Nine weeks before Black Friday I read twelve DTC storefronts line by line. Eight of them had a defect a shopper could see: a sold-out bestsellers grid, an error message under every buy button, a savings badge that disagreed with the price. None were hard to fix. All of them are hard to notice once the traffic arrives. So here's the short list first, then the detail, page by page.
Amais BajwaFractional growth lead27 Sep 202622 min read119 action items
Share
Most Q4 plans I get shown are media plans. Budgets, channels, creative calendars. The store those plans send people to gets a line that says "site updates" and a date in the last week of November.
That's backwards. Paid traffic at peak is the most expensive traffic you'll buy all year, and it lands on whatever the store happens to look like that morning. A cart that hides the shipping cost, a discount that stacks to a price you never agreed to, a hero product showing as sold out — each of those quietly taxes every dollar of media behind it.
8 of 12
Audited stores with a defect a shopper could see, nine weeks before Black Friday
My audit, 27 Sep 2026
46%
Say an unexpected cost is why they abandoned a mobile checkout
Tapcart 2026 survey, 1,205 shoppers
17%
Start their holiday shopping during BFCM week. Everyone else has already begun
Tapcart 2026 survey
This page is the store-side plan. It starts with the twenty things I'd do first, in plain language, with how much each one matters and how much work it is. After that comes the detail for each part of the store, with the data behind it. The audit itself is at the end, with names.
One note on the numbers. Most of the shopper figures come from Tapcart's 2026 survey of 1,205 mobile shoppers. That is stated intent from people recruited as mobile shoppers, so read it as direction rather than as a forecast for your store.
Start here
The twenty things I'd do first
If you only have time for one screen of this page, it's this one. Tick them off as you go; your ticks stay saved in this browser.
Impact is how much the task moves revenue or protects margin, out of five. Lift is how much work it takes: an easy lift is a decision, a setting or an afternoon; a heavy lift needs design or development time, so start those first.
The download holds all 119 items on this page: these twenty, the detailed checklists below and the launch-night steps, with a column to mark each one done.
The calendar
What each window is actually for. The dates assume a code freeze on 13 November, two weeks before Black Friday.
This weekDecide, then freeze the inputs
Offer written on one page with a margin floor. Peak product list frozen in a single sheet. Product titles rewritten. Duplicate collections merged. Early-access page live and collecting emails.
To 17 OctBuild the structure
Peak collections created at stable URLs. Product attributes filled in for a browse path that isn't product type. Every existing discount audited and the combining rules drawn up. Site speed measured.
18 – 31 OctBuild the experience
Gifting menu and guides built. First screen of the product page rebuilt. Cart drawer, progress bar and recommendations shipped. Wallets switched on. Sold-out states tested with a zeroed variant.
1 – 13 NovTest, then stop changing code
Gift guide on the first screen of the homepage. Worst-case cart tested and approved by finance. Tracking, consent and error alerts checked. Rollback rehearsed. Code freeze on 13 November.
14 – 26 NovMerchandise and warm the list
Gifts and sets move to the front of the menu on 20 November. Sale dates published so people can register. The offer revealed in three moments. Support briefed. Launch-night runbook written, with names.
27 – 30 NovRun it and watch it
Sale goes first in the menu. A test order checked within fifteen minutes of launch, in the US and in Canada. Stock and prices checked at each day's own peak hour. Every discount checked an hour after it ends.
1 – 24 DecSell delivery, not price
Sale becomes last chance. The announcement bar becomes shipping cut-offs. Product pages carry the order-by date. Gift cards take the hero once the ground-shipping date passes.
26 Dec – 5 JanBoxing Week and the self-gift window
The post-Christmas collection goes live on a scheduled swap. Gift-card spending gets its own merchandising. The review is written by 5 December, and the peak customers are tagged for next year.
01 · the detail
Catalogue & product data
Product data is what every other decision sits on. Get the naming and the collections wrong in September and you spend November working around it.
This is also where the AI question actually gets answered. Two-thirds of shoppers in Tapcart's survey are at least open to using AI to shop this season, and an assistant can only recommend what it can read. Clean titles and structured attributes are how a machine understands, compares and suggests what you sell.
