In a frontend design round for a hotel-booking web app, which clarifying questions change the design, and how does each answer change it?
answer
- ask only what forks the design
- device mix and network quality
- who arrives from search engines
- languages, right-to-left, accessibility
- items per screen, offline or not
basics
~10 sAsk what forks the design: device mix, network quality, search-engine traffic, languages and right-to-left, accessibility, data volume per screen, offline needs. Each answer moves a decision: first-response content, script weight, layout mirroring, paging, scope.
solid answer
~50 sI'd ask only questions whose answer changes something I draw. **Devices and network**: half the traffic on low-end phones over flaky mobile data means a tight script budget, the map loaded after the list, and a booking form that survives a dropped request. **Search traffic**: if most visitors arrive from search engines, the results and hotel pages must return meaningful HTML in the first response; how to deliver that is a deep dive. **Languages**: eight, one of them right-to-left, means no hard-coded strings, locale-aware prices and dates, and a layout that mirrors. **Accessibility**: the map cannot be the only way to choose a hotel. **Volume**: results and map pins per screen decide paging and pin clustering. **Offline**: booking needs the network, so offline is out of scope, and saying so is a decision. An internal staff dashboard would answer most of these the other way.
go deeper
Remember the list: devices, network, search traffic, languages, accessibility, volume per screen and offline, and that each answer should change something you draw.
For each question, say the consequence: a script budget, meaningful first HTML on landing pages, a mirrored layout, a non-map path to every hotel, paging and clustering.
Prioritise the two or three questions that fork this product, state assumptions where the interviewer defers, and revisit them when a later decision rests on them.
Notice which answers are business bets, such as a ninth language or a market on slower networks, and keep the design cheap to change along those axes.
## Ask questions whose answers fork the design A design round opens with a deliberately vague prompt: "design a hotel-booking website". **Clarifying questions** turn it into something you can design. In a **frontend** round the questions that matter are the ones whose answer changes a decision in the browser: what the first response contains, how much script ships, how the layout behaves, how much data one screen holds. A question that changes nothing you will draw is time spent without score. The preferred move is to propose the likely answers yourself and let the interviewer correct them, then write the agreed list where both of you can see it. ## The questions and what each answer changes | Question | Hotel-booking answer | What it changes | Internal analytics dashboard | |---|---|---|---| | Which devices? | Half on low-end phones | Script budget, interaction cost, image sizes | Desktop browsers | | How good is the network? | Flaky mobile data | Small payloads, retry, form data that survives failures | Office network | | Do pages need to be found by search engines? | Most traffic is organic search | Meaningful HTML in the first response for landing pages | No: everything is behind sign-in | | Which languages? | Eight, one right-to-left | Externalised strings, locale formats, mirrored layout | One or two | | What accessibility do we commit to? | Keyboard and screen-reader users can book | A non-map path to every hotel, operable carousel, announced errors | Same commitment, different screens | | How much data on one screen? | Dozens of results, hundreds of pins | Paging, pin clustering | Thousands of table rows, several charts | | Offline? | No, booking needs the network | Out of scope | No | The right-hand column is the point: the **same questions** give a different product, so the design is driven by the answers, not by a template. ## Devices and network "Half our users are on low-end phones over flaky mobile networks" changes several things at once: - **Script weight** matters: every kilobyte must be downloaded, parsed and run on a slow processor before the page responds well, so heavy pieces such as the map load after the results list that carries the actual content. - **Interaction cost** matters: a filter tap must not redo more work than the change needs. - **Failure is normal**: the three-step booking form must keep what the guest typed when a request fails and retry the submit without booking twice. That last property is part of the API contract. ## Search traffic When most visitors land from a search engine, the results pages for a destination and the hotel pages are **landing pages**. They need their main content in the HTML the server sends, because a page whose content appears only after script runs waits in the documented crawler's rendering queue, for a few seconds or longer, and depends on that render succeeding. It is commonly reported that other crawlers may not run script at all. The booking steps, by contrast, never need indexing. Which rendering approach each page gets is a deep dive of its own; this stage only records that the requirement exists. ## Languages and right-to-left Eight languages mean: - no hard-coded strings in components; - prices, dates and currencies formatted per locale; - room in the layout for text that runs longer in translation; - a URL per language version, so each can be found and shared. **One right-to-left language** changes layout, not just text: - spacing and alignment written in terms of start and end rather than left and right; - mirrored directional icons, such as back and next arrows; - step indicators in the booking form that read right to left; - mixed-direction text, such as a Latin-script hotel name inside a right-to-left sentence. ## Accessibility, volume and offline 1. **Accessibility** is stated as a requirement and turned into consequences: the results list must offer every hotel the map shows, the carousel must work without a pointer, and booking errors must be announced. What full conformance demands is separate accessibility work. 2. **Data volume** decides whether the results page pages or scrolls and whether the map clusters pins. 3. **Offline** is a scope decision. Saying "not offline: booking needs live prices and availability" is a valid answer that removes a whole layer of design. When the interviewer says "you decide", state the assumption that makes the design harder rather than easier, write it down, and say what would change if it turned out wrong.
- The interviewer answers your device question with 'you decide'. What do you do?State the assumption that makes the design harder rather than easier (for a public booking site, mid- to low-end phones on mobile networks), write it beside the requirements, and say what would change if it were wrong. That keeps the round moving and shows you know which answer forks the design.
- Which of these questions matter less for an internal analytics dashboard, and which matter more?Search traffic drops out because staff sign in, and devices are desktops on office networks, so first-load weight matters less. Data volume grows: thousands of rows per table and several charts on one screen. How fresh the numbers must be and whether people export them become the questions that drive the deep dives.
- Why ask about right-to-left specifically rather than just 'internationalisation'?Because it changes layout, not only strings: spacing and alignment by start and end instead of left and right, mirrored icons such as back arrows, step indicators that read the other way, and mixed-direction text like a Latin hotel name in a right-to-left sentence. Translation alone changes only the words and their length.
- Opens the clarifying questions by asking which framework the team uses.
- Treats indexing as later polish for a site fed by organic search.
- Assumes desktop browsers on fast Wi-Fi without asking about the device mix.
- Equates internationalisation with translating strings, ignoring right-to-left layout and formats.
- Asks many questions without saying which design decision each answer changes.