holocron
中/en
← Business
Business

Jobs to Be Done

One line: Customers don't buy a bundle of product attributes — they "hire" a product to get a specific job done. That's the lens both episodes use to split apart markets that look homogeneous from the outside: the Nintendo episode uses it to find customer jobs nobody else was bothering to serve; the NVIDIA episode uses it, in a single passing line, to frame LLM training as merely the current high-water mark on a much broader spectrum of compute-hungry jobs — not the whole story.

The Mechanism — start from the job, not the product's specs

  • Surface-similar products can be doing completely different jobs: David's example is Halo Infinite versus Fortnite — to "an alien, or let's say... a 60-year-old person" watching him play Halo Infinite against someone else playing Fortnite, "on the surface, it looks pretty similar" (both are live-service games where you shoot at people), but "the job to be done of those two experiences is completely different, and those are very different markets" (Nintendo 之二：主机战争（1989-2023）). Ben generalizes the same point one layer further: hanging out with friends inside a game, killing time on Mario Kart in your living room, and playing a slot-machine-style game on your phone during a flight are "such different human experiences and jobs to be done" that it's "totally crazy to call them all gaming" (Nintendo 之二：主机战争（1989-2023）).
  • A product with no obvious positioning can be nailing two completely different jobs at once: the Gameboy — dim black-and-green screen, slow — was simultaneously serving "games for kids" and "games for a casual adult killing time," and David isn't even sure Nintendo planned it that way: "I don't know if Nintendo thought about this consciously" (Nintendo 之二：主机战争（1989-2023）).
  • The job nobody is taking seriously is where the biggest untapped market hides: the moment Nintendo of America's Minoru Arakawa saw the Tetris demo, he understood — "This is not a kid's game. This is an adult's game." — which is what produced the "Punish Your Father" Father's Day campaign and the airplane-magazine ads aimed squarely at business travelers; 46% of US Gameboy players ended up being adults (Nintendo 之二：主机战争（1989-2023）).
  • The same vocabulary can also place a whole technical/resource spectrum, not just segment human customers: in the NVIDIA episode, "jobs to be done" isn't used to slice up a customer base at all — it's used to locate LLM training on the spectrum of compute-hungry tasks in general: "LLMs are the current maxima in human history of jobs to be done that require a ton of compute." The implication is that most AI workloads (including image-generating diffusion models) don't look anything like LLMs, so compute demand can't just be extrapolated off the LLM-training curve (NVIDIA 之三：AI 时代的黎明（2022-2023）).

Cases in This Library

Nintendo: The Console Wars — the core lens of the handheld analysis (used twice, then named explicitly at the wrap)

This is where the material engages with jobs-to-be-done most deliberately and at the greatest length. It first surfaces in the Gameboy/Tetris story (Playbook #2): rather than positioning the Gameboy as a cheaper toy for kids, Nintendo of America — prompted by the Tetris demo — recognized it was serving two completely different jobs at once, and built "Punish Your Father" and the airport-magazine ads to go straight at the adult, casual-killing-time audience. It surfaces again in the market-segmentation discussion (Playbook #11), where the Halo-vs-Fortnite example is used to split "gaming" into casual / mid-core / core tiers, which the hosts then map onto three distinct corporate strategies — Sony (core: top spend, best graphics), Microsoft ("the Netflix of gaming": cross-device subscription), Nintendo (mid-core plus quality-focused casual). At the wrap, David singles jobs-to-be-done out by name as one of the episode's biggest takeaways, noting the handheld market "really highlights this," and summing up the Gameboy and DS as "hyper serving two jobs to be done" — games for kids and games for casual adult players — "Nintendo just got it right." He extends the point once more, more loosely, to mobile gaming: even though the experience is bad, "a lot of people that are playing those games, that game is accomplishing a job to be done for them" (Nintendo 之二：主机战争（1989-2023）).

NVIDIA Part III — a single, borrowed use, not a developed framework

Unlike the Nintendo episode, this one never treats jobs-to-be-done as a standalone analytical tool — the phrase appears exactly once, inside Bear Case #5 (the training-to-inference debate), and it isn't applied to a customer segment at all. Ben uses it to bound how much of the compute story LLM training actually represents: "even today the vast majority of AI workloads don’t look like LLMs" — LLMs are just the current historical high point on the spectrum of "jobs to be done that require a ton of compute," while a lot of other work — diffusion-model image generation, for instance — falls among the tasks that "don’t require an entire internet of training data and a trillion parameters to pull off." The line does rhetorical work against the assumption that compute demand simply keeps extrapolating off the LLM-training curve (NVIDIA 之三：AI 时代的黎明（2022-2023）).

Discussion and Boundaries

  • The two episodes don't use this symmetrically: the Nintendo episode treats jobs-to-be-done as a named, repeatedly-applied analytical tool that gets its own callout at the wrap; the NVIDIA episode uses the phrase once, more as a borrowed turn of phrase to place LLM training on a compute spectrum than as a framework being argued through. This page records both as they occur rather than forcing them into the same shape.
  • How thin the framework's footprint is in this library: at the wrap, David explicitly measures jobs-to-be-done against 7 Powers 护城河框架 — "We talked about this sometimes on Acquired, not a lot, certainly not as much as we talked about seven powers and other ideas" — which is also an honest description of the source material itself: these two episodes are, so far, the only place it appears in this library.
  • Flagged as possibly unintentional on Nintendo's part: the Nintendo episode carries its own hedge — David says "I don't know if Nintendo thought about this consciously" — meaning the hosts aren't claiming Nintendo actually designed the Gameboy using jobs-to-be-done language at the time; this is Acquired applying the lens retrospectively, not a claim about Nintendo's real internal reasoning.