The company was betting on the next product. I made the current one earn more.
Jimdo was busy building what came next. I was the designer on checkout, which meant I was sitting on the part that was already paying for all of it, and that almost nobody was looking at. So while the company built the new thing, I spent eight months making the current one earn more.
Improving something that already works is harder than fixing something broken. There is no villain and no broken step to point at, so every change has to be argued for. I found the changes with the best ratio of effort to return, and then argued for them.
I was the designer on checkout, and because checkout is where the money changes hands, that stretched into growth. We also spent long stretches without a PM, so I picked that up too: I pulled and analysed the funnel data myself, wrote the hypotheses, prioritised them, and stayed through dev handoff and A/B results. I saw opportunities everywhere in that funnel and did not wait for them to reach a backlog. I put them there.
Everyone agreed checkout needed to change.
Nobody could say where to start.
Most of the company's attention, correctly, was on the new product. Which left the funnel that was taking money that week with nobody watching it closely.
The catch is that it was converting above industry average. When a flow is beating benchmark, "it feels dated" is not an argument, it is a risk, and rightly nobody was going to spend engineering time on it. I ran a design critique with the design team over the whole flow, to turn a shared hunch into something written down by more than one person. Every note on that screen has an author who is not me, which is what made it usable in the next conversation.
It confirmed what I suspected about the flow and about pricing, but a wall of notes is not an order to work in. So I took it to the data: Mixpanel and Tableau for where people were leaving, usability testing in Maze for what they hit on the way.
There was a lot to do.
I did not know Mixpanel when I started. I asked the data team to teach me, and then started building my own boards, simple ones first and harder questions as I went. A question you can answer yourself is a question you ask more often, and a good half of what ended up on this page came from questions that were only worth asking once they were cheap.
Too many steps
A long, multi-step checkout, adding screens to a flow whose job is to collect a payment.
Nothing being offered
No upsell strategy at all, and not one add-on prompt anywhere in the flow.
Pages with no owner
In-product landing pages, each with its own UI and a weak hierarchy.
Missing payment methods
No Apple Pay, and none of the options people now expect to find.
Weakest on mobile
Higher drop-off on phones, and few trust signals to hold anyone there.
Five problems, and nowhere near the engineering time to take on all five. Rebuilding a checkout that was outperforming benchmark was never on the table, and it should not have been. So the question was never what would the ideal flow look like. It was which of these moves the number for the least effort, and how do I prove it before anyone spends a sprint on it. Everything below comes out of that question.
Four fronts, and one thing every one of them had in common
Every problem on that list had been sitting there for a long time, and the reason was always the same: somebody had decided it once, and nobody had checked since. So the job was less about having better taste than the people before me, and more about being the one willing to go and look.
Three of the four came with something we could not agree on. Each one was settled by evidence rather than by where people sat on the org chart.
A marketing restyle was already scheduled. I used it.
A visual refresh was already planned for other reasons, so engineering time was allocated. That is the cheapest moment in the year to change anything, and I took the brief further than a restyle: clearer prices, plan highlights rewritten around value instead of feature lists, and consistency with the rest of the flow. Plan cards were reordered highest to lowest, left to right, from an A/B test rather than a preference.
Red is a very good marketing colour, and we had already given it a job
Prices, discount badges and the countdown bar were all red. Everybody knew exactly why: red creates urgency, and marketing had an A/B test from three years earlier where it had won. That is a real argument, not a habit. It was also colliding with the design system, where red is what an error looks like, on a subscription people are meant to keep for years.
Red sells urgency, and we have the test that says so. Nothing since has suggested otherwise.
That test is three years old and red already means something else here. One week to re-run it, and I take the hit if I am wrong.
A full week, no measurable difference in either direction. Red had not been earning conversion, it had just been sitting there, and there was no way to know that without looking. Which gave us two things worth more than a point of conversion. Red went back to meaning one thing in the product instead of two, so an error looks like an error again and a discount does not. And the pricing page stopped being governed by a result nobody could re-examine, so from then on we could keep testing it without carrying somebody's three-year-old conclusion into every decision.
Faster, and worth more per customer
Add-ons moved inside the flow instead of living somewhere you had to go and find.
We removed redundant steps, added Apple Pay, and connected Stripe. That last one is less boring than it sounds: Stripe has its own opinions about how payment inputs look and behave, and the work was balancing what it insists on against a design system with its own rules. Neither side gets everything.
Then we embedded add-ons into the flow and tested where they landed best. Adoption went up, which turned checkout from a step people get through into a place the business earns.
The more valuable thing it bought us was a place to ask questions. Once add-ons live at the moment of payment, you can put a feature that does not exist yet in front of people who are actively spending money, and find out whether anyone wants it before a single sprint goes into building it. That is a far better signal than a survey, because the person answering has their card out.
We ran those honestly. Clicking told you "thanks for your interest, we'll let you know when it's ready" and put you on a list, so the click bought you something. A door that opens onto nothing returns the same data and spends a little of your customers' trust to get it.
Which number a person should see first, and where the detail belongs
This one starts on the pricing page and ends here, which is exactly why it mattered. The business wanted the price shown before tax, with an asterisk, because it is the smaller number and our customers are businesses who can reclaim VAT. The problem is what that does to the order summary: you advertise one figure and then charge another, and the difference turns up at the last step of a flow you have spent months smoothing out.
