Fix identity resolution in three steps. Verify the email at capture with double opt-in. Build a deterministic spine from the keys you already trust: logged-in state, transactions, loyalty, company domain and address. Then measure identity coverage as a tracked metric and use probabilistic matching only to extend the edges. Get coverage past 60% and the guessing stops.
Most identity projects start by shopping for a graph. This one starts with the keys you already hold. Without a shared key linking email, device ID, loyalty number and CRM ID to one profile, data stays fragmented no matter how many integrations you buy (CDP.com). So the work is not more plumbing. It is deciding what the key is and wiring everything to it. Here is the sequence, with owners and what to measure at each step.
Step 1: verify the email at capture
Email is the long-term key. It survives job changes, device swaps and cookie loss, so it carries more of your identity weight over time than any other signal. That means its quality decides the quality of everything you later link to it. A bad email at capture poisons the whole profile joined to it.
So verify at the front door, not in a cleanup later. Double opt-in on every capture point: forms, checkout, loyalty sign-up, in-app. A confirmed email is worth more than three unconfirmed ones, because you can build on it with confidence. And capture more than one touchpoint while you have the person’s attention. Ask for the personal email as well as the work one, and keep them apart. The personal email is the key you protect for the long game, because it outlives every job the customer will ever hold.
Owner: the person who owns capture, usually marketing operations or lifecycle. Measure: percentage of new records captured with a verified email, and the average number of verified touchpoints held per customer. Both climbing is the signal that the front door is sound.
Step 2: build the deterministic spine
Now assemble the spine from keys you can trust. Deterministic identity is the backbone: logged-in state, transactions and loyalty number are facts, not inferences. Where a customer logged in, paid, or scanned a loyalty card, you know who they are. Start there and link outward.
Add the joins you can make safely. Company-domain email links colleagues: two people on the same domain work at the same business, which is safe to infer and useful for B2B and for separating work from home. Precise address matching links a household: a shared, precisely matched address usually means one home, giving you the household view without guesswork. For families, join members into one account, but have one member invite the other so consent stays clean. Then cross-link the rest to the same profile: email, cookie, device and in-app login. One person, one profile, many signals feeding it.
Keep probabilistic matching out of the spine entirely at this stage. Build the backbone from certainty first. You extend it later, deliberately, in Step 3.
Owner: the data or engineering lead who owns the customer profile store. Measure: the count of distinct deterministic keys resolving to a single profile, and the duplicate rate falling as merges land. Fewer profiles than last month, for the same customers, is progress.
Step 3: measure coverage and supplement at the edges
You cannot manage identity you do not measure. Make identity coverage a tracked metric on the dashboard, reviewed like any other number: what share of your active customers can you confidently link to one deterministic profile. Below roughly 60% coverage, downstream attribution and personalisation produce guesses, not decisions (Improvado, 2026). That line is the target. Everything above it compounds, because every layer you build on identity inherits its coverage.
Only once the deterministic spine is measured and climbing do you add probabilistic matching, and only to extend the edges. Use it to reach the customers you cannot yet resolve with certainty, never as the backbone. Probabilistic matching degrades as signal loss worsens, and signal loss worsens every year as cookies die and consent tightens. Lean on it and your identity gets weaker over time. Lean on the deterministic spine and it gets stronger, because you own the keys.
Owner: the analytics or CRM lead who reports the metric, so coverage is a defended number, not a footnote. Measure: identity coverage as a percentage of active customers, tracked month on month, with 60% as the floor and higher as the goal. Watch the share of decisions made on deterministic identity versus probabilistic. You want deterministic carrying the weight.
How to train your team to hold the fix
A coverage number is a moment. Holding it is a habit. Identity decays the same way it fragmented: one unverified capture, one unlinked signal at a time, unless you change how identity enters and ages.
Set a rule that no new capture point ships without email verification wired in. Make identity coverage a standing line on the marketing and data review, refreshed each cycle, so a drop gets noticed the month it happens, not the year. Give the customer profile a named owner responsible for coverage, not just for keeping the lights on. And when a new source of identity appears, a new app, a new loyalty tier, a new region, treat wiring it into the deterministic spine as part of the launch, not a backlog item.
The culture you are building is simple to state and hard to keep: resolve identity on purpose, on keys you trust, and measure it like revenue. It runs against the reflex to buy a graph and move on, so it needs a leader defending it. That defence is the difference between a one-off coverage lift and a spine that stays strong.
Where Morphy helps
We run a four to eight week identity resolution engagement. We map the deterministic keys you already hold from logins, transactions and loyalty, wire email verification into your capture points, and cross-link the signals that safely join, company domain and precise address included. Then we stand up identity coverage as a tracked metric on your dashboard.
The defined metric is that coverage rate: we baseline where your deterministic identity sits today and set a target that lifts it past the 60% line where attribution and personalisation stop being guesses (Improvado, 2026). No new identity graph to licence. We build the spine from the keys you already own, because the identity is already in your data. It is just not joined up yet.
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 does identity resolution fail, and how do you fix it?. Post 2 of 25 in the Customer Data Maximization series.
Frequently asked questions
How do you build a deterministic identity spine?
Start with the keys you can trust: logged-in state, transactions and loyalty number. Link those to a verified email, then cross-link company domain, precise address, cookie, device and in-app login to the same profile. Deterministic identity carries the weight. Probabilistic matching only extends the edges (CDP.com).
What identity coverage should I aim for?
Track identity coverage as a metric and get it above 60% of active customers. Below roughly 60%, downstream attribution and personalisation produce guesses, not decisions (Improvado, 2026). Every point of deterministic coverage you add turns another slice of your data into something you can act on.
Why verify email at the point of capture?
Email is the long-term key that everything else links to, so its quality sets a ceiling on everything downstream. Verify with double opt-in at capture, not in a cleanup six months later. Fixing a bad email at the front door is far cheaper than repairing every profile joined to it.
How long does an identity resolution fix take?
The first coverage lift fits inside a four to eight week engagement, because you build on identity you already hold from logins, transactions and loyalty. You are not buying an identity graph. You are wiring the deterministic keys you already own into one profile and measuring the result.