Mapping the fashion funnel: awareness to repeat purchase

Your ad platform reports a 2.4% conversion rateconversion rateThe percentage of visitors or prospects who complete a desired action (purchase, sign-up, contact form), calculated as conversions divided by total opportunities.View full definition →. The store back office says 1.85%. The analytics property lands somewhere in between. Nobody is lying: three tools are counting three different denominators, and one of them is crediting a purchase to a click from nine days ago. Until the funnelfunnelThe customer journey from awareness to purchase, typically Awareness, Interest, Consideration, Decision, Action, with prospects narrowing at each stage.View full definition → has one agreed stage list, one event per stage and one denominator per rate, every performance meeting becomes a meeting about the data.

This lesson maps the stages a fashion shopper passes through, names the event that marks each one, and pins each rate to its denominator. What you paid for the traffic, and what the buyers are worth over time, are other lessons in this module.

The fashion funnel, stage by stage

Five gates, five events. Each one leaks, and each one has a rate that only makes sense against the right base.

*Event names follow the Segment ecommerce spec, which most CDPs and tag managers mirror. Rough industry ranges for these rates as of late 2025 exist in public sources such as the Baymard Institute; the benchmarks lesson in this module is where those numbers get argued about.*

Notice the add-to-cart row. Against sessions it reads 8.5%. Against product viewers it reads 19%. Both are called "add-to-cart rate" in the wild, and a team that mixes them will congratulate itself for a 10-point improvement that came from a change in traffic mix. Write the denominator into the metric name in your dashboard: `atc_rate_per_pdp_viewer`. It is ugly and it ends the argument.

Where the leakage happens

Add-to-cart leakage

Only 8,500 of 45,000 product viewers added something. Hesitation here is usually about price clarity, imagery, or missing information (fabric, model height, return terms).

There is a counting trap unique to apparel. A shopper who adds a medium, changes her mind and adds a large fires two Product Added events. Count events and your add-to-cart rate inflates by 10 to 15% on size-heavy categories like denim. Count distinct carts containing at least one line, and the number holds. Same fix for the size-swap in the mini cart, which some themes fire as add plus remove plus add.

Checkout leakage

8,500 carts, 3,900 checkouts started: about 54% lost before the checkout page. Baymard's long-running research puts documented average online cart abandonment near 70% across ecommerce, measured from cart to completed order rather than cart to checkout start, which is why the two figures differ. The usual causes:

• Surprise shipping costs revealed at checkout
• Forced account creation
• No preferred payment method (Klarna, Apple Pay, PayPal)
• Slow mobile checkout

The fashion-specific killer: size and fit

A large share of apparel abandonment is size and fit uncertainty. Shoppers add two sizes "to be safe", then stall. Others leave to find a size chart and never come back.

Fit anxiety reappears after purchase as a return, which is why the funnel must not stop at Order Completed. Fire an `Order Refunded` or return event keyed to the order ID, the SKU and the stated reason, then compute the version of conversion that matters:

Net conversion = orders that are kept, not orders placed. If 1,850 order and a quarter come back, you kept about 1,388. Gross conversion flatters every channel that sells fit-risky product.

Fixes worth instrumenting so you can read their effect: detailed size guides, fit-finder quizzes, "model is 178cm wearing size M" annotations, customer fit photos, virtual try-on. Vendor-quoted return reductions deserve scepticism; run the fit tool on half your PDPs and read the difference yourself.

🎬 [VIDEO: "How Fashion Brands Reduce Returns" - youtube.com - a practical overview of fit tech and return-reduction tactics in apparel ecommerce]

Engagement metrics that predict purchase

Before someone converts, behaviour signals intent. Worth a tracked event each:

• Product page depth: three or more products viewed in a session correlates strongly with buying.
• Zoom and image interactions: zoom on fabric detail is a proxy for touch, and zoomers convert higher.
• Size guide opens: a positive signal of serious intent, as long as the session does not end there.
• Wishlist adds: strong predictor of a later purchase, especially before sale events.
• Return-visitor sessions: apparel buyers commonly visit two to four times before ordering. First-session conversion is rare, which is exactly why last-click session-based conversion rates read low.

engagement_rate = sessions_with_action / total_sessions

# Example
sessions_with_action = 62000   # scrolled, zoomed, added to wishlist, etc.
total_sessions       = 100000
engagement_rate      = 62000 / 100000
# = 0.62  -> 62%

Rising engagement with flat conversion means the top of the funnel works and checkout or fit does not. Diagnose the gap instead of buying more traffic.

