Every D2C brand says they personalise. Showing someone's first name in an email subject line is not personalisation. Showing a different product to someone based on what they've browsed is closer. Rebuilding the entire store experience around what you know about a specific customer — that's personalisation. Here's how to actually build it.
The personalisation hierarchy: from easy to powerful
Think of personalisation as a hierarchy. Most brands operate at levels 1-2. The real revenue opportunity is at levels 3-4.
Level 1 — Name and order history: First name in email subject lines, order status references. Everyone does this. Marginal lift. The baseline.
Level 2 — Segment-based messaging: Sending different emails to "new customers" vs "repeat buyers" vs "lapsed customers." Better than nothing, but you're still addressing groups, not individuals.
Level 3 — Behaviour-based personalisation: Showing different product recommendations based on what a specific customer has viewed, bought, and interacted with. This is where real lift begins. A customer who's bought a protein supplement and viewed BCAAs gets a different homepage experience than a customer who's bought multivitamins and viewed omega-3s. The same store, different experience.
Level 4 — Predictive personalisation: Using AI to predict what a specific customer wants before they've expressed it — based on patterns from similar customers, their own usage history, and timing signals. Showing a customer a product they haven't searched for yet because your model predicts they're ready for it. This is the frontier for D2C.
The data foundation you need before personalisation makes sense
Personalisation is only as good as the data feeding it. Before implementing any AI personalisation layer, audit your data quality:
Customer identity resolution: Do you have a reliable way to identify returning visitors? Anonymous browsing data is nearly useless for personalisation without being connected to a known customer profile. The practical solution: first-party data collection at every reasonable opportunity — quiz completion, email signup, wishlist, review submission. Each creates a cookie→email link that makes your visitor data actionable.
Clean purchase history in Shopify: Your product taxonomy needs to be consistent for purchase history to be meaningful. If the same category has been labelled three different ways in Shopify ("supplements," "supps," "health products"), purchase history analysis produces confused signals. Audit your product metafields and collection structure before relying on purchase history for recommendation logic.
First-party preference data: Explicit preferences (quiz answers, wishlist items, saved sizes) are more valuable than inferred behaviour for early-stage personalisation. They're also more defensible as third-party data becomes less available. Invest in collecting them.
Meaningful interaction history: "Added to cart but didn't buy" is a strong signal. "Viewed product page for 45 seconds and left" is a moderate signal. "Loaded the page and bounced in 2 seconds" is noise. Your analytics setup needs to distinguish these.
The personalisation stack for a Shopify D2C brand
The practical implementation for a brand doing ₹5Cr–₹30Cr/year on Shopify:
Email personalisation (highest ROI, lowest complexity): Klaviyo with dynamic content blocks. Build a single email template where product blocks swap based on customer segment (purchase history + predicted next category). A "product you might love" block that shows skincare to skincare buyers and supplements to wellness buyers — in the same campaign, from the same send. This is Klaviyo 101 but most brands aren't using it. Setup: 2-3 days.
On-site recommendation personalisation (medium ROI, medium complexity): Rebuy or Nosto for personalised product recommendations. Homepage "recommended for you" section that varies per visitor. Post-add-to-cart upsells that are contextual to what's in the cart. Setup: 1-2 weeks. Cost: ~1% of influenced revenue (Rebuy's model) or flat monthly fee (Nosto).
Quiz-led personalisation (high ROI for the right categories): For beauty, wellness, supplements, food/beverage — a quiz that maps customer inputs to a personalised product set. The quiz data feeds into Klaviyo for personalised ongoing communication. The key investment is in the recommendation logic quality and the quiz UX. Setup: 3-4 weeks for a well-built implementation.
Post-purchase personalisation (high ROI, low implementation cost): The most underinvested area. A customer who just bought X is a warm audience for Y and Z if the recommendation is right. The Klaviyo post-purchase flow with dynamic product recommendations based on the specific item they bought is setup in a few hours and compounds with every purchase.
How to measure personalisation lift accurately
The right way to measure personalisation is with a controlled holdout group — a segment of customers who receive the non-personalised experience — compared against the personalised segment over the same time period.
Without a holdout group, you're measuring correlation, not causation. Sales go up after you implement personalisation — was it the personalisation, or was it better ad creative you launched the same week? Or seasonal demand? A holdout group controls for all external variables.
Practical holdout methodology for most D2C brands: When launching a personalisation feature, randomly assign 20% of your audience to "control" (sees the non-personalised experience) and 80% to "test." Run for 30 days. Compare average order value, conversion rate, and email revenue per subscriber between the two segments.
The metrics to measure: for on-site personalisation, focus on AOV and items per transaction (personalisation should increase basket size). For email personalisation, focus on click-to-purchase rate (personalised emails should drive higher-quality traffic, not just higher open rates). For quiz personalisation, quiz-to-purchase conversion rate is the primary metric.
Be patient: personalisation systems improve over time as more data accumulates. The lift from a recommendation engine in month 6 is typically 40-60% higher than the lift in month 1. Don't judge a personalisation investment only on first-month results.
FAQs
What personalisation level are most D2C brands stuck at? Levels 1–2 — first-name email personalisation and broad segment-based messaging. The real revenue opportunity sits at levels 3–4, behaviour-based and predictive personalisation, which most brands haven't built yet.
What's the highest-ROI, lowest-complexity personalisation to implement first? Email personalisation via Klaviyo dynamic content blocks — a single template that swaps product blocks based on customer segment, built in 2–3 days, using infrastructure most brands already pay for.
How do I know if personalisation is actually working, not just correlating with other changes? Run a controlled holdout — randomly assign roughly 20% of your audience to a non-personalised control group and compare AOV, conversion, and email revenue per subscriber against the personalised group over the same 30-day window.
How long before a personalisation system shows its real lift? Be patient through at least month 3–6 — the lift from a recommendation engine in month 6 is typically 40–60% higher than month 1, as the underlying data accumulates and the model improves.
Our take
"We personalise" almost always means level 1 or 2 in practice — a first name in a subject line, a segment-based send. Real personalisation, the kind that measurably moves AOV and retention, lives at levels 3 and 4 and requires a genuine data foundation most brands haven't built. Start with email personalisation via Klaviyo dynamic blocks; it's the fastest path to level 3 with infrastructure you likely already have. If your "personalisation" is really just a first-name merge tag, talk to us — there's a lot of revenue between where most brands are and where the technology already allows.
If any of this sounds like your situation, talk to us. We'll tell you exactly where your revenue is leaking and what it would take to fix it. Explore Strategy & Consulting →