The trap I saw most clearly: one store's product page showed its search-engine keyword string as the page heading, with the name customers actually use sitting underneath it. Anyone asking an assistant where to buy that product by name is being matched against the weaker of the two. Keep the keyword string in the SEO field and let the title be the product's name.
The other pattern worth copying is a way to browse that isn't product type. "Cleansers, treatments, moisturisers" makes the shopper do the translating. Only four of the twelve stores offered a path by need, such as concern, scent or benefit. The ones that did built it from real product attributes, which is why it scales and maintains itself.
A deep sitewide cut is usually the most expensive way to buy demand you already had.
The evidence points the same way from two directions. In Tapcart's survey, 42% of shoppers expect to spend more this BFCM and 13% expect to spend less, so the cut-back crowd is a small minority. And Salesforce's Cyber Week 2025 figures showed average selling prices up 6% with order volume still up 2%. Demand isn't the constraint. Capturing it without giving away the margin is.
The most common answer for "how much do you need" has been 20–25% in three surveys running. Roughly three in ten say they need 30% or more. And only about 5% say they won't buy from a brand they love without any discount at all.
The part that costs real money: stacking
Shopify sorts discounts into three classes, product, order and shipping, and you choose which classes may combine. Automatic discounts and codes can combine too. So an automatic discount left over from a campaign two seasons ago will sit on top of your Black Friday code unless somebody switched it off.
A single order can take up to five product or order codes plus one shipping code, and a store can have at most 25 automatic discounts running, including ones created by apps. The order of the maths matters as well: order discounts are calculated after product discounts, so two 20% discounts come to 36% off, not 40%. Model the compounding, not the sum.
Ten ways to run an offer
These are my judgements from how each offer works, not measured effects. Nobody publishes results by offer type. Use the table to shortlist two, then work out what each leaves you per order.
Offer
What it does for you
How it goes wrong
Sitewide percentage
Simple, and what shoppers expect. Worst for margin: every order is discounted, including the certain ones.
It discounts demand you already had, and trains your list to wait.
Spend tiers
The strongest lever on order value. The discount is paid for by the extra item.
Tiers copied from a competitor rather than your own order values move nobody.
Buy one, get one
The most-preferred promotion among shoppers. You give away product cost, not a share of revenue.
It can't combine with other product discounts, and it doubles your stock forecast.
Gift at a spend threshold
Works like a tier with a better story. The cost is one small product.
No cap, or a gift nobody wants.
Free-shipping threshold
Lifts the basket by about one entry-priced item.
Set from a round number instead of your own shipping cost.
Bundle or set price
The strongest per order. You chose the mix, so you chose the margin.
The bundle undercuts your hero product at full discount depth.
Mystery gift
Clears slow stock at cost.
Used as a dumping ground. A disappointing gift hurts the next campaign.
Bonus loyalty points
Best for margin: a future cost at your redemption rate, not today's price.
Invisible before purchase. If it isn't on the product page, it changes nothing.
Early access only
Concentrates the offer on your keenest buyers, with no public discount.
"Early" that is twelve hours and the same depth won't work twice.
No discount, added value
Price held. Good for the brand, weaker for volume.
Needs real pull. Without it you simply lose the week.
Free shipping has no standard
A spread of more than double across nine stores selling broadly similar things. The threshold is an output of your own shipping cost and your own order values, not something to copy. The same goes for Canada: one of these brands runs $40 in the US and $60 in Canada, which is a round local number rather than a conversion.
The cheapest conversion lever you have, and the one most teams change too late.
Hard navigation and poor search sit in the top tier of what shoppers say frustrates them on mobile. Four of the twelve stores had a gifting entry in the menu nine weeks out. The other eight were relying on the shopper to invent one.
The best pattern I found for running several offers at once was a drawer. One store folds six live offers behind a single line in its announcement bar; tap it and you get all six, each linked, with one terms page behind them. Two other stores rotate their bar messages instead, which means a shopper who reads one never learns about the other three.
That chart is why the menu has to change four times rather than once. By 20 November about 80% of shoppers have already started, so the menu's job shifts from helping people discover to getting them somewhere fast.
1 NovOpen the season
Gifts enter the main menu, with price bands and recipient guides underneath. The seasonal tile in each menu swaps to the gift guide. A Sale entry appears only when the first offer is really live.
