The interviewer is not grading your marketing vocabulary. In a typical first-round launch prompt, the score sheet rewards one decision: go, no-go, or pilot, with the arithmetic that makes that call inevitable. Most candidates still open with Product-Price-Place-Promotion. I stop listening about ninety seconds in.
The thesis is simple. A launch case is incremental contribution after cannibalization, tested against a year-1 volume you can defend without inventing TAM. Fail that test and the rest of the tree is decoration. After this piece you should be able to size a credible year-1, subtract displaced core-product profit, find breakeven units, and recommend with a kill condition.
What "failure rate" actually means
Prep folklore loves "95% of new products fail." That number floats around attributed to Clayton Christensen, but what he and co-authors actually wrote in The Innovator's Solution is sharper and less meme-friendly: more than 60% of development efforts are killed before launch, and of the ~40% that ship, another ~40% never become profitable — so roughly three-quarters of development spend does not succeed commercially. Different definition, different number.
Retail-shelf data is more concrete. Nielsen's 2015 Global New Product Innovation work reported that of more than 60,000 new European SKUs, only about 55% were still in distribution at 26 weeks and only 24% survived a full year. A later NIQ analysis (2023) pushed back on the doom narrative for a different success definition: among new items that already had national distribution, 52% grew sales in year two; across all new items the "vitality" rate was 31%. MIT Sloan Review, using almost 9,000 new products that achieved broad distribution at a large U.S. retailer, found only about 40% still sold three years later.
| Source | What was measured | Rough takeaway |
|---|---|---|
| Christensen et al., Innovator's Solution | Development spend and profitability | ~75% of development money fails commercially |
| Nielsen (2015), European SKUs | Still in distribution at 12 months | ~76% exit within year one |
| NIQ (2023), national-distribution items | Sales growth in year two | ~52% "vital" |
| MIT Sloan retail study | Still on shelf at year three | ~60% gone |
When a candidate says "most launches fail," I ask which definition. Shelf death, unprofitable P&L, or abandoned R&D are three different cases. In the room, pick one and move to your client's numbers.
Adoption is an S-curve, not a year-1 share grab
Frank Bass's 1969 Management Science model treats first purchases as a mix of innovators (external influence) and imitators (word of mouth). Cumulative adoption is an S-curve; peak sales come later than the press release. Everett Rogers's adopter categories — innovators ~2.5%, early adopters ~13.5%, early majority ~34%, late majority ~34%, laggards ~16% — are statistical ideal types on a normal timing distribution, not a guarantee that your SKU will hit 16% of category buyers in month twelve.
In interview math that means: year-1 volume is usually a thin slice of eventual peak, unless the firm already owns the channel and the buyer. Attach to an installed base beats "category × share." If you quote Rogers percentages as if they were your forecast, you are performing diffusion theory, not solving the case.
Public companies rarely disclose model-level launch P&Ls. What they do disclose is how launches move the whole book. Apple's FY2025 Form 10-K shows iPhone net sales of $209.6 billion (up from $201.2 billion in 2024) and states plainly that new product introductions affect net sales, cost of sales, channel inventory, and demand as buyers wait for the next release. That is cannibalization and channel fill, written into a filing — the same tension you should name in a consumer or hardware launch case.
The four kill tests (say them, do not mnemonic them)
| Kill test | Question | Number the interviewer wants |
|---|---|---|
| Reachable demand | Who buys in year 1 from our pipeline? | Units or accounts, not TAM |
| Unit economics | Contribution after variable cost and cost-to-serve | $ per unit / account |
| Cannibalization | What core revenue dies when this sells? | Displaced units × core contribution |
| Capability / channel | Can we sell and ship without inventing a new company? | Cycle time, slotting, capacity |
Price lives inside buyer willingness and contribution. Promotion lives inside channel. Spec lives inside the job. You may use 4P words; you do not need a speech.
Ned's rule. Never recommend a national launch whose year-1 volume sits below breakeven on incremental contribution (new contribution minus cannibalized core). If you need heroic share or zero cannibalization to clear the bar, the honest answer is pilot or no-go.
Worked example: PeakLine HardFreeze X
Prompt (invented). PeakLine sells soft-sided coolers to U.S. outdoor retailers. Last year: 1.2 million SoftCooler Pro units at $95 ASP, variable cost $52, contribution $43/unit. Engineering built HardFreeze X, a hard cooler with a rechargeable ice pack. List ASP $189, variable cost $98, contribution $91/unit. Fixed launch spend (tooling, first production run, retailer slotting, launch media): $22 million. Sales wants national launch for summer. CFO wants go / no-go.
Step 1 — Incremental contribution, not list margin
Every HardFreeze that sells will displace some SoftCooler demand. Use a base cannibalization rate of 0.25 SoftCoolers lost per HardFreeze sold (trade-up inside the same brand; invent a rate, state it, stress-test it).
Incremental contribution = $91 − (0.25 × $43)
= $91 − $10.75
= $80.25 ≈ $80 / unit
If someone quotes "$91 contribution, so we are fine," they failed the case. The firm only keeps about $80 of new economics per unit after the soft cooler that did not sell.
Step 2 — Breakeven volume
Breakeven units = $22,000,000 / $80 ≈ 275,000 units in year 1
Raise cannibalization to 0.40 and incremental falls to $91 − $17.20 = $73.80; breakeven rises to ~298,000. That sensitivity belongs on the whiteboard.
