the customer already has and what they are trying to do, not from what other people also bought. That is impossible if you cannot say who the customer is.

The same applies when something starts asking on a customer’s behalf. It will present itself as the customer and ask about their prices, their orders and what fits the thing they bought last year. Answering requires knowing who that is, which makes it a customer record problem long before it becomes an agent problem.