20 NovRouting, not discovery
Gifts and sets move to the first two positions. The early-access signup becomes a menu item. The bar carries the sale dates and the signup, not a discount.
27 NovThe sale has one job
Sale first, gifts second, bestsellers third. The bar states the offer, the threshold and the end time with a timezone, and links to the terms. Live before midnight Eastern and checked on a phone.
1 DecSell delivery
Sale is renamed last chance and points at what remains. From about 10 December the bar's only job is the shipping cut-off. Once the ground date passes, it becomes gift cards.
The first screen on a phone is roughly one image and three lines of text. That's the whole budget.
Three of the audited homepages led with a product. Only one of them told a first-time visitor what they'd get: spend $50 and receive four mini fragrances. The second led with an award and the third with a launch. Both are fine things to say, but neither answers the question a stranger arrives with.
The bar is rising, too. In Tapcart's survey 91% say a great mobile experience matters, and the share calling it "extremely important" went from 41% to 47% in a year. The one trend moving in a brand's favour is trust: marketplaces fell from 66% to 61% as the most trusted place to buy, while a brand's own app rose from 43% to 47%.
A small thing that reads as a big one
Three numbers for one claim
One store's page description promised "over 100,000 customers" while its hero and its testimonials both said 150,000. On two stores, several different products showed an identical review count, because the count belonged to the whole collection rather than to each product. Specific proof only works while it is consistent, and search engines read the page description too.
Three homepages, one switch
Before the sale, the hero's job is to turn visits into permission: the dates, an email field, no discount stated. During it, the hero states the offer, the threshold and the end time as real text, with sale and gifts reachable on the first screen. After Cyber Monday, it leads with the shipping cut-off, then switches to gift cards once that date passes. Build all three as one section with a selector in the theme editor, so the change after the freeze is a setting and not a deploy.
The product page's job has changed. A lot of people now arrive to confirm, not to be persuaded.
Tapcart's argument is that by the time a shopper reaches a product page, an AI assistant has often squeezed days of research into minutes. If that's right, a page that makes its case slowly is answering a question the shopper already settled somewhere else.
I'd hold that loosely. Adobe measured visitors arriving from AI tools and found them spending more time on site and viewing more pages, which is the opposite pattern. The reading that survives both is narrow and practical: make confirming fast, without stripping the page. Treat the first screen as a confirmation panel. Everything that argues, the story, the science, the long ingredient explanation, moves below it.
The best thing I saw on any product page
A claim with its evidence attached
One store footnotes its headline claim, defines what the claim means and discloses the study behind it: a double-blind, third-party, twelve-week trial with 22 participants, plus five numbered references to published papers. That is what evidence looks like when it is made visible.
The failure I saw was the reverse. A carefully worded benefits section sat on the same page as an FAQ that made stronger claims with no qualification at all. The headline is rarely the problem. It is usually a second block of copy nobody checked.
Baymard's figure, averaged across 50 separate studies, is that 70% of carts are abandoned. The largest single reason shoppers give on mobile is a cost they didn't see coming. Both are fixable inside the cart.
Every one of the twelve stores used a slide-out drawer rather than a separate cart page. The better ones made the drawer do the selling. Two ran a progress bar with two goals, free shipping first and a gift further on, so there was still a reason to add something after the first goal was met.
The recommendations inside the cart are the highest-value lever on this page. Merchants using one recommendation platform, Rebuy, accounted for about $1.7bn over BFCM 2025, close to 12% of everything sold on Shopify that weekend. One haircare brand reported a 25.9% lift in order value across the weekend, with in-cart recommendations driving close to 60% of the extra revenue. That is one merchant and not a benchmark, but the direction is consistent.
Every missing payment method is a group of abandoned carts. Shop Pay alone carried 32% of Shopify's BFCM 2025 orders. And among Gen Z, 64% expect Apple or Google Pay against 62% who expect to pay by card, so for the youngest shoppers the wallet now outranks the card.
Two defects worth searching for tonight
Raw template code and a permanent error message
One cart showed a line of unprocessed template code where the free-shipping amount should have been. Another store printed "There was an error adding the item to your cart" under every add-to-cart button on the page, whether or not anything had gone wrong. Even when hidden from view, text like that is still read by screen readers, search engines and AI assistants.
