Customer data
What is customer data and which of it do you need during a conversation?
Customer data is everything you know about a customer. During a conversation only a small part of it counts, and that part has to be there within a second.
In short
Customer data is the whole of what you record about a customer: who they are, what they bought, what contact you had and what was agreed. That data rarely lives in one system. The distinction that matters in practice is between data that lives in a source and must be current, and data you record and keep yourself.
Why a copy of your order data is a problem
Order data changes. A parcel leaves, a payment arrives, a return is approved. If that information is copied over to your service system once a night, an agent works all day with last night's picture. The customer then knows more than the agent, and that's the most expensive kind of conversation there is.
What you want is for the status to be pulled from the source when the conversation opens. Your webshop knows the order status, the carrier knows where the parcel is, your accounting knows whether it's paid. The service system is where those three come together, not where they're stored.
Which data belongs next to the conversation?
The test is the number of actions. Every piece of data an agent looks up in a second screen is ten to thirty seconds per conversation. At a thousand conversations a month that's several working days of looking things up, and it adds nothing to the answer.
- The latest orders with status
- Not just the order number, but where the order stands now and when it shipped. This covers most incoming questions at a webshop.
- Track and trace from the carrier
- The current scan, not the link you emailed at dispatch. The difference between "in transit" and "refused at the pickup point yesterday" is the entire answer.
- Open returns and credits
- A customer asking where their money is wants to know which step it's stuck in. Without this, every answer becomes a promise to call back.
- Earlier contact about the same topic
- If this is the third message about the same order, the agent should see that before sending the standard reply.
- Commitments that were made
- What a colleague promised last week determines what you can say today. This is the item most often held in someone's head instead of a system.
How do you match a conversation to the right customer?
Email is the starting point and immediately the weakest link. People order with a different address than they email from, use a work address for a private order or let a partner order. So matching on email alone leaves a share of conversations with an empty customer card.
What works better is a ladder: email first, then phone number, then an order number from the text of the message, then postcode with house number. And when nothing matches, ask. An agent asking for an order number is better off than a system attaching the wrong customer.
What the GDPR asks of you here
Customer data you show next to a conversation is personal data. That means three things you must be able to demonstrate: why you keep it, for how long, and who can access it. The last is most often forgotten in service teams, because everyone sees everything by default.
Practically: don't store payment details in conversation notes, keep special category data such as health off the customer card unless you have a basis, and make sure a deletion request is genuinely executable. A system where you can't delete a customer is a problem that only surfaces when someone asks.
Where it goes wrong
Showing everything instead of the right thing
A customer panel with forty fields is as unusable as an empty one. The agent scans and misses precisely the one thing that counts. Show five things by default and put the rest behind a click.
Showing data from a nightly export without a timestamp
If your data isn't live, state how old it is. An agent who knows the status is from last night will check it before passing it on.
Giving the AI more data than the agent
When an AI agent is connected to customer data, the same boundaries apply as for a person. Define which fields it may read and which it may never mention in an answer, and test that with a conversation where the customer explicitly asks.
Not naming a source when data conflicts
If your webshop holds a different address than your CRM, it must be settled which one wins. Without that agreement each agent picks their own and parcels go to whichever address happened to be on top.
What to do this week
Sit next to an agent and count across ten conversations how often they open a second screen, and for what. That list is nearly always short and nearly always the same: order status, track and trace, return status. Those are the three connections you build first, in that order.
What this looks like in a system
In Cuego the Hub panel next to the conversation pulls data live from the source. The order comes from your webshop or order system, the tracking status from the carrier itself, the return from the return process. Matching runs through a ladder of email, phone number and order number from the message text. The AI agent uses exactly the same data as your agents, within the boundaries you set per field.
CommerceRelated terms
WISMO
WISMO is the question of where the order is. It's rarely a location question and nearly always a certainty question, and that difference determines what you can do about it.
Read onCRM
A CRM stores who your customers are and what you agreed with them. The value isn't in the storage but in what comes out of it during a conversation.
Read onWorkflow
A workflow is a sequence of steps that runs by itself once something happens. The skill is choosing which steps deserve it.
Read on