Instrumenting the funnel: one event, one owner

A tracking plan is a document, not a setting. It lists every event, its properties, who owns it and which dashboard consumes it. Segment, which sells the customer data layer under discussion here, structures this around two calls: track for behaviour and identify for who the person is. If identify only fires at checkout, everything before the sale stays anonymous and your repeat rate is computed on account holders only.

Three failure modes that quietly rewrite the numbers:

1. Client-side loss. Ad blockers, ITP and abandoned tabs mean browser-fired purchase events undercount the order table, often by 5 to 15%. Make the server-side order record the source of truth for the purchase row, reconcile monthly, and treat any gap above a few percent as a broken funnel rather than a rounding issue.
2. Guest checkout. If most orders skip account creation, a repeat rate keyed to customer IDs will under-read badly. Match on hashed email, and accept that a shopper who buys with a work address and later a personal one shows up as two people.
3. Off-property purchases. Glossier has sold through Sephora in North America since 2023. A customer's second purchase can happen on a shelf that fires none of your events, so your repeat rate reads lower than reality while wholesale sell-through reads fine. Zara has the same seam in reverse: an online session that ends with a store visit or a click-and-collect pickup looks like abandonment unless store transactions are joined back to the web profile.

Session definitions matter too. Most analytics tools close a session after 30 minutes of inactivity, so a shopper who browses on a phone at lunch and buys on a laptop at night generates two sessions and one order unless a user ID stitches them. Your denominator inflates, your conversion rate deflates, and no campaign caused it.

Retention: the 320 who came back

Of 1,850 buyers, 320 bought again within 12 months: a repeat purchase rate of about 17%. Say which repeat rate you mean, because three circulate:

• repeaters divided by first-time buyers in the same cohort window (used above)
• repeat orders divided by all orders in the period
• customers with two or more lifetime orders divided by all customers ever

The second flatters brands with a small loyal core; the third only ever rises. Pick one, date-stamp it, and keep it.

Two more measures belong in the plan. Time to second purchase in apparel often falls between 3 and 6 months, so a cohort read at 90 days is premature. Cohort retention tracks each month's buyers as a group: if the January 2026 cohort holds 40% at 3 months and 15% at 12, the failure is mid-lifecycle (lapsed email, no reorder prompt, no seasonal reactivation) rather than acquisition quality.

Practical retention plays in apparel

• Post-purchase fit follow-up ("how did the size work?") that feeds the size data back into the PDP
• Drop notifications keyed to previously purchased categories
• Back-in-stock and price-drop alerts on wishlisted items
• Loyalty rewards aimed at the second and third order, where the drop-off is steepest

Putting the funnel together

Read the whole thing as a diagnostic, not a scoreboard:

• Weak session to product view? Ads or landing pages mismatch intent.
• Weak product view to cart? Imagery, price or information gaps.
• Weak cart to purchase? Checkout friction or fit anxiety.
• Weak first to repeat purchase? Product satisfaction, sizing, or lifecycle marketing (or an off-property purchase you never recorded).

Fixing the widest leak beats buying more traffic. Doubling to 200,000 visits at 1.85% gives you a larger copy of the same problem, plus a bigger acquisition bill.

Key Takeaways

• One event per stage, one denominator per rate. Add-to-cart reads 8.5% against sessions and 19% against product viewers; name the base in the metric.
• Deduplicate apparel-specific noise. Size swaps fire repeat Product Added events and inflate cart metrics on denim and footwear.
• Reconcile client-side events to the order table. Blockers and ITP commonly hide 5 to 15% of browser-fired purchases.
• Track returns as a funnel stage. 1,850 orders with a quarter returned is about 1,388 kept customers, and net conversion is the number that survives contact with finance.
• Repeat rate depends on identity. Guest checkout and wholesale shelves like Glossier at Sephora both suppress it; store pickup at a brand like Zara suppresses online conversion. Stitch the profiles or read the numbers wrong.