Spec audit of a real Dublin bakery's website, done to show how I work. Mannings didn't ask for it, and nothing was changed on their site.
People already wanted the cake. Ordering it was the hard part.

Context
Mannings Bakery has three shops in Dublin, on Thomas Street and in Tallaght and Blanchardstown, and sells cakes online for next-day and later delivery. Its cake photos carry a Blas na hÉireann 2023 finalist badge. Mannings didn't ask for this audit.
The real question
When someone has already decided to buy a birthday cake, what gets in their way when they try to order it on their phone?
My judgement: with an occasion cake, most of the persuading happens before anyone reaches the website. People know the bakery and they know the date. What decides the sale is whether ordering feels clear, quick and certain. So this is a conversion audit: it looks only at the path from wanting the cake to putting it in the basket.
How it was checked
- A phone-sized screen and the public website only. No login, no order placed.
- Three visits over ten days: 6, 11 and 15 November 2026.
- One cake, the Oreo Chocolate Cake, followed from its page to the basket.
- 32 public reviews, read for what customers say goes wrong.
Every finding is labelled: Seen on the live site, Reported by customers, or my Judgement.
What I found: three things standing between buyers and the basket
Ranked by how surely each one stops a sale and how cheap it is to fix.
1. A card marked "optional" blocks the order. (Seen and tested)
The greeting card is labelled optional. Skip it, press Add to basket, and an error appears at the top of the page while nothing goes in the basket. Candles and the cake topper are set up the same way. My judgement: some people will scroll back and fix it. Others will ring the shop or give up.
The fix: make "No card, thanks" the default, and the same for candles and toppers. Anyone who wants one still picks it, and skipping an extra can never block an order.

2. Four different answers to "When do I need to order?" (Seen on three pages)
The homepage banner says order before 5pm for next-day delivery. Further down the same page it says before 3pm, on selected cakes. The cake page says 48 hours' notice, which may be right for that cake. The FAQ says usually 48 hours, often 24. The 5pm and 3pm clash was there on all three visits. My judgement: someone ordering tonight for tomorrow can't tell which deadline applies to their cake.
The fix: one deadline, worded the same way everywhere, with each cake's notice time shown right beside its Add to basket button. Mannings confirms the real cut-off.

3. Nothing on what happens if nobody's home. (Seen, and reported by customers)
The delivery page says drivers come between 8.30am and 5pm and text an estimated time. None of the Delivery, FAQ or Gift a Cake pages says what happens if the person receiving the cake is out. Two of the testimonials on Mannings' own homepage are from people sending cakes from abroad, and one public review describes a cake sent from the US going back to the shop because nobody was home, with the delivery fee lost.
The fix: one short paragraph setting out the policy on the delivery and gift pages, plus boxes for the recipient's phone number and delivery notes at checkout, if they aren't there already. The policy itself is Mannings' call.
Read the full audit (PDF, 7 pages).How we'd know it worked
No business should take a fix on faith, so each one comes with its own check:
- Measure first. Take the last four weeks of numbers from the site's analytics. For the first fix, that's how many cake-page views end in Add to basket.
- Change one thing at a time, a week apart. Each fix has its own number, so each can be judged on its own.
- Compare four weeks later. Keep a change only if it beats the "before" number without making something else worse. For example, paid card and topper sales shouldn't drop sharply.
The cake page checked already runs Google Analytics and Meta's (Facebook's) tracking code, the tools that count visits and clicks. My judgement: the "before" numbers for the first two fixes most likely exist already. The third would come from Mannings' own delivery records. And because December changes cake orders, the comparison should use the same weeks last year where possible.
The plan: three fixes in three weeks, no new website
- Week 1: unblock the basket.
- Week 2: one deadline, everywhere.
- Week 3: peace of mind for gift senders.
- Weeks 4 to 7: read the numbers, keep what works, undo what doesn't, and report back in plain English.
Results
None. This is an audit, not a project Mannings ran. I had no access to their sales or visitor numbers, so none are claimed. I didn't see the checkout, and I followed only one cake, from its page to the basket.
How I worked
No team. I directed the audit and chose what mattered and in what order. AI tools did the legwork: running the site checks and screenshots, organising the evidence, drafting, and cross-checking each claim against a dated screenshot or public source. The final calls are mine.
Why it's here
This is how every project with me starts: an audit of what you already run, before anything new gets built. The full audit is seven pages, with the screenshots, each fix drawn out and the test plan.
Read the full audit (PDF, 7 pages).Want the same check on your order or enquiry page? Let's talk.