With about twenty minutes left in a hotel-booking web app's frontend design round, which one or two deep dives would you open, and why?
answer
- what this product lives or dies by
- rank by stated requirement, not comfort
- landing pages: search traffic on cheap phones
- the booking flow on a flaky network
- skip styling, tooling, logging, framework
basics
~20 sOpen the areas the stated requirements make hardest: how the results and hotel pages reach crawlers and low-end phones quickly, then the booking form on flaky networks. Skip styling, tooling and framework debates, and say why you chose.
solid answer
~40 sI'd rank candidates by what this product lives or dies by and which stated requirement each carries. First, **how the results and hotel pages are delivered**: most traffic arrives from search onto those pages, on low-end phones, so indexing and LCP meet there; that is a rendering-per-page-type deep dive. Second, **the three-step booking flow on a flaky network**: it is where revenue converts, and a dropped request must neither lose the guest's input nor book twice. If time remains, keeping filters, list and map consistent on the results page. I would not spend it on styling, build tooling, logging or which framework. I'd announce the choice with its reason and let the interviewer redirect. For an internal staff dashboard the picks change completely: large tables and charts, not indexing.
go deeper
Know that the deep dives are a choice: one or two areas tied to stated requirements, not a tour of every topic you know.
Explain your ranking: which requirement each candidate carries, whether it is unique to this product, and whether it changes the architecture.
Pick under time pressure, say why aloud, name what you are deliberately leaving out, and know which pick comes next if time allows.
Recognise deep-dive choice as engineering prioritisation: a team's effort should follow the risks that its traffic, users and revenue actually create.
## Why the choice itself is graded The last stage of the RADIO frame, **optimizations and deep dives**, usually gets the largest share of a frontend design round; the frame's authors suggest about 40 % of the session. There is rarely time to cover every area, so **choosing** what to open is part of what is assessed. A good choice shows you understood the product and its users; a poor one (a tour of everything, or the topic you happen to know) shows you did not connect the requirements to the design. What an interviewer reads from the choice: - whether you remember the non-functional requirements agreed twenty-five minutes earlier; - whether you can tell what is **unique** to this product from what every website needs; - whether you can say *why* out loud, briefly, before spending the time; - whether you can leave something out on purpose and say so. ## A way to rank candidates Before opening anything, rank the candidates quickly against four tests: 1. **Does a stated requirement depend on it?** A deep dive that carries an agreed non-functional requirement (search traffic, low-end phones, a right-to-left language) is worth more than one nobody asked for. 2. **Is it unique to this product?** What defines a booking site is landing pages that must be found and fast, and a booking flow that must not fail. Generic concerns apply to every site. 3. **Does it change the architecture?** Topics that move boxes or arrows are worth more than implementation details. 4. **Can you go deep?** Your strengths count, provided they serve the product. ## The hotel-booking picks | Candidate | Requirement it carries | Verdict | |---|---|---| | How the results and hotel pages are delivered | Organic search, LCP on low-end phones | **Open first** | | The three-step booking flow on a flaky network | Flaky network, revenue at the last step | **Open second** | | Filters, list and map staying consistent | Low-end phones (INP), shareable result views | Third, if time remains | | Photo carousel performance | LCP on the hotel page | Fold into the first pick | | Live price updates | No stated freshness requirement | Leave out | | Styling approach, build tool, logging | None | Leave out | The **first pick** is where two stated requirements meet: most visitors arrive from search onto results and hotel pages, on cheap phones over unreliable networks, so both indexing and load speed are decided by how those pages are delivered. The **second pick** is where the business makes its money and where the flaky network does the most damage: a guest who loses typed details, or is charged twice after a retry, does not come back. Notice what the table does *not* contain: the substance of either deep dive. How each page should be rendered, or how the booking submit stays safe to retry, are designs of their own. This stage is about **choosing and justifying**. ## What not to open Topics that rarely change the architecture belong at the bottom unless the prompt makes them central: - which **component framework** to use, or a comparison of frameworks; - the **styling approach** or a component kit, unless theming or scale is the question; - **build tools**, linting and deployment pipelines; - **logging and analytics** integrations, unless they shape the user experience; - generic advice such as "minify and compress", unless performance at scale is the core problem. ## Same frame, different product The value of ranking by requirement shows when the product changes. For an **internal analytics dashboard** used by a few hundred signed-in staff on desktops, indexing drops out entirely and first-load weight matters less. The hard parts become **screens holding thousands of table rows and several charts**, and how fresh the numbers must be. Same frame, entirely different deep dives, because the requirements differ. ## Announcing the choice - Say the one or two picks and the requirement each serves, in one or two sentences. - Offer the interviewer the chance to redirect; some have a topic in mind. - Keep a rough time box per pick, and say what you are deliberately leaving out so it reads as a choice rather than a gap. - If the interviewer redirects, follow, and connect the new topic back to a stated requirement where you can.
- Why rank the booking flow above keeping the map and list in step?Because the stated requirements bite harder there: the flaky network threatens a guest's typed details and could cause a double booking, which costs money and trust, while the map enhances a list that already delivers every result. If the interviewer's product were mostly browsed by map, that ranking would flip.
- How would the picks change for an internal analytics dashboard?Indexing drops out because staff sign in, and desktops on office networks make first-load weight less pressing. I'd open how screens with thousands of table rows and several charts stay responsive, and how fresh the numbers must be, rather than how pages reach crawlers.
- Is it wrong to pick a deep dive because it is your strength?Not when it serves the product: showing depth where you have it is encouraged. An accessibility specialist opening focus and error handling in the booking steps is relevant here, because it carries a stated requirement. Opening the internals of a JavaScript engine is not, however well you know them.
- Opens a deep dive on the styling approach for a site fed by search.
- Tries to touch every area for a minute instead of one or two in depth.
- Picks deep dives without tying them to any stated requirement.
- Spends the deep-dive time comparing component frameworks.
- Chooses the same deep dives for a public booking site and an internal dashboard.