The candidate answers correctly and the panel note says the opposite. Feedback comes back as "unclear thinking", "could not structure a response", "knows the tools but cannot explain them", or a bare "communication". The candidate replays the round afterwards and cannot find the mistake, because there was no mistake in the content — every fact was right, every decision was defensible, and the grammar was fine. The tell, if anyone had recorded it, is one answer of ninety to a hundred words delivered as a single sentence: no full stop, no pause, each point folded into the one before it, joined end to end with "and", "so" and "because". Filler words cluster near the start ("so basically", "actually", "like") and then disappear as the speaker settles into the run.
Any spoken round, and this is the single most common way a well-prepared fresher loses one. It appears most in the second and third question of an interview, once the nerves have peaked and before the candidate has found their rhythm, and it is worst on questions the candidate knows best — a question you know cold is a question you can say fast, and speed is the whole problem here. It shows up equally in English-medium and regional-medium candidates, and equally in students with strong and weak vocabulary, which is the clearest evidence that this is not an English problem. Online rounds make it worse, because there is no room feedback telling you the listener has fallen behind. Linked module: 6.7 (Communication for Indian students).
Two mechanisms combine, and neither is about language.
The first is physiological. Under pressure the body speeds up speech, shortens breath and shrinks the pauses between clauses. This happens to everyone, including the interviewer. Left alone, it produces a run-on: the mouth stays ahead of the plan, so the speaker keeps adding connectors — and, so, because — because stopping feels like failing. Filler words are a symptom of the same gap, not a vocabulary problem. Notice that the rushed answer is usually the shorter one in seconds, which is why "speak more" is the wrong advice.
The second is structural, and it is the one that actually gets scored. A listener holds an answer by dividing it into parts. If you never mark a boundary, they receive one undivided block and can hold none of it. Five correct decisions with no boundaries between them arrive as zero decisions, and the panel writes down what they can recall, which is nothing specific. The note "unclear thinking" is therefore accurate about what the panel experienced and wrong about what the candidate knew — the delivery was scored as the thinking, because delivery is the only part of the thinking anyone in the room can observe.
The confirming test takes one minute and needs no interviewer. Record any answer, wait ten minutes, play it back, and try to write down the separate decisions in it. If you cannot list them from your own recording, no panel could. Then measure the longest stretch with no full stop. Anything past twenty-five seconds is the failure in this entry.
What this is NOT: it is not accent, it is not grammar, and it is not a regional-medium disadvantage. The paragraph that failed was well-formed English. Working on pronunciation here fixes nothing, because nothing was mispronounced.
Three changes, none of which touch the content of your answer.
1 — The signpost sentence. Before any answer longer than thirty seconds, say how many parts are coming. "Two columns had missing values, and I treated them differently." "There were three things I tried, and the third worked." This costs three seconds and it opens boxes in the listener's head that the rest of your answer can fill. Say a small number — two or three, never six.
2 — Audible full stops. After each part, stop. A full stop the listener can hear is what divides your answer into things they can hold. In practice this means finishing a point with a short sentence rather than trailing into "and also…", then closing your mouth for two seconds. Two seconds feels enormous to you and completely normal to them. A pause reads as thinking, never as a gap.
3 — Slow the first ten seconds on purpose. Speech rate is set in the opening line and the rest follows it. Deliberately say your signpost slower than feels right; you will land near normal, because your internal clock is running fast.
The rehearsal that installs all three, over about two weeks: pick ten questions you can already answer, record each one, and score only for shape — spoken length, filler count, longest run without a full stop, and how many decisions a listener could repeat back. Re-record the worst one changing nothing but the shape. Do not rewrite the content, do not add vocabulary, and do not work on your accent. The whole repair is where the stops go.
What not to do. Do not try to slow down by speaking each word slowly — that sounds unnatural and is hard to sustain. The pace comes from the pauses between sentences, not from stretching the words inside them.
Write a signpost sentence at the top of every prepared answer in your notes, so the shape is rehearsed rather than invented in the room. Practise aloud, always: a silent read-through installs the content and none of the delivery, and the delivery is what fails here. Keep one number in your head as a guardrail — no stretch of speech longer than twenty-five seconds without a full stop — and check it against your own recordings weekly rather than by feel. In mock rounds, ask your partner for exactly one piece of feedback: "list the separate points you heard." Their list is your score, and it is more useful than any opinion about how you sounded. Before a real round, say your first three answers out loud once, because the run-on almost always happens in the first five minutes, before the voice has settled.
Q: "Take me through how you handled the missing values in that dataset." Strong answer: open with a signpost naming the number of parts ("two columns, and I treated them differently"), give each part as decision-then-reason in two or three sentences, stop audibly between them, label anything extra as extra, and close by offering to go deeper. Weak answer: the same facts, correct, delivered as one hundred-word sentence — the panel hears one noise and writes "unclear". If you notice mid-answer that you have been running on, the recovery is a single sentence: "Let me put that more clearly — there were two things." Nobody marks you down for restructuring out loud; they mark down what they cannot follow.