Most Accenture candidates practice the wrong thing. They stack ten "tech" prompts — cloud, ERP, AI — then open every one with a profitability tree. In the room, the interviewer is not scoring whether you know the word "migration." They are scoring whether you recognize the shape of the ask in the first ninety seconds and pick the move that shape rewards.
Accenture's own Case Interview Workbook already names those shapes: the great unknown, the parade of facts, and the back of the envelope. This piece teaches you to practice against those three openings, work one original payback case with dual-run math you can reproduce, and leave with a single assignment you can finish today. For process and track differences, use the Accenture case interview guide.
Why industry labels waste your reps
Accenture is a solutions and professional services firm with roughly $69.7 billion in fiscal 2025 revenue and about 779,000 people as of 31 August 2025, per its financial performance summary. That scale shows up in case flavor — technology, operations, and implementation sit next to classic strategy — but flavor is not a scoring dimension.
I score cases the way most Big Four and strategy interviewers do: clarity of problem definition, logic of the structure, quantitative hygiene, and whether the recommendation can survive contact with a client who has to execute. Reciting "cloud" or "platform" does not move those marks. Naming the dual-run period, the sequencing risk, or the fact you are deliberately ignoring does.
So stop building a library of industry costumes. Build a library of openings.
The three shapes Accenture already published
The workbook is not secret folklore. Accenture still links it from careers materials, and the case tips on its careers blog push the same habits: ask what success means, stay flexible, talk the thinking out loud. Cases in that workbook typically run 45 to 60 minutes total, with 30 to 45 minutes on the case itself.
| Shape | What you are given | Opening move that scores | What fails |
|---|---|---|---|
| Great unknown | One thin sentence, almost no data | Clarify objective and constraints before any framework | Dumping a memorized tree on an undefined decision |
| Parade of facts | Excess detail, mixed signal and noise | Name which facts you park and which two drive the answer | Trying to use everything; never converging |
| Back of the envelope | A number to estimate, thin inputs | Build a visible driver chain, sanity-check the result | Silent arithmetic with no units or check |
Those three map cleanly onto how interviewers allocate attention. In a great unknown, I listen for clarifying questions. In a parade, I listen for triage. In an envelope, I listen for structure of the estimate, not the final digit.
Ned's rule. If you cannot name the shape in one sentence before you draw a tree — "this is a parade; I will discard half the facts" — you are practicing recognition of industries, not performance under the scoring sheet.
Worked example: HarborLine dual-run payback
Invented company, invented numbers. Cover the solution and try the structure first if you want a clean rep.
Prompt. HarborLine Shipping, a regional carrier with $820 million in annual revenue, runs its booking and fleet systems in two owned data centers. The CIO wants a full public-cloud move in three years. The board asks: does the migration pay for itself on infrastructure economics alone, and if so when?
Shape. This is a back-of-the-envelope wrapped in a technology story. Do not open with "revenue, cost, competition." Open with: current run cost, target run cost, one-time migration cost, and the dual-run overlap where you pay for both.
Facts you would request (and then receive):
| Input | Figure |
|---|---|
| Current annual run cost (two data centers) | $74 million |
| Steady-state cloud run cost | $51 million |
| One-time migration spend (over three years) | $110 million |
| Dual-run length | 18 months |
| Dual-run legacy load while cloud ramps | 65% of current run cost |
Arithmetic you can reproduce:
- Steady-state saving = $74M − $51M = $23M per year.
- Dual-run penalty: 18 months at 65% of $74M ≈ $48.1M per year × 1.5 years ≈ $72M of overlap cost (order-of-magnitude; keep units attached).
- Total cost to arrive = $110M migration + ~$72M overlap ≈ $182M.
- Simple payback after completion ≈ $182M / $23M ≈ eight years.
Recommendation that scores. On infrastructure economics alone, HarborLine should not do a full three-year lift-and-shift. Recommend a phased cut: migrate the two workloads with the worst cost per transaction first, quantify any revenue or risk benefit the CIO claims separately, and only expand once dual-run length is proven under twelve months. The risk to watch is schedule slip that lengthens dual-run and destroys the already weak payback.
The trap is not the division. It is forgetting the overlap branch. Candidates who "know tech" still lose here because they total migration spend against annual savings and never price the period when both environments are live.
