Frontend work at a company that makes business applications for finance, human resources and supply chains, and a console for its cloud services, means dense screens used by accountants, buyers and administrators for hours. Prepare to discuss large grids of invoices, long forms with drafts, many languages and number formats, showing the progress of a long-running cloud operation, and totals that must match the exported spreadsheet. What to show here: dense, exact screens that professionals can trust all day. Each question below has what the interviewer is checking, an answer frame and a sample spoken answer to adapt.
How Oracle hires, in its own words, and its questions for every job →
“I built a purchasing screen for buyers who compared supplier quotes all day. They asked for every column on one screen, and my first clean design with fewer columns was rejected. I watched them work and saw that they scanned columns side by side, so scrolling or tabs broke their flow. I kept all columns, but aligned numbers to the right, grouped related columns under headings, let users hide or pin columns, and put rarely used details in a panel that opened without leaving the row. The result was your result.”
Weak answers hide information to make the screen look clean and never test with expert users.
“We replaced an old data entry screen in a finance application that some users had used for years. I first recorded how they used it: they never touched the mouse, and they knew the field order by heart. In the new screen I kept the same field order and every keyboard shortcut, even where the new design could have been different. We let users switch between old and new for a period, collected their complaints, and fixed the top ones before removing the old screen. Some users were still unhappy at first, and I answered each one. The result was your result.”
Weak answers switch everyone at once and dismiss complaints as resistance to change.
“The browser must never hold millions of rows. The server does sorting and filtering and returns one page at a time, with a total count. In the browser I render only the rows visible on screen, reusing elements as the user scrolls. Column widths, order, hidden columns and the frozen first column are client state that I save per user. A frozen column is two scroll areas kept in step. Totals at the bottom come from the server for the whole filtered set, not the loaded page, or users will think the numbers are wrong. Keyboard navigation between cells matters for these users.”
Weak answers load all rows and sort in the browser, or compute totals from the visible page.
“I would save a draft to the server every time the user pauses or leaves a step, so closing the laptop loses nothing. Each step validates its own fields when the user tries to move on, with messages next to the field, but a draft can be saved even when it is incomplete. On submit, the server validates the whole claim again, since the rules about limits and policy live there, and the client maps the server's errors back to the fields and steps. Attachments upload as soon as they are chosen, and the form keeps their ids. A summary step shows everything before submission.”
Weak answers keep the draft only in the browser, or block saving a draft until every field is valid.
“First, no text in the code. Every string comes from a resource file with a key, and the layout must accept text that is much longer than the English. Numbers, dates and currency are formatted by the built-in locale functions, because a decimal comma or a different digit grouping typed by hand will be wrong somewhere. I would write layouts with start and end instead of left and right, so the page mirrors for right to left languages, and check icons that imply direction. I would test with a fake long language and a right to left locale from the first sprint, not at the end.”
Weak answers think localisation means translating strings, and format money with string functions.
“The request should return at once with a job id, and the server owns the state. The page shows the resource in a creating state and asks the server for updates, either by polling at a growing interval or through a push channel. I would show the steps that are known, and avoid a fake progress bar that guesses. If the user leaves and comes back, the page must not rely on anything kept in memory. It loads the resource and its state from the server and continues from there. When the job fails, the page shows the reason and a retry, and the failed resource does not vanish silently.”
Weak answers show a spinner that is lost on refresh, or a progress bar that moves on a timer.
“I would take it seriously, because for finance users a few cents means the numbers cannot be trusted. I would get the same data and compare the screen and the spreadsheet row by row. Usually the two totals are computed in different places: one adds rounded row values, the other rounds the sum, or one uses floating point in the browser. The fix is that a single place, normally the server, computes totals with exact decimal arithmetic, and both the screen and the export use that result. I would add a test with the reported data and check the export path for other totals.”
Weak answers dismiss a few cents as a rounding thing, or fix the screen and leave the export different.
“Interfaces for professionals are judged on a different scale. An accountant does not want delight, she wants to close the month with no errors and no waiting. I find that a demanding and honest kind of frontend work. In my current role I build screens for a purchasing team, and the moments I am proud of are small: a grid that stayed fast at a million rows, a form that never lost a draft. Here those problems come at a larger scale and in many languages, and I want to learn how they are solved across a whole suite of applications.”
Weak answers speak about beautiful design and do not mention the professionals who use the screens.
No figures on this page: pay depends on the level, the city and the offer in front of you. When the offer comes, the free salary negotiation script helps with that part.
ClapAssist is an AI interview assistant for Mac and Windows. It listens to the interview on your computer and shows you what to say, in short lines you can read while you talk. Your live interview audio and screen are never stored. Your resume and notes are saved to your account so the app fills them in on any computer. It is hidden from screen share on every plan, including Free; only you can see it.