Dubai · Web Design
Conversion rate optimization in Dubai — changes shipped, not just recommended.
We audit where visitors leave, rank the fixes by what they cost against what they are worth, then build and measure them on the site you already run. What you keep is a test log: every change made, and whether the conversion rate moved.
- Audit before anything changes
- Analytics in your accounts
- Consent handled under UAE PDPL
What's included
Six things a CRO engagement does to a live site.
Every one of them lands on the site you are already running, rather than in a document about it.
An audit of where the site loses people
Every step from arrival to enquiry or checkout, walked on the templates that actually carry traffic — desktop and phone, English and Arabic. It ends as a written list of the points where people leave, ordered by how many leave there.
Measurement repaired before anything is judged by it
Plenty of sites count page views and treat the result as conversion data. A conversion gets defined once, the events get wired to match, and the reports agree with the platform. Skip this and every decision afterwards is made against figures that were never right.
Forms and checkout, field by field
The longest drop in most funnels is a form. Fields nobody reads come out, validation stops rejecting phone numbers and addresses that are perfectly valid here, and the payment methods people in this market already carry on their phone sit where the eye is.
A/B testing where the traffic supports it
A split test needs enough conversions per variant to separate a real difference from noise. Where the volume is there, a change runs as an experiment with its sample size and stop date agreed in advance. Where it is not, we say so instead of selling a test that cannot conclude.
The changes built and released
Written, reviewed on staging, then released to the live site — in the theme, the CMS or the codebase, whichever the site runs on. This is the step that separates optimisation from advice about optimisation, and it is the one most CRO engagements hand back to you.
Arabic and English read as two funnels
A bilingual site has two audiences on two journeys, and an average hides both. Traffic, drop-off and form completion are reported per language, because a mirrored layout can be sound in one direction and quietly broken in the other.
Who it's for
Three sites worth optimising now.
And two that are not: a site with almost no traffic has nothing to learn from yet, and a site being replaced within a few months will not keep the findings.
Traffic arrives and nothing happens
Search and paid are both working, sessions are up, and the enquiry count has not moved with them. The media spend is buying attention the site cannot convert, which makes every further dirham of it worth less than the last one.
Sessions have grown year on year and enquiries have not.
A store that loses people at checkout
Products get found, carts get filled, and the payment step is where the number collapses. Checkout is the most measurable part of a website and usually the least examined — it is also where a missing local payment method or a badly built address field costs real orders.
You cannot say which checkout step loses the most people.
A rebuild being argued for on instinct
Someone senior thinks the site looks dated and a budget is being drawn up around that opinion. An audit costs a fraction of a rebuild and answers the question underneath it — whether the design is the problem, or whether one form, one step or one page is.
Nobody in the room can name the page where the money is lost.
How we deliver it
Five stages, and nothing ships on a hunch.
The order matters: measurement is repaired before anything gets judged by it, and every change is logged on the day it goes live.
01
Baseline
Tracking is checked and fixed first, then the current conversion rate is recorded over an agreed window. Everything after this is read against that number, which is why it is worth the week it takes rather than a figure taken on the first day.
02
Diagnose
The funnel walked on real devices, session review, form-field analysis and drop-off broken out by page, device and language. Ends with the losses written down and ordered by size, not by how easy they are to fix.
03
Prioritise
Each fix scored on the effect it is expected to have against what it costs to build, then agreed with you before anything is written. Cheap and obvious goes first. A whole-template rewrite waits until there is evidence for it.
04
Build and release
Changes written, reviewed on staging, then released. Every one is logged with the date it went live, because a change nobody recorded is a change nobody can read a result against three months later.
05
Read and repeat
The result read at the point agreed in advance rather than the moment it looks good — stopping a test early is the most common way a programme produces a win that does not survive. Wins stay, losses are reverted and recorded, and the backlog is re-ordered on what was learned.
Stack
Boring technology, chosen so the next developer recognises it.
Nothing here is exotic. That is the point: an unusual stack is a lock-in mechanism dressed up as innovation.
- Front-end
- Next.jsReactTypeScriptTailwind CSSAstro
- CMS & commerce
- WordPressShopifySanityPayloadWooCommerce
Deliverables
What lands, and what you keep afterwards.
The analytics property, the testing tool and the repository are all in your accounts, so the programme can carry on whether or not we are still running it.