How to practice each shape without ten full cases
You do not need a new industry every night. You need three short drills that isolate the opening.
| Night | Shape | Twenty-minute assignment | Pass condition |
|---|---|---|---|
| 1 | Great unknown | One sentence prompt only. Spend five minutes writing clarifying questions out loud, then a two-branch tree. | You ask what "success" means before you structure |
| 2 | Parade of facts | A dense one-page fact pack. Circle the two facts that move the decision; write one sentence on each discarded cluster. | You can defend every parked fact in fifteen seconds |
| 3 | Back of the envelope | One estimation ask. Driver chain on paper, units on every line, one sanity check against a known benchmark. | A stranger can follow the chain without asking what a number is |
Accenture's careers guidance is explicit that interviewers want you to ask questions and define success rather than rush a clever answer. That is the great-unknown muscle. Parade practice is rarer on prep sites and more common in real rooms, because client materials are noisy. Envelope practice is where units and dual-run logic live.
Group cases and Strategy-track discussion formats (often called Potentia in candidate folklore) are separate skills. They still reward the same closing habit: a decision, a number or test, and a risk. Do not invent firm process detail that Accenture does not publish; the recruiting FAQ states that interview count and format vary by role.
A scoring heuristic from the interviewer's side
When I debrief a candidate-led Accenture-style case, I use a rough 10-point sheet. Treat the numbers as Ned's estimate of weight, not as Accenture's confidential form.
| Bucket | Points | What earns them |
|---|---|---|
| Clarify / shape recognition | 2 | Correct opening questions; named constraints |
| Structure | 2 | Two or three testable branches, not a template dump |
| Triage or estimation hygiene | 2 | Facts parked with reasons, or drivers with units |
| Decisive math | 2 | One calculation that changes the recommendation |
| Recommendation and risk | 2 | Decision + number + what would change your mind |
If you finish a practice case with a beautiful tree and no recommendation with a number, you left four points on the table. If you nail the math but never clarified the objective, you left two at the start — and those are the two I stop listening for later, because I no longer trust the rest.
Practice this today
Do one HarborLine-shaped payback on a different invented client in thirty minutes: write the four branches, invent plausible run costs, force an 12–24 month dual-run, and close with yes/no plus a risk. Then run a scored math or structure drill so someone other than you grades the tree.
CoachNed keeps everything open for seven days with no card; then $120 for a recruiting season or $49 a month. Start with a five-minute first rep at /start, or book the live voice case at /interview. For volume, use /consulting-cases and the /drills/math and /drills/structure engines. Process context sits in the Accenture case interview guide.
Frequently asked questions
Where can I find Accenture's official case interview practice material?
Accenture publishes a Case Interview Workbook covering the six-step solving loop and the three case shapes. Work it for method, not for dated industry examples; the shapes still match how interviewers open.
How long is an Accenture case interview?
Accenture's workbook describes cases that typically last 45 to 60 minutes, with about 30 to 45 minutes on the case itself. Exact length and number of interviews still vary by role, as the firm's hiring FAQ notes.
Do Accenture cases always involve technology?
No. Technology and implementation flavors are common because of what the firm sells, but the scoring sheet still rewards structure, math, and a commercial recommendation. A cloud label without dual-run or sequencing thinking is just costume.
Should I practice group cases for Accenture?
Assessment-centre formats that include group discussion appear in some processes; Accenture does not publish a fixed worldwide script. Practice the behaviors that transfer: propose a structure, own a number branch, and synthesize a recommendation before time expires.
How many Accenture practice cases do I need?
Fewer full cases, more shape drills. Six to eight deliberate reps across the three shapes, plus one timed payback with dual-run, beats twenty unread "tech" case PDFs. Stop when you can name the shape and the opening move before you draw. CoachNed is not affiliated with Accenture or any other firm named here.
Sources
- Accenture Case Interview Workbook (FY19 PDF) — official six-step loop, misconceptions, and three case shapes; typical 45–60 minute case length. Checked 2026-09-24.
- Secrets to a successful case-study interview | Accenture — firm careers guidance on asking questions, defining success, flexibility, and talking through assumptions. Checked 2026-09-24.
- Financial Performance | Accenture — fiscal 2025 revenue of $69.7 billion (12 months ended 31 August 2025). Checked 2026-09-24.
- Recruiting and Hiring Process | Careers FAQs | Accenture — interview count and format vary by role; assessments do not replace interviews. Checked 2026-09-24.