In a typical first-round slate of eight remote cases, I score two or three candidates as if they never worked the problem at all. They did the math. They just never said it. Phone and video cases are not softer versions of the room interview. They are the same score sheet with the scratch paper removed from view, so anything you do not speak does not count.
That is the thesis most guides miss. Candidates treat remote prep as lighting, eye contact, and "virtual presence." Those help a little. What moves the score is auditability: can the interviewer reconstruct your structure, formula, units, and implication without seeing your page? After this piece you will have a spoken audit script for structure and math, a short exhibit readout that survives a bad share, a recovery play when the link dies, and one practice rep that mirrors the medium.
The score sheet did not change. The channel did.
Firms score remote cases on the same axes they use in person: structure, quantitative rigor, judgment on data, and clear recommendation. McKinsey's interviewing page frames the problem-solving interview as discussing a client scenario so the firm can see how you structure ambiguity, handle facts, and articulate a conclusion. Bain's interviewing page is blunt that consulting cases reward logical structure, precision, and creativity, and that there is not always one right answer. Neither firm publishes a "discount for Zoom" line.
What changed is the evidence stream. In a conference room I can glance at your tree, catch a unit error mid-calculation, and nod when a branch lands. On video I get a face and a voice. On a phone case I get only the voice. Silent competence looks like fumbling. Explicit, slightly slower narration looks like a partner in a client call.
| Signal | What I can see in person | What I get on phone or video |
|---|---|---|
| Structure | Glance at your tree | Only the branches you name |
| Math | Scratch work on the table | Formula, units, answer, implication |
| Exhibits | Shared paper or whiteboard | Axis names and one takeaway |
| Alignment | Nod, frown, lean-in | Verbal check-ins you initiate |
| Recovery | Easy to pause the room | You must name the drop and restart |
McKinsey also publishes assessment integrity expectations that apply to virtual interviews: participate alone, use one screen, discard notes afterward, and do not record, use generative AI, use a calculator, or use prepared notes. Treat those as the floor. A quiet room is not optional polish; it is part of the integrity bar.
Bandwidth is a floor. Narration is the skill.
Zoom's system requirements recommend about 1.2 Mbps up and down for 720p one-to-one video, and only 60–80 kbps for VoIP audio. Most home connections clear that. Stability beats peak speed. Wired Ethernet or sitting next to the router beats a heroic "HD" setting on flaky Wi-Fi. McKinsey's own virtual tips on the interviewing page are practical: quiet space, headphones, pen and paper within reach, camera on, virtual backgrounds off, and disable note-taking AI bots that may join and record.
Do that once, the day before. Then stop optimizing the room. The remaining gap is whether your thinking is inspectable when nobody can see the page.
Ned's rule. If the interviewer cannot audit the step without seeing your paper, it did not happen. Say the quantity, the formula, the units, and the implication before you go quiet for arithmetic.
Open so the interviewer can correct you early
Remote cases punish the wrong structure harder than in-person ones, because misalignment stays invisible longer. Open in four moves, spoken as complete sentences:
- Restate the objective in plain business language.
- Ask one clarifying question that would change the tree.
- Ask for 20–30 seconds to structure.
- Present two or three named branches, then say which one you want first and why.
Bad open (common on phone): "Okay so profitability… drivers… um… market?" The interviewer has nothing to correct.
Better open: "I want to confirm the goal. GreenWave wants to know whether a $4.2 million plant retrofit pays back inside three years on contribution margin alone. One clarifying question: are we treating volume as fixed, or can the retrofit unlock capacity? Give me half a minute, then I will walk a short structure."
That last sentence is not filler. It buys you thinking time without sounding like you are stalling, and it invites a course-correction before you build the wrong tree.
Make structure audible with counts and names
Two habits fix most "unclear framework" scores on remote cases.
First, declare the count before the list: "Three branches." Then name each once, in one breath each. Second, re-name the branch every time you return to it. "Second" is not a label. "Customer acquisition cost" is.
A workable spoken tree for GreenWave:
- "Three branches. First, incremental contribution from the retrofit. Second, upfront and operating cost of the retrofit. Third, risk and timing if volume or price miss."
- "I would start with incremental contribution, because if the volume upside is thin, the cost debate does not matter."
On video you may briefly hold the paper to the camera after you have spoken it. On phone, the spoken version is the only version. Offer an alignment check once: "Does that structure work before I dig into contribution?" Then dig. Do not ask permission after every sentence.
Narrate math before you compute
Silent math is the highest-frequency remote fail I see. The fix is a five-line audit trail. Speak it even when the arithmetic is easy.
