Skip to content
tezvyn:

Metrics

400 bites tagged Metrics — interview questions with model answers, and 60-second explainers.

LLMs & Generative AI1 min read

Evaluating image generation: FID and IS

FID compares feature distributions of real and generated images, lower is better; Inception Score rewards confident, diverse classes but ignores real data. knowledge of generative image metrics.

Design Systems1 min read

Quantifying the cost of a missing component

Scan repos for duplicate hand-rolled tables, count implementations and lines, estimate redundant build and maintenance hours plus defect rates. Whether you justify investment with evidence of waste.

Design Systems1 min read

Isolating the design system's velocity impact

Use comparison groups, difference-in-differences or before-after baselines, and control for confounders; cite mechanism data like component reuse. Whether you can argue causation, not correlation.

Design Systems1 min read

Constructing a design system ROI formula

ROI