The list you build in October is the list that buys in November.
About 80% of shoppers begin before Thanksgiving week and only 17% begin during it. Demand shows up weeks before anyone expects a deal, and that gap is the one window where you can win attention without cutting price.
Scarcity is a rising motive: the share of shoppers buying early because they fear a sell-out went from 33% to 42% in a year. But the honest version is the one that works. None of the twelve stores ran a countdown timer in late September. What they ran instead was a waitlist, a "back in stock" label in the menu and notify-me on sold-out products.
Reveal the offer in three moments
13 NovThe dates
Publish the sale window and the early-access window, with one field to register. No discount stated. It's the only countdown you can run honestly, because the deadline is real.
20 NovThe type of offer
Say whether it is tiered, buy one get one, or a gift with purchase. Still no number. This is the message that lets a shopper decide to hold their budget for you.
24 – 26 NovThe number
The depth, to early access first, with the code applied by the link. One audited store gates this to members and calls the event its only sale of the year.
One last finding, if you have an app. 42% of shoppers discover a brand's app on the brand's own website, against 9% through email or text. And 28% download before they've bought anything. The best place to promote the app is the page the shopper is already on.
Four in ten holiday purchases aren't gifts. Almost nobody merchandises for that.
Buying for yourself has gone from 31% to 40% of holiday shopping in three years, and for Gen Z it is already the majority. A plan built entirely around gifting ignores a large share of why people are actually buying. None of the twelve stores had a self-gift guide. It is the clearest open goal I found.
Build guides as collections, never as landing pages. A collection brings filters, sorting, the product feed and quick-add with it. A landing page brings none of that and has to be rebuilt every year. It is also why one store could have recipient guides and three price bands live nine weeks out, without a build sprint.
Four kinds of guide are worth having: by recipient, by price, by need, and for yourself. Name the last one the way customers talk, "treat yourself", not "self-gift".
Canada
Boxing Week is a second season
About 8.6 million Canadians, roughly one adult in four, shop Boxing Week. The reported pattern is that Black Friday rewards impulse while Boxing Day is increasingly planned. The stretch from 26 December to 5 January is gift-card spending and buying for yourself, which is the opposite merchandising job from December.
Two journeys, one store. Greeting a ten-time customer with a first-order pop-up tells them you don't know who they are.
Where
First-time visitor
Returning buyer
Homepage hero
The offer, the threshold, the returns promise, the review count
Early access or loyalty balance, a reorder path, what's new since they last came
Welcome offer
Shown. It is the cost of winning them
Hidden. Showing it gives away a discount you never needed to give
Product order
Most-reviewed first
New, and what they haven't bought yet
Product page
Returns window and guarantee near the price
Stock, shipping cut-off, subscribe price, reorder
Way in
A discovery set, trial size or sampler
The full size, the bundle, the subscription
The offer
The discount. This is what you are buying them with
Access, bonus points or an exclusive, not a deeper percentage
The right-hand column is where margin is protected. The survey's own recommendation is to spend discount budget on winning the shoppers who are close to buying, and to protect margin on the ones who were always going to.
TaskImpactLift
10 · the detail
Technical readiness & launch night
This is the section that stops a good plan being undone by a deploy.
Slow pages and a buggy experience sit in the top tier of what shoppers say frustrates them on mobile, and a site that was too slow or crashed is one of the named reasons for abandoning a checkout.
The finding that changes the plan most: across $2.88bn of tracked BFCM 2025 revenue, Triple Whale found a different peak hour on every day of the weekend, including 3am on Saturday and 4am on Sunday. So error alerts have to reach a phone rather than an inbox, and the stock-and-price check has to run four times, once a day, at that day's own peak.
The freeze I'd recommend is 13 November, covering the theme and the checkout. Content and merchandising changes carry on through the theme editor and the admin. That is the whole reason the homepage versions and the menu changes above are built as settings.
The usual failure isn't that nobody knows what to do. It's that three people do the same thing and nobody does the fourth. Put a name beside each line. For the rollback, aim to be live again within thirty minutes of deciding, and rehearse it once so that number is real.