The lower number wins more clicks, and the asterisk covers us.
Show the total, say it includes VAT, and break it down in the order summary where someone is actually checking the maths.
It kept turning up in research we were running for other reasons: shown one price and charged another, people said a version of the same thing, that it felt like a trick. A quick qualitative round to check it and every single person said it. Nobody needed a chart after that. It shipped my way, and the useful part is why: not because I argued better, but because a price is one promise made twice, and the two have to match.
Pages nobody had claimed
The feature landing pages inside the product had been waiting a long time for a team to pick them up. There were even designs already drawn. They belonged to nobody, so nothing happened. I pushed to take Business Listings, and my team ran it even though it was not our area.
One interface
The same UI across every feature page, instead of one per team that touched it.
Value, not features
Content rewritten around what the feature does for you.
One call to action
Standardised CTAs, so every page asks for the same thing the same way.
Upsells in pattern
Upsell flows streamlined into the shape used everywhere else.
Pages made dynamic
A starting point if you already have the feature, a selling point if you do not.
It moved conversion on that feature by 2%. Not spectacular on its own, but it came from about a week of work on a page that had been ignored for a long time, and it was the proof that the rest of the pages were worth claiming too. A whole surface went from dead weight to a growth touchpoint because somebody finally put their name on it.
Name the amount, do not promise the value
Our engineering lead had seen how WordPress handled long-term plans and brought it to the team, which is worth saying because good ideas do not only arrive from the people whose job it is to have them. We took it, researched it, and landed on a banner at the final step: the exact amount you save by switching to yearly or two years, not a vague promise of value. Copy across the flow carried the same idea, so the saving came up more than once.
It worked. More people moved to yearly and multi-year plans, which lifts lifetime value and retention together. And then we spent a surprising amount of time on what colour it should be.
The banner as it shipped. Not the colour I wanted.
Green already meant something, and we were about to make it mean two things
Green was our reserved colour for positive system feedback: it confirmed, it never sold. Purple was the upsell colour. Using green to sell something bends a rule the whole product relies on, and once it bends in one place it bends everywhere. The VP of Product wanted green.
Green. It reads as a good thing, because it is a good thing.
Purple, or green stops meaning anything anywhere else.
I was never arguing that purple would convert better, and green performed, so green shipped and it is what you are looking at. My case was about what the colour would cost us everywhere else in the product, and that is not something a checkout metric can see. So I disagreed and committed, and built it in green.
Everything above was cheap to ship because of something I built in the first weeks
Order summaries, billing cycle selectors, payment method logos, add-on cards, legal footers, success and error screens. None of it existed. Checkout was the only place in the product that used any of it, and the design system was not structured to hold it, so none of it lived anywhere reusable.
So the first thing I did at Jimdo was build the library. One master file where the specifications live, changed there and flowing down: variants covering every breakpoint and state, naming agreed with engineering so a component means the same thing in Figma and in the code, each set documented with where it is used and when not to use it.
That is the whole reason the eight months above look the way they do. A test is only worth running if the variant is cheap to build, and after the library existed most of them were an afternoon. The investment is what made the experiments affordable, and it is the practice I had learned to build at BlackRock, brought into a team that did not have it.
The master file. Every set named, described, and published for the team.
Measuring is what made disagree and commit actually work
Everyone says they disagree and commit. It only holds if the disagreement has somewhere to go. Without that, committing means the quieter person stopped talking, and the same argument comes back in three weeks with the same people and the same opinions, because nothing ever answered it.
Agreeing in advance how each one would be settled is what made it work. A week of traffic for the red. A round of research for the price. Committing costs you nothing when you know the question is going to get answered, and you can back something you argued against without it being a defeat, because the decision belongs to the evidence rather than to a person.
The saving is not the conversion points. It is the months nobody spent arguing about things nobody could settle.
- Rewrote and restructured 4+ in-product landing pages
- Redesigned the full checkout experience
- Restyled the pricing page
- Added Stripe and Apple Pay
- Introduced upsell banners, modals and pricing highlights
- Created prompts for yearly and two-year plans
- A/B tested add-on flows and placement
- Maintained the team library and design documentation in Figma
On a funnel that already worked
conversion on Business Listings, from a facelift on a page nobody owned
faster checkout, from removed redundancy and better payment integrations
drop-off through the checkout flow
Higher long-term plan adoption
Savings messaging that names the exact amount pushed people toward yearly and two-year plans, which lifts lifetime value and retention together.
More add-on revenue
Upsell points built into the checkout itself rather than bolted on afterwards, and tested for where they landed best.
Better design ops
A checkout and upsell library in Figma, standardised across breakpoints and states. This team did not have that practice before.
Where the value was hiding
The whole company was looking forward, which was the right call and left the present unattended. Nothing in this case was a discovery. It was all sitting in plain sight, in the flow customers were paying through that week, waiting for somebody to spend a quarter on it instead of on the next thing.
Looking after what already exists is not a lesser job. It is frequently the cheapest quarter a company has, and it is the one nobody volunteers for, because there is no launch at the end of it and no story to tell unless you go and write one.
Hola, I'm Lu. Senior Product Designer, 8+ years across different industries. I start with the real problem, make complexity understandable, and stay close to the outcome.
More about me →