Omnichannel identity resolution across Shopify, Klaviyo and POS systems
Omnichannel identity resolution means knowing that the person who ordered online on Tuesday is the one who paid at your counter on Saturday. Match on email and phone first, add loyalty IDs, use guesses only for reporting, and your loyalty numbers start to line up with revenue.
This guide is for retailers and direct-to-consumer brands with a store online and a counter in person. It covers the keys that link one customer across systems, when a best guess is good enough, how to get an email or phone at the register without slowing the line, and the privacy rules that apply.
Why loyalty and revenue reports disagree
Each system counts a different slice of the same customers, so their totals never agree on their own:
- A POS only knows who bought when the customer gave a phone number or email for a receipt or a loyalty account. Every other sale at the counter is anonymous.
- Loyalty reports count members. Revenue reports count every sale. The gap between them is mostly the anonymous sales.
- Klaviyo credits revenue by its own attribution window, and it cannot see sales outside the store it is connected to. A shopper who clicks an email and then buys in person is invisible to it.
- Many POS systems send the day's sales in one batch at closing, while Shopify records each order the moment it is placed. A report run at noon compares a full day online with yesterday in store.
- Refunds and returns land in different places. An online order returned at the counter can count as a sale in one system and a refund in another.
Identity resolution fixes the first three. The last two are timing and bookkeeping, so line up the dates and the refund rules before you blame the matching.
Deterministic keys: email, phone, loyalty ID, card token
Deterministic matching links records only when they share an exact key. It is the safe default for POS and ecommerce customer matching, because when it says two records are one person, they are. These are the keys, strongest first:
- The best shared key. Shopify and Klaviyo usually have it, and a POS has it when the customer asks for an e-receipt or joins loyalty. Lowercase and trim it on both sides before you compare.
- Phone
- The key most likely to be given at a counter. Put every number into one international format first, or the same number written three ways looks like three people.
- Loyalty ID
- Exact inside your loyalty program, and useless outside it until it is tied to an email or phone. Make the online account and the in-store card the same account.
- Card token
- Links purchases made with the same card inside one payment system. Tokens from different processors do not match each other. Treat it as an internal key, check your processor's terms before using it for marketing, and never store the card number.
Clean the keys before you match
Most missed matches are formatting, not missing data. “Jane@Example.com ” and “jane@example.com” are one person. So are “(617) 555-0100” and “+16175550100”. Lowercase and trim emails, put phones into one international format, and only then compare. Matching on phone as well as email catches the customers who never gave an email.
Probabilistic matching and its acceptable error rate
Probabilistic matching links records that look like the same person without sharing an exact key: the same name and postcode, the same shipping address, a similar pattern of visits. It finds more matches. It is also sometimes wrong, and the two kinds of wrong cost different things:
- A false match merges two people into one. One customer looks far more valuable than they are, and the wrong person can get a message meant for someone else.
- A missed match splits one person into two. Omnichannel customers are undercounted, but nobody gets the wrong message.
How much error is acceptable
There is no single safe number. Set the tolerance by what the match is used for:
- For totals and trends, some error is fine, as long as you know roughly how much. Check a sample of guessed matches by hand each month.
- For anything a customer sees, like emails, offers or account details, use exact keys only. A false match there is a privacy problem, not a reporting one.
- Keep guessed matches labeled as guesses, and never merge the underlying records for good. You can undo a label. You cannot easily undo a merge.
Capturing identity at the register without hurting throughput
The register is where most identity is lost, and also where a long checkout costs you most. To unify in-store and online customer data without slowing the line:
- Ask for a phone number. It is quicker to say and type than an email, and it doubles as the loyalty lookup.
- Offer an e-receipt. It gives the customer a reason to share an email, and it ties the sale to the address they use online.
- Ask once, at the end, in the same words every time. A short, steady script is faster than a different question at each till.
- Let customers finish sign-up later. A QR code on the receipt can collect the rest at home.
- Ask at returns and pickups too. Those customers often already have an online order with their email on it.
Consent, PII handling, and regional constraints
Matching uses personal information, so the rules that govern it apply. This is a summary of common practice, not legal advice. Check the details for your locations with your lawyer.
- An email given for a receipt is not permission to market to it. Keep a separate record of who agreed to marketing, by channel.
- Marketing texts in the United States generally need the customer's prior written agreement. Treat text consent as its own thing.
- In the EU, the UK and Canada, marketing email generally needs consent, with limited exceptions for existing customers in some places.
- Several US states give customers the right to see and delete their data and to opt out of its sale or sharing. Matched profiles are part of that data.
- Collect only what you use, limit who can see full profiles, and hash emails and phones before you share them with any ad platform.
Measuring what omnichannel customers are actually worth
Once records are matched, the value question is simple to ask and easy to get wrong.
- Report your match rate first: the share of in-store revenue tied to a known customer. Every other number rests on it.
- Compare customers who buy in both channels with those who buy in one, over the same period, by revenue per customer and repeat visits.
- Be careful with the result. Your most loyal customers are the ones most likely to shop both ways, so some of the gap is who they are, not what the channels did.
- Compare people who started the same way, such as first-time online buyers, and see which ones later came into the store.
- Check each tool's own credit against matched sales. If Klaviyo claims a sale, see whether the matched customer bought online, in store, or at all.
askotter matches these records the way this guide describes: emails lowercased and trimmed, and phone numbers put into one format before matching. The Shopify connection is live today. Klaviyo and POS connections are on the integration roadmap. Point-of-sale data comes in through the vendor's API where one exists, or a daily export. See Shopify + POS systems, Shopify + Klaviyo and POS systems + Klaviyo for what each pair lines up.