The evidence
The live audit: twelve stores
Twelve DTC beauty and wellness storefronts, read on 27 September 2026. Everything here is what the page actually showed that day, not what a brand is known for.
Sets are everywhere. A gifting structure is rare. A self-gift guide didn't exist anywhere.
Worth copying
ILIA runs six offers at once in one expandable drawer, with a single terms page behind them.
OneSkin shows five published references on its homepage, and footnotes its headline claim with the study design and the number of participants.
Moon Juice states the source, dose and form of each ingredient, shows the cost per serving beside the price, and lets shoppers filter reviews by age and concern.
Salt & Stone has a complete shop-by-scent structure in its menu, with "back in stock" labels on menu items.
Sol de Janeiro merchandises its loyalty rewards on the homepage, priced in points.
Drunk Elephant had gift guides by recipient and by price live nine weeks out, and a bundle discount that applies without a code.
Kosas prints the number of shades on every product card, and runs a two-goal progress bar in the cart.
The Bare Home puts refills in the main menu and offers a gift note in the cart, with the character limit stated.
Defects on live sites
Salt & Stone: every product on the homepage bestsellers grid showed as sold out, including the launch the hero was promoting. The costliest defect in the audit.
Sol de Janeiro: an error message printed under every add-to-bag button on the page.
Glossier: raw template code in the cart where the free-shipping amount should be.
Teangle: prices shown in cents and internal product IDs on product cards, and three different figures for one customer-count claim.
The Bare Home: unedited theme placeholder copy live in a signup form, and one review count shown on three different products.
OneSkin: a "Save 10%" badge on prices that imply about 23%, and image descriptions naming the wrong product.
Drunk Elephant: terms from a January promotion still on the page in September.
Moon Juice: one product under three different names.
Jones Road: the menu and the cart pointing at two different collections, both called Kits.
Store by store
Store
Gifts in the menu
Browse beyond product type
Free shipping from
What stood out
Glossier
No, sets only
No
$40
The hero is a gift at $50, not a discount
Drunk Elephant
Yes, by recipient and price
Partly, editorial guides
$40
The only full gifting structure, nine weeks out
ILIA
Yes, plus a gift quiz
No
not stated
Six offers in one drawer with one terms page
rhode
No
No
$45
Early access behind a login, on the homepage
Kosas
No
No
$50
Build-your-own kit: any three, 20% off
Jones Road
No
No
$55
No discount anywhere on the site
Salt & Stone
Yes
Yes, by scent
$50
Members-only early access to its one sale a year
Sol de Janeiro
Yes, holiday sets
Yes, by benefit and scent
$35
Rewards shown on the homepage, priced in points
Moon Juice
No, gift card only
Yes, by benefit
$65
Cost per serving shown beside the price
OneSkin
No
No
$79
Published research cited on the homepage
The Bare Home
No, gift cards only
No
C$100
Refills in the main menu
Teangle
No
No
not stated
The hero button opens a bundle builder
How I did it: I read the US storefront for each brand, except The Bare Home and Teangle, which are Canadian stores. I read every homepage, and one product page in full, as each was served on the day. I didn't complete a checkout, and what a page shows can vary by location, so treat this as a snapshot.
If you do nothing else
Decide the offer in writing, with a floor. Almost every expensive mistake on this page starts with an offer nobody wrote down: the stack that went too deep, the bundle that undercut the hero, the discount that never ended.
Take the surprises out of the cart. Full price shown early, every wallet switched on, a sold-out state that leads somewhere. None of those needs a discount, and all of them compound.
Start the heavy lifts this week. There are only five in the top twenty, and they are the ones that can't be rushed in November: the first screen of the product page, the product data, the cart drawer, the recommendations inside it and the homepage switch.
Then stop changing code. The freeze is the cheapest insurance you can buy. Everything you'll want to change during the sale should already be a setting.
And spend an hour reading your own store the way I read these twelve, on a phone, as a stranger. Eight out of twelve had something broken in plain sight. The odds are that yours does too, and it's far cheaper to find it now.
I'm a fractional growth lead. I build the data layer your growth team is missing, then run media, creative and email against it. If you'd like your store read the way I read these twelve before the traffic arrives, that's a conversation I'm happy to have for free.
Amais Bajwa
Fractional growth lead