Fix fragmented customer data in three steps. Name one authoritative source per data domain before you buy anything. Connect those sources on deterministic keys you can trust, with clear write rules. Then appoint one named owner for quality across every touchpoint. Ownership first, tooling second. That order is the whole fix.
Most silo projects start with a purchase and end in disappointment. This one starts with a decision that costs nothing but nerve. You already have the systems. What you lack is agreement on which one wins when they disagree. Here is the sequence, with owners and what to measure at each step.
Step 1: pick the master for each data domain
Do not look for one system to rule everything. Look for one authoritative source per domain. Transactions own one system. Identity owns another. Consent owns a third. Each domain gets a single home, decided on purpose, written down.
Start with the domains that carry revenue and risk. For transactions, the master is where billing sits, because billing is where identity is verified with money. It holds the cleanest link between a person and what they actually did. For identity, the master is your CRM spine. For consent, it is whatever system your legal obligation actually flows through, not wherever a marketer last ticked a box.
Resist the reflex to buy first. A CDP bolted onto fragmented sources becomes another partial store (Netguru, 2025). It reads the same conflicting inputs and hands them back faster. The reason to decide ownership before tooling is simple: 68% of data leaders name silos their top concern, up 7 points year on year (DATAVERSITY, 2024). Spend is rising and the problem is widening. Tools are not closing the gap. Decisions are what close it.
Who does it: the data leader, with the business owners of each domain in the room, because this is a call about authority, not architecture. What to measure: the count of data domains with a named authoritative source. Aim for 100% of your revenue-critical domains assigned, even where the honest answer is “we need to move this later”.
Step 2: get the connecting keys right
A master is only useful if you can connect the other systems to it. Connect on keys you can trust, not keys you can guess. Deterministic beats probabilistic every time, because a fact you can act on today beats a vendor’s confidence score you cannot.
Email is your primary key, so protect its quality at the point of capture. Company-domain email links colleagues at the same employer, imperfect but useful, so keep personal and work addresses separate rather than merging them. Precise address matching links a household. These three carry most of the load. Trust them over any probabilistic graph a vendor offers you.
Then set write rules, and this is the part teams miss. Naming a master means nothing if four systems can still overwrite the same field. For every connected field, decide which system is allowed to write it. Everything else reads that field, it does not change it. Consent updates flow from the consent master outward. Address changes flow from the transaction master outward. One writer per field, many readers.
Do not wait for every record to reconcile. Draw the Venn diagram of your systems and look at the overlap. Work the records that resolve cleanly today and use them now. Perfection is the enemy of revenue, and a single customer view that arrives in two years arrives too late to matter.
Who does it: a data engineer or integration lead, paired with the domain owners who know what each key really means in their system. What to measure: match rate on your primary key, and the percentage of connected fields that have exactly one authorised writer. Both should climb every week.
Step 3: appoint the owner
Tooling makes data connectable. It does not keep it connected. That takes a named person or team who owns data quality and governance across every touchpoint, with the master systems as their spine. Without an owner, the fix decays the same way the mess grew: one new tool, one new channel, each quietly starting its own copy of the customer again.
Give the owner real authority, not just a dashboard. They approve which system becomes the master when a new domain appears. They hold the write rules. They review match rates and chase the drift before it spreads. This is a standing job, not a project that ends. It costs an org-chart line, not a licence, which is exactly why it gets skipped and exactly why it works.
This is the move that separates a clean quarter from a durable fix.
Who does it: a named data owner or small governance team, reporting to whoever owns the revenue the data serves. What to measure: time to resolve a data-quality issue, and the number of domains still without an authoritative source. The first should fall, the second should reach zero.
How to train your team to hold the fix
A clean data spine is a moment. Keeping it clean is a habit, and habits need rules people can repeat without you in the room. The fragmentation grew because every team optimised its own tool. The fix holds only when teams learn to defer to the master instead.
Set one rule above all others: no system writes to a field it does not own. Make it a condition of every new tool that enters. Publish the domain-to-master map somewhere everyone can see, so a marketer knows exactly where consent truth lives and an analyst knows exactly where transaction truth lives. Run a short session whenever a new source connects, so the write rules travel with the tool rather than living in one engineer’s head.
The culture you are building is one sentence: decide the master, then defer to it. It runs against the instinct of every team that wants its own tidy copy. That is why it needs a leader defending it out loud, and an owner enforcing it quietly. Together those two hold the line that keeps the spine a spine.
Where Morphy helps
We run a four to eight week customer data foundation engagement. We map your systems by domain, name the authoritative source for each, and design the deterministic keys and write rules that connect them. You leave with a domain-to-master map, a working set of connecting keys, and a named owner model your team can actually staff.
The defined metric is your primary-key match rate: we baseline it across your systems and set a target for the records that carry revenue. We work the systems you already own, because the fix is a decision you have not made yet, not a platform you have not bought yet. Where a single spine has done this at scale, the reference is Tesco Clubcard at dunnhumby: roughly 12 million records connected to one master, which is what let a financial-services line go from zero to 63 million dollars in twelve months.
Go deeper on customer data maximization
Three ways forward. Pick the one that fits where you are.
- Get the playbook. Practical notes on turning the customer data you already own into revenue, straight to your inbox. Join the newsletter at the foot of this page.
- Take the assessment. Score your customer data maximization in four minutes and see your top revenue blockers. Start the assessment →
- Book a meeting. Bring your data problem. Leave with a prioritised fix, not a platform pitch. Book a call →
The playbook companion to Why is your customer data fragmented, and how do you fix it?. Post 1 of 25 in the Customer Data Maximization series.
Frequently asked questions
What is the first step to fix fragmented customer data?
Designate one authoritative source per data domain before you buy anything. Transactions own one system, identity another, consent a third. 68% of data leaders name silos their top concern, up 7 points year on year (DATAVERSITY, 2024). The fix is an ownership decision first and a tooling decision second.
Will a CDP fix my data silos?
Not on its own. A CDP bolted onto fragmented sources becomes another partial store (Netguru, 2025). It reads the same conflicting inputs and inherits their conflicts. Decide which system is authoritative for each domain first, then use tooling to connect them with clear write rules.
How do you connect systems once you have named a master?
Connect on deterministic keys you can trust. Email is the primary key. Company-domain email links colleagues at one employer. Precise address matching links a household. Set clear write rules so only the authoritative system for a field can update it. Everything else reads, it does not write.
Who should own data quality across the business?
One named person or team, accountable for quality and governance across every touchpoint. Ownership is an org-chart line, not a licence. Without it, the master you named drifts back into fragments as new tools and channels each start keeping their own copy again.