Worked example you can reproduce. GreenWave's retrofit claims +8,000 units per year at $95 contribution per unit, with a $4.2 million capital cost and $180,000 extra annual opex. Payback on contribution alone:
| Step | What you say | Number |
|---|---|---|
| Quantity | "I am calculating annual incremental contribution." | — |
| Formula | "Units times contribution per unit, minus extra opex." | — |
| Inputs | "8,000 units, $95 each, less $180k opex." | 8,000 × 95 = 760,000; 760,000 − 180,000 = 580,000 |
| Capex | "Payback is capital over annual contribution." | 4,200,000 ÷ 580,000 ≈ 7.2 years |
| Implication | "That is outside a three-year hurdle on contribution alone, so I would not recommend on volume upside alone." | — |
Spoken as one block: "I am calculating annual incremental contribution. Formula is units times contribution per unit, minus the extra opex. Using 8,000 units at $95, that is $760,000, less $180,000 opex, so about $580,000 a year. Capex of $4.2 million over $580,000 is roughly 7.2 years. That misses a three-year hurdle on contribution alone, so I would not green-light on volume upside alone; I would next ask whether price or mix improves contribution enough to cut payback under three."
Notice what the interviewer can score without seeing your page: correct formula, sane units, ballpark division, and a business conclusion. A candidate who mutters for 40 seconds and then says "seven years?" earns a weaker math mark even with the same answer.
If you need quiet for the last arithmetic, say so: "Give me ten seconds to finish the division." Then return with the number and the implication. Unannounced silence still reads as stuck.
Exhibits: name, pattern, implication
Shared slides on video and verbal exhibits on phone fail the same way. Candidates read cells. I am listening for judgment.
Use a four-beat readout:
- Ask for 20–30 seconds.
- Name the chart and the axes.
- State one pattern.
- Translate to a decision, then stop.
Example: "This is revenue growth versus margin by segment. Enterprise grows slower but carries roughly double the margin of SMB. If the goal is profit in three years, I would prioritize enterprise economics next, not raw growth."
If the share is blurry, ask once to enlarge a number. That is careful. Narrating every bar is not.
Presence is pacing, not theater
"Virtual presence" advice often drifts into performance tips. From the scoring chair, presence is mostly load reduction for the listener.
- Speak about 10% slower than feels natural. Packet loss eats the ends of sentences.
- Pause after each branch. Do not fill the pause with "so" or "um."
- Look at the camera for the final recommendation on video; on phone, the recommendation must be answer-first because there is no face cue.
- If you need time, name it. "Thirty seconds to structure" beats rushing into a half tree.
Do not apologize more than once for a tech hiccup. Fix it, restate the last point, continue. McKinsey's virtual guidance also says to ask the interviewer to repeat a question if it was unclear. Use that. Guessing through a garbled prompt is how remote cases go off-map.
When the connection drops
Treat a drop like a client call, not a crisis. Keep the recruiter number and a charged phone with a hotspot ready. If audio dies, say calmly that you lost the line, switch to phone audio or a dial-in if the invite has one, and restart from the last completed step: "Before we dropped, I had contribution at about $580k a year and payback near 7 years. Shall I continue from risk, or do you want me to restate the structure?"
Do not pretend nothing happened. Do not rebuild the whole case from minute one unless asked. One clean restart sentence is enough.
Practice the medium, not just the content
Typed drills build structure and math speed. They do not train the remote channel. You need reps where the only evidence is your voice.
Concrete assignment for today:
- Set a 18-minute timer.
- Run CoachNed's live voice case at /interview (CopperCup, spoken turns, seven-score debrief). It is itself a remote spoken case: no interviewer can see your notes, so the same audit habits apply.
- After the debrief, replay only the first three minutes of your recording or notes. Mark every place you went silent on math or used "second branch" instead of a name.
If you want a shorter warm-up first, take three typed turns at /start, then move straight to voice. Everything on CoachNed is open for seven days with no card; after that it is $120 for a recruiting season or $49 a month.
Frequently asked questions
Are phone case interviews easier than in-person?
No. The case and the score sheet are the same. Phone removes the visual channel entirely, so structure and math must be spoken or they do not score. Convenience is not ease.
Can I use a calculator in a virtual case interview?
Usually no. McKinsey's published integrity expectations forbid calculators and generative AI in assessments and interviews unless an approved accommodation applies. Follow the invite for your firm, and practice mental math with units spoken aloud.
How do I show my structure if the interviewer cannot see my paper?
Speak a count, then name each branch, then name the priority branch and why. On video you may briefly hold the page up after speaking it. On phone, the spoken version is the only version.
Should I look at the camera or the screen during a video case?
Look at the screen while you listen and work. Look at the camera when you deliver the recommendation and when you check alignment. Constant camera stare looks rehearsed; never looking up looks disengaged.
What should I test the day before a remote case?
Join the exact platform link, confirm mic and camera, disable AI note-takers, charge a phone hotspot, and write the recruiter backup number on paper. Then leave the hardware alone and rehearse narrated math. CoachNed is not affiliated with McKinsey, BCG, Bain, Deloitte, or any firm named here.
Sources
- Interviewing at McKinsey — problem-solving interview framing and published virtual interview tips (quiet space, headphones, camera on, disable AI note tools). Checked 2026-09-24.
- McKinsey Assessment Integrity Expectations — alone, one screen, no recording, no calculator, no generative AI, no prepared notes. Checked 2026-09-24.
- Bain Interviewing — consulting case interviews reward logical structure, precision, and creativity. Checked 2026-09-24.
- Zoom system requirements (Windows, macOS, Linux) — bandwidth guidance for 720p one-to-one video (~1.2 Mbps) and VoIP audio (60–80 kbps). Checked 2026-09-24.