AHZ Associates

A student in Dhaka deciding whether to apply to a UK university has one question ahead of all the others, and it is not about rankings. It is what this is going to cost me, in money and in mistakes. AHZ Associates places international students at UK universities and charges those students nothing, so its website exists mostly to make an unfamiliar process legible: here is every step, here is what we do at each one, here is where our office is in your country. Getting that right across sixteen countries is a content problem wearing a software costume.

Why it exists

A H & Z Associates Limited was incorporated in June 2012 and is registered at 16 St Clare Street in London EC3N, under the SIC code for educational support services. Its own account of itself is that it was founded by two brothers who came from Bangladesh to the UK to study, that it now works with UK institutions from offices around the world, and that it sits inside a larger group alongside a charitable foundation.

Two audiences read the site and they want opposite things from it. A prospective student wants to know what happens next, in what order, and what it will cost them. A university partner wants to see an organisation that looks capable of handling applications properly. The site has to serve both without either feeling like it walked into the wrong building.

The cost answer is the one that does the most work. Students pay nothing, and the site says so plainly rather than filing it in a fee FAQ, alongside a free eligibility test and a free consultation booking. Worth being straight about what is not on the page: the site does not explain how the business is funded, so anything you might assume about who pays instead is not something we can verify from the outside.

How it works

The service is published as eight steps rather than summarised: free consultation, shortlisting universities, application and document review, review of the statement of purpose and letters of recommendation, scholarship advice, choosing between offers, student visa support, and accommodation support. Further pages cover counselling, pre departure preparation and post arrival help with settling in, banking and health services. Setting the whole path out in public is the point. A reader who can see all eight steps is far likelier to start than one invited to submit an enquiry and wait to find out.

On top of that spine sit country sections with their own landing pages and application forms for Bangladesh, India, Pakistan, Nepal, Sri Lanka, Nigeria and Egypt, a global offices directory with filtering across sixteen countries, subject browsing across areas including business and management, engineering and technology, law, and medicine and health, and the usual blog, news and events sections with a downloadable guide and a newsletter.

Three separate systems carry all of it, which is easier to follow as a table than as prose.

System	What it is for	How it runs	What can be seen from outside
Main marketing site	The public journey, country sections, offices directory, editorial	One published site behind a CDN, with imagery served from object storage	Everything
Application portal, on its own subdomain	Managing a live application	A separate deployment on its own web server	A logo and a loading state; the rest is behind sign-in
Global site, on another subdomain	Study destinations beyond the UK, including Australia, Canada, Ireland, Germany and the USA	A different host again	Everything, and its figures do not match the main site's

The older .co.uk domain redirects permanently to the .com, which is one of the few pieces of this estate where the correct behaviour is simple and easy to confirm.

The decision that shaped it

The choice worth defending is that the marketing site and the application portal are separate systems, on separate hosting, with separate release cycles.

Marketing pages change constantly, in a hurry, at the hands of people who are not engineers. An application in progress is somebody's visa timeline. Couple the two and a rushed content deploy can interrupt a student halfway through a form, while a portal release has to be scheduled around a campaign. Split apart, the blast radius of each change stops at its own boundary. Country variance sharpens the argument further: varying marketing copy by country risks an embarrassing page, whereas varying an application flow by country means different validation rules, document lists and states that real people are already sitting in.

The cost of that split is not zero. Two deployments, two sets of dependencies to keep current, a shared design language that has to be maintained by hand in two places, and a seam a student crosses at exactly the moment they are most anxious. Sign-in state does not travel across it for free. Anyone choosing this shape should price the seam honestly rather than assuming separation is free because it sounds tidy.

The second decision is where country differences live. The rule we work to is that anything true everywhere is stored once and rendered many times, and only genuine local fact gets a variant: office addresses, intake dates, document lists, the local proof of a photograph and a phone code. Seven country sections with four page types each is twenty-eight pages before you add a second language, and the multiplier that catches people out is not authoring, it is review. Somebody has to read all twenty-eight after any change and confirm they still agree. That job has no owner by default, which is exactly why it never happens. We set out the full method, including holding the variance as data with a review date attached, in what to vary by country, and what to keep identical.

Where it stands today

All three sites are live. The main site and the portal sit behind a CDN, the global site runs somewhere else, and the content management system behind the main site is not something we can name with confidence: the image filenames and the shared bucket suggest a group platform, but that is inference from the outside rather than a fact from the inside.

The portal is the honest gap. It renders behind authentication, so an unauthenticated visitor gets a logo and a loading state, and the routes you would guess at return nothing. We have not exercised what sits behind that screen, and describing screens we have not seen would be inventing a case study.

Then there are the numbers, which are the most visible symptom of a large estate drifting. Counts of partner institutions, counsellors, offices, staff and enrolled students appear across the homepage, the about page and the sister site, and they do not agree with one another. None of them was wrong on the day it was written; each was accurate once, then one got updated and the others did not. Claims about satisfaction rates and award counts are marketing statements with no source attached. We have not repeated any of those figures here. Sixteen is what the offices directory actually lists, so sixteen is the only count on this page, and everything else stays with the company publishing it.

We have been on the other side of this. Our own civic platform, Jonosakti, still shows headline totals that came out of a seed migration written before launch, and the only reason the real activity can be told apart from the seeded rows is that a separate live count exists for figures produced by actual accounts. Nobody sets out to publish a fixture file as a statistic. It happens because the number was true enough on the day and nothing was attached to it saying where it came from.

What this says about how we work

Large content sites rarely fail with an outage. They fail slowly, through drift, until the Nepal page contradicts the India page and a student turns up at an appointment with the wrong documents. Nothing about a stale page announces itself, which is why the engineering that matters on a project this shape is structural: one source for anything shared, variance held as data rather than as duplicated prose, a named owner per variant, and a review date rendered where a reader can see it.

The same discipline shows up in smaller work. Keeping content in a database with its history intact is what let us move Taxcare Accountancy off a legacy system without dropping the addresses people had already linked to, and the inventory and redirect method behind that is exactly what a country section restructure needs too.

The other half of the job is knowing what to leave alone. A portal holding live applications is not somewhere to try out an opinion about interface design, and a student mid-application is not an audience for our preferences. That restraint is part of what we sell: the willingness to say that a piece of a system is working, is nobody's priority, and should be left where it is.