- A written conversion audit of the live site, ordered by where the largest losses are
- A measurement plan, and the analytics events wired to match it
- A prioritised change backlog, each item scored on expected effect against build cost
- The changes themselves, released to the live site through staging
- A test log — what was tried, the dates it ran, and what the number did afterwards
- The baseline reading, and each month's comparison against it
- A tracking and consent note: what is recorded, on what basis, and how long it is kept
The changes that did nothing stay in the log. A programme that reports only its wins is a highlight reel, and it makes the next decision worse — because nobody can see what has already been ruled out.
Pricing
What conversion work costs.
AED 6,500
per month
That is the Growth care tier — retained hours and monthly reporting, which is how a continuing optimisation programme is billed here. A one-off audit with the changes implemented is scoped and quoted on its own, because the size of the site decides it rather than a page count. The smaller Launch plan at AED 1,500 a month covers updates, backups and monitoring rather than optimisation work. Prices exclude 5% VAT, and the retainer cancels on thirty days' notice.
What changes the priceWhat that covers
- A conversion audit of the live site, written down
- Analytics and event tracking repaired first
- A ranked backlog of changes, scored against build cost
- The changes built and released, not just listed
- Experiments run and read where the traffic supports them
- Monthly reporting against an agreed baseline
Retail and e-commerce
Every step from product page to payment is countable, and the volume is usually there to conclude a test rather than argue about one. The local detail then decides more than the layout does: a checkout built around a postcode in a market that does not use one, or one that omits the wallet and instalment options shoppers here already have on their phone, loses orders that no amount of visual polish recovers.
Real estate
The conversion happens on a listing rather than a homepage, and it is an enquiry rather than a purchase — so the form is effectively the product. What it asks for decides how many arrive: a required budget field filters out browsers, which is either the point or the loss depending on how the sales team works. Most of it happens on a phone, in two languages, and the two behave differently.
Optimisation is the only one of our design services in Dubai that starts from a site already carrying traffic. Where the structure underneath is the problem rather than one step in one journey, a redesign is the cheaper answer — and where the page does not exist yet, what you need is a landing page, not a test.
Questions
Asked before a CRO programme starts.
Often not, and that changes the method rather than the work. A split test needs enough conversions per variant to separate a real difference from noise, which most UAE business sites do not produce inside a month. Below that volume the honest approach is diagnosis and sequential change: fix what the audit shows is broken, release it, and read it against a baseline agreed before it went live.
The first changes usually ship within weeks, but a reading you can trust takes longer than the change does. A test has to run for whole weeks, because weekday and weekend behaviour differ, and it has to be called at the point agreed up front rather than the moment it looks good — stopping early is the most common way a programme produces a win that later evaporates.
A continuing programme runs as a monthly retainer from AED 6,500 — the Growth care tier, which is retained hours plus reporting against an agreed baseline. A one-off audit with the changes implemented is quoted on its own, because the size and platform of the site decide it rather than a page count. Prices exclude 5% VAT, and the retainer cancels on thirty days' notice.
Optimise if the structure is sound and one or two steps leak; redesign if nobody can name the page where the money is lost. Testing on a site you intend to replace within a few months buys findings that will not survive the rebuild. The cheap way to settle it is the audit — it costs a fraction of a rebuild and tells you which of the two you are actually buying.
Yes — what we need is access rather than authorship: the analytics property, the CMS or the repository, and whoever currently deploys. The audit reports early whether the platform can carry the changes — some inherited themes and page builders make a form or a checkout step expensive to alter, and that reorders the work.
Yes, with consent and with the recording configured not to capture what it should not. Under the UAE Personal Data Protection Law that means asking before behavioural tracking starts, masking form fields and anything personal inside a replay, and keeping recordings only as long as the analysis needs them. It is a setup decision, and far easier to get right before the tooling goes on than after.
We do. The change is written, reviewed on staging and released to the live site, which is the part that separates this from a recommendations document. Where your own developers own the deployment, the change is handed over as a pull request with a description of what it does and what it is expected to move. Either way the site carries the change, not a report about it.
It gets recorded and the original stays live. A change that moves nothing is still a result: it closes off a line of thinking cheaply and stops the same idea being proposed again next quarter. What does not happen is the losing tests quietly dropping out of the report — a log of nothing but wins tells you nothing about what has already been ruled out.
Ready to build something that lasts?
Tell us what you need. You will get a scoped response with an indicative price — from an engineer, not a salesperson.