MARCH, 2026 · SKY GROUP LTD. UK / SWEETIES AT THE TILL
Sweeties at the Till - Checkout Upsell
Sky Stream’s checkout step had zero upsells. By identifying this and addressing the underlying problem, we increased the product attachment rates to almost 250% per product.
From fake door test to shipped feature. A product design case study for Sky Stream. (Sky Stream is Sky UK Ltd.’s streaming subscription platform)
+500%
Attach Rate
Ultra HD & Dolby Atmos (TV Experience) Attachment
+256%
Attach Rate
Sky Cinema (Premium TV Pack) Attachment
What is Sky Stream checkout?
Sky Stream is Sky’s TV streaming subscription, sold entirely online. Customers pick a base plan and work through a three-step purchase flow: Delivery → Account → Payment.
The final step is the Order Summary page, a clean breakdown of what the customer is buying before they pay. At that moment, the customer’s intent is as high as it’ll ever be. They’ve made a decision.
But, until this project, that Order Summary page had nothing on it except the basket. No suggestions. No prompts. No upsell of any kind.
No upsell.
And data that proved it was costing us.
A missed opportunity hiding in plain sight
Sky sells TV streaming packages online. During checkout, customers pick their base plan and land on an Order Summary page before paying.
That page had zero upsell. No suggestions, no add-ons, nothing. Customers who’d skipped premium TV packages like Sky Sports, Sky Cinema, Ultra HD & Dolby Atmos (UHD) earlier in the journey on the TV Packs page hit the Order Summary and were given no second chance.
The data made the cost of that gap hard to ignore.
A fake door test had already been run.
No real product behind it. Just a prompt.
1.6%
CTR on Sky Sports prompt
0.75%
Upsell rate. With no real product yet.
If people click on something that doesn’t exist, what happens when we build the real thing?
Checkout is the most fragile moment in any purchase funnel.
This was a design problem as much as a commercial one
The opportunity was obvious. But I knew that acting on it carelessly could do real damage.
Checkout is where customers are most likely to second-guess themselves. Any extra visual noise, more decisions, more things to read could increase the chances of customers abandoning the purchase entirely. Killing conversion to boost attachment would be a net loss.
Adding upsell cards above or near the CTA was seen internally as a real threat to CVR. Before a single design was made, I had to accept that the wrong call here had consequences.
The Goal
Drive incremental attachment of premium add-ons at checkout: Sky Sports, UHD, Sky Cinema.
The Constraint
Don’t introduce friction that causes customers to abandon. CVR must stay flat.
The design brief was to find where these two things could coexist.
What competitors do and what I chose to do differently?
I looked at how other streaming and e-commerce platforms handle checkout upsell. Two patterns came up constantly:
Pop-up Modals
Interrupt the checkout flow with a full-screen or overlay prompt. High visibility, high friction. CTR goes up; so does abandonment.
→ Too aggressive for a fragile funnel moment.
Generic banners
Promotional strips at the top or bottom of the page. Low friction, but also low relevance. They are not tailored to what the user is actually buying.
→ Ignored because they feel like ads.
The right approach was something in between - relevant, quiet yet impactful and below the line of payment anxiety.
Card Size Test
Iterations Explored
We explored two card sizes on the Order Summary page, a smaller compact card and a bigger, more visual one. The bigger cards lost. They added real visual weight and pushed the CTA further down the page, which is risky on mobile, where every extra scroll is a chance to lose the sale. They also felt closer to a marketing banner than a personalised recommendation, working against the "Recommended for you" framing.
The smaller cards had no real downside, just a one line description and a single "add" CTA, easy to scan and act on. This wasn't only a design call. After the fake door test performed well, we ran a real A/B test between the two card sizes with actual user data to settle it. The smaller cards won, and that's the version that shipped.
Compact card — the version that shipped
Larger, image-led card — tested, didn't ship
Three principles I designed around
01
Curation
The upsell cards were pushed based on two factors - customer intent and the packs the customer already had in the basket. Sky Sports and UHD were prioritised - highest commercial value, highest intent signals. The upsell needed to feel earned, not forced.
02
Personalisation
The copy ‘Recommended for you’ was used to signal the customer that this upsell is based on their selection and basket, not a blanket promotion. That single copy decision changed the perceived relevance of the card.
03
Placement
Iterations were created in collaboration with the Product Manager and development squad. Position of the card (above and below the CTA) was tested in the fake door test, and card size (a compact card versus a bigger, more visual one) was later settled with a real A/B test.
Small, image-less cards above the CTA won. Counterintuitively, subtlety outperformed visibility and didn’t hurt CVR.
Checkout Upsell Cards, from experiments to permanent feature
Experiment
Fake Door Test
Before designing anything real, the team ran a fake door test.
A prompt at checkout with no product behind it. Just a button.
The question was simple: would anyone even click?
1.6% CTR and a 0.75% upsell rate on a ghost prompt was enough signal to move forward. People wanted to see more. The job now was designing something that respected that intent without making them regret staying.
Bake In · Sweeties at the Till
The Feature (Mobile)
I designed small, image-less upsell cards placed above the CTA on the Order Summary page shown only to users who didn’t already own the product.
My call: keep the cards compact. No images. No hero sections. Just a product name, a brief description, a price, and a single CTA. Every element you add at checkout is a new reason to hesitate.
“Recommended for you” framing was a specific copy decision I pushed for based on competitor analysis. It signals relevance rather than promotion, which matters at this stage in the journey.
Bake In · Sweeties at the Till
The Feature (Desktop)
On desktop, the Order Summary spanned wider, so I placed two upsell cards side by side, one premium product, one lower-value option.
I kept the card size and layout consistent with mobile. The same logic applied: if the user already owned the product, they didn’t see it. Priority order was a high-value premium first i.e., Sky Sports or Sky Cinema, then UHD, then Sky Cinema.
I placed the cards above the Continue CTA, visible without scrolling, but below the price summary, so the user processed what they were paying before seeing what else was available.
Checkout upsell cards are now permanent on Sky Stream
Following two experiments, a fake door test and then the fully designed feature, the cards are now baked in as a permanent part of the checkout experience.
CVR remained flat across both experiments. The biggest design risk didn’t materialise.
CVR impact: flat across both experiments. The core design constraint held.
+500%
Attach Rate
Ultra HD & Dolby Atmos (TV Experience) Attachment
+51
Incremental UHD sales / month
Ultra HD & Dolby Atmos (TV Experience) Attachment
+256%
Attach Rate
Sky Cinema (Premium TV Pack)
+40
Incremental Sky Cinema sales / month *premium
Sky Cinema (Premium TV Pack)
Three things I’d carry into the next project.
Subtlety at checkout builds trust
Small, image-less cards consistently outperformed larger image-led ones in CTR. I went into the project expecting bigger to mean better. It didn’t. At a vulnerable moment in the funnel, restraint is a design decision, not a compromise.
Content is as important as placement
‘Recommended for you’ wasn’t just copy. It changed how the card was perceived. Surfacing a card that said ‘you might like this’ felt different from one that said ‘here’s a deal.’ That single framing decision shaped how relevant the upsell felt.
The biggest design risks are often wrong
Cards above the CTA were seen as a conversion threat. They weren’t. CVR stayed flat. This taught me to pressure-test assumptions early. The risk that feels most obvious to a team isn’t always the one that materialises in data.
Role
Senior Visual Designer
Responsibility
Product Planning, Product Design, Documentation
Timeline
Planning & Designing
2 months
Dev Handover & Build
2 months
Experiment & Final Feature Bake - In
3 months