Amazon — Device Economics — Product Sense & Metrics

Role context: Data Scientist II, Device Economics (Decision Science / DSO) · Est. study time: 35 min · 4 practice questions

How to prepare for this role

Device Economics is a forecasting / econometrics / portfolio-economics role, not product analytics or recsys. It's judged on understanding the loss-leader device economics (downstream ecosystem value, not device margin), forecasting demand — including new products with no history — evaluating forecasts honestly, and measuring substitution/cannibalization causally. Amazon's loop also tests Leadership Principles (STAR) and ML depth + breadth.

The day-to-day is owning science-based forecasts that drive multi-million-dollar decisions: pre-launch and annualized demand forecasts for Amazon Devices, substitution/cannibalization across the portfolio, and economic models for go/no-go launches — communicated to PMs, marketing, and leadership who must trust the numbers. The instinct that separates strong candidates: knowing that device unit sales/margin is the wrong lens (devices are loss leaders) and that the real objective is downstream ecosystem value.

Where to spend your prep time:

  • Device-economics product sense (this article) — the loss-leader model, downstream impact, and why a forecast must model portfolio substitution.
  • Forecasting and causal inference — new-product forecasting, honest forecast evaluation (bias, asymmetric costs), and measuring cannibalization without an experiment. See the A/B Test & Causal Inference section (reframed around forecasting + causal).
  • SQL, econometrics, and communication — be fluent in SQL/ML, think like an economist (elasticity, substitution), and practice building trust in a forecast with non-technical leaders.

The through-line: forecast demand and substitution well, value devices in downstream-ecosystem terms, and communicate uncertainty leaders can act on.

What Device Economics actually is

Amazon Devices (Echo, Fire TV, Kindle, Ring) sells hardware often at or below manufacturing cost — the strategy is to "make money when customers use the products, not when they buy them." The identity to carry into every answer: a loss-leader hardware portfolio whose real value is the downstream ecosystem spend it drives (Prime engagement, shopping, subscriptions, content, ads), so the economics are device margin plus downstream impact. Amazon's internal downstream impact metric assigns each device a value based on the customer's Amazon-ecosystem spending after buying it — which is why the Devices division could run large losses (reportedly ~$25B over 2017–2021) and still be a rational bet.

Two things make this different from a normal analytics job:

  • Device margin is the wrong objective. Because devices are loss leaders, the metric that matters is downstream impact / device lifetime ecosystem value, not hardware profit. A device that loses money on the unit but drives years of Prime engagement and Amazon spend can be a great investment.
  • Forecasting drives the decisions, and the portfolio cannibalizes itself. Pre-launch and annualized forecasts size supply and go/no-go bets, and a new device substitutes demand from existing ones — so a forecast must model the whole portfolio and separate incremental demand from substituted demand. And a forecast is only useful if it's unbiased and decision-aware, not just accurate on average.