A single customer record is the archive in which the online store, the ERP and the CRM recognise the same person as one customer, with one system allowed to correct their data. In 2025 only 41.9% of Italian SMEs analysed their own business data, against 83.6% of large companies (Istat, the Italian national statistics institute, "Imprese e ICT", reference year 2025, press release of 15 December 2025, companies of 10–49 employees vs. large companies) — and it's hard to analyse data that three different systems tell three different ways.
The case is always the same, only the names change. A customer buys on the online store as a guest and becomes a record with a name and a delivery address. The same customer asks for an invoice and the ERP creates a second one, keyed to the VAT number. Then they call about a large quote and the salesperson enters them in the CRM from scratch, because neither of the earlier archives ever crossed their desk. At that point the company has three customers where there is one, and none of the three knows how much that customer has actually bought.
It isn't a software problem: each of the three systems works as it should. It's that nobody decided which of the three has the last word when the data disagrees. Once that's settled, marketing stops being a line of spend that generates scattered leads and becomes the lever that brings measurable orders to a customer record that keeps count — the difference between spending on advertising and knowing what that spend produced.
What "a single customer record" means
The customer record is the set of data that identifies a customer: company or personal name, VAT or tax number, addresses, contact details, commercial status. When the same person exists with slightly different data in several systems, you have duplicate records. The technical term for the remedy is master data management: the discipline of establishing a single "true" version of each piece of reference data and distributing it to the other systems. The system that holds that true version is the master system for that field: the only one allowed to change it, while the others receive it read-only.
Why does the same person become three different customers?
Each system creates the record the moment it needs one, with whatever key it has at that moment — almost never the same as the other two:
| System | When it creates the customer | Key it usually uses |
|---|---|---|
| Online store | At the first order, often as a guest | |
| ERP | At the first invoice | VAT / tax number |
| CRM | At the first sales conversation | Company name typed in by hand |
None of the three keys is wrong in itself: the problem starts when three different keys stay attached to the same person, with nothing to match them up. It's the same knot described in the article on how to integrate CRM and ERP: there too the answer runs through choosing a master system, but here the subject is the customer record itself, not the flow between two systems.
Which fields have to stay identical across the three systems?
Not every piece of customer data needs the same discipline. Some fields have to be identical everywhere; others can stay local to one system:
| Field | Recommended master system | Why |
|---|---|---|
| Company name, VAT/tax number | ERP | It's tax data: it has to be correct by law before it's correct for marketing |
| Deal status | CRM | It's the only system designed to represent it |
| Delivery address | Online store | It changes most often and the customer keeps it up to date there themselves |
| Email and phone | The system of first contact | One collection point beats three versions to reconcile |
How to build it, in practice
- Choose the unique key. VAT or tax number, not the email: a person can have several email addresses, rarely more than one tax identity.
- Designate the master system for the customer record. Usually the ERP, because tax data originates there and carries legal constraints; the CRM and the online store receive it read-only.
- Export the three archives and compare them on the chosen key, to count how many real customers are hiding behind how many records.
- Merge the duplicates. Below a few hundred customers you do it by hand, row by row; above that you need a matching rule (same VAT number, or same email domain plus same postcode) and human review only on the doubtful cases.
- Switch on one-way synchronisation from the master system to the other two, on a sample first, then across the whole record base.
- Close the side door: stop the CRM and the online store from creating a new customer from scratch when that customer already exists in the ERP with a valid VAT number.
When it isn't worth it
This is my judgement, not a figure: if the company has a few dozen active customers and the people who sell know them by heart, a formal single-record project is probably oversized. In that case a simpler discipline is enough — one spreadsheet or a light CRM as the point of truth, updated by hand — until the number of customers outgrows people's memory. A single record and master data make sense when there are already three or more systems and duplicates run into the dozens, not when the problem is simply that nobody has yet written things down in one place.
What it costs
There is no fixed price: it depends on how many systems are involved, how many records have to be compared, and how dirty they are. The spend has three parts: an analysis that maps the archives and estimates how many duplicates there really are; the merge itself, which scales with the number of records to compare and clean; and the synchronisation between the systems, which is one-off technical work plus light maintenance over time. Anyone working on this kind of project, this author included, normally quotes a fixed scope for the analysis alone before pricing the rest: it's the only way not to promise a price on a problem nobody has measured yet.
Frequently asked questions
Where do you start if the company doesn't have a CRM yet?
With the ERP and the online store: they're almost always already there and already disagreeing with each other. The CRM comes later, once the underlying record base is clean — installing it first just feeds it the same mess.
Does the unique key always have to be the VAT number?
For business customers, yes, almost always. For private customers (with no VAT number) you use the tax code where available, otherwise the combination of email and phone number, knowing that it's a weaker key and needs checking more often.
How long does it take to merge the records?
It isn't something you can generalise: it depends on the number of records and how dirty they are, not on how many people work on it. An analysis that quantifies the problem usually takes a couple of weeks; the merge itself only follows once the real numbers are visible.
What happens if two systems hold different addresses for the same customer?
The master system for that specific field wins — in most cases the online store for the delivery address, the ERP for the billing address. If it was never decided which wins, that is exactly the symptom described at the top.
In short
A single customer record isn't software you buy: it's a decision about who has the last word when the systems disagree, applied with discipline. You need a unique key, a declared master system, a real comparison between the existing archives, and synchronisation that starts from one point only. Without that step, every marketing report built on top of three different record bases tells three slightly different versions of the same truth.
It's the first of the steps that make an SME's data readable, before you even talk about automations or assistants — the same principle set out in AI in SMEs: the problem is the data. If your company already has several systems that don't talk to each other and you want to know where I'd start, the systems analysis is designed to map exactly this before proposing any solution. The first step is a thirty-minute call to tell me which systems you use today.
Sources
Istat, "Imprese e ICT", reference year 2025, press release of 15 December 2025 (figures on business data analysis, companies of 10–49 employees and large companies, comparison with 2023). There is no official national statistic on the share of SMEs with duplicate customer records across different systems: where no verified figure is available, this article says so explicitly and separates Istat data, illustrative example and the author's own assessment.
Sergio Selvelli, 17 September 2026.