Step 3 — Year-1 volume without TAM theatre
Do not size "all coolers in America." Build from PeakLine's reach:
| Source of year-1 demand | Logic | Units |
|---|---|---|
| Loyalty / email upgrade offers | 400k engaged buyers × 12% convert | 48,000 |
| Core retail partners' first-order commit | Three chains, demoed planogram | 140,000 |
| Spillover / online Direct | Residual after stockouts | 30,000 |
| Base year-1 | 218,000 |
Rogers would remind you that 218k is still early-adopter territory relative to a multi-year peak. Bass would remind you that imitator-driven volume arrives later. Neither model rescues a year that sits 57k units under a 275k breakeven.
Stretch case (sales fantasy): 18% loyalty convert + 200k retailer fill = ~302k. That clears 275k only if cannibalization stays at 0.25 and retailers actually take the inventory. I would not bet a national mold spend on the stretch.
Step 4 — Recommendation
No-go on full national launch. Fixed spend of $22m needs ~275k incremental-economics units; the defendable base is ~218k. Recommend a two-region paid retail pilot with capped tooling (cut fixed to ~$8–10m), measure true SoftCooler displacement for one season, and re-ask go/no-go when observed cannibalization and sell-through are known. Kill if pilot sell-through implies national volume under 250k or if cannibalization exceeds 0.35.
That is the answer shape: number, incremental economics, volume gap, conditional path. Not a brand story.
Where candidates leak points
TAM × arbitrary share. "U.S. cooler market is $4B, we take 2%…" The interviewer cannot audit the 2%. They can audit 12% of a 400k list.
Ignoring cannibalization. Same brand, same aisle, same season — assume some displacement unless the case says greenfield category.
Confusing list margin with incremental contribution. SoftCooler's lost $43 is as real as HardFreeze's $91.
Treating Rogers/Bass as a forecast engine. Use them to push year-1 down, not to invent precision.
Launching because capability exists. A finished mold is a sunk cost argument in disguise. The decision is forward-looking cash.
When this tree is the wrong tool
If core SoftCooler profit just collapsed, diagnose profitability before inventing a SKU. If the "product" is a new country, that is market entry. If the question is which growth vector among four, start with Ansoff sequencing. If the firm cannot ship anything because of an org freeze, that is not a GTM plan.
How I want you to open
"I would treat this as year-1 volume versus breakeven on incremental contribution after cannibalization, not as a 4P plan. SoftCooler loses about $11 of contribution for every four HardFreeze units at a 0.25 displace rate, so we keep roughly $80 per unit and need about 275k units to clear $22m fixed. My base demand from loyalty and retailer commits is about 218k, so I would not launch nationally — I would pilot, measure cannibalization, then revisit."
That opening already sounds like someone who has sat on the other side of the table.
Practice this on a clock
Take any launch prompt and force three lines before structure: (1) reachable year-1 units, (2) contribution net of cannibalization, (3) breakeven versus that volume. Then run a timed structure rep so the kill tests become muscle memory — everything on CoachNed is open for seven days, no card; then $120 for a recruiting season or $49 a month.
Or start a five-minute typed case at /start and score whether your first turns actually threaten a go/no-go.
CoachNed is not affiliated with McKinsey, BCG, Bain, Deloitte, or any other consulting firm named in this article.
Frequently asked questions
How do you structure a new product launch case interview?
Open with reachable year-1 demand, unit economics, cannibalization, and channel capability. Recommend go, no-go, or pilot with a numeric kill condition. Save marketing levers for how you would close a gap you have already measured.
What is a typical new product failure rate?
It depends on the definition. Nielsen's European SKU work found only about a quarter still in distribution after one year; NIQ's later "vitality" metric shows roughly half of nationally distributed items growing in year two. Christensen's development-spend framing implies most R&D money does not produce a profitable product. State which definition you mean.
How do you account for cannibalization in a launch case?
Estimate displaced units of the core product per new unit sold, multiply by core contribution, and subtract from new-product contribution. Stress-test the displace rate. Public filings (for example Apple's discussion of product introductions and channel inventory) show the same economic tension without giving you a model-level rate — you still have to invent a coherent one in the case.
Should I use the Bass model or Rogers curve in the interview?
Use them as intuition: year-1 is early on an S-curve, so do not assume peak share immediately. Do not fit p and q parameters on a whiteboard unless the interviewer hands you a time series. Attach rates and retailer commits beat diffusion algebra.
What is the difference between a launch case and a market entry case?
Launch usually assumes you already sell to the buyer or channel and asks whether this offer should exist. Market entry asks whether you should show up in a geography or segment at all. If the product is the entry, say so and merge the trees.
Sources
- Nielsen Global New Product Innovation Report (2015 PDF) — European SKU survival: ~55% at 26 weeks, ~24% at one year. Checked 2026-09-24.
- NIQ: The CPG Innovator's guide to vitality (2023) — 52% of nationally distributed new items grow in year two; 31% across all new items. Checked 2026-09-24.
- Bass, F.M. (1969). A New Product Growth for Model Consumer Durables. Management Science (JSTOR) — S-curve adoption; innovator vs imitator purchase timing. Checked 2026-09-24.
- Apple Inc. Form 10-K for fiscal year ended Sept 27, 2025 — iPhone net sales $209.6B; product introductions affect sales, costs, and channel inventory. Checked 2026-09-24.
- MIT Sloan Management Review: Why Great New Products Fail — ~40% of broadly distributed new products still sold after three years. Checked 2026-09-24.
- Christensen, C.M., Raynor, M.E. The Innovator's Solution — ~60% of development efforts killed pre-launch; of those shipped, ~40% unprofitable. Book cite; no public URL for the passage. Checked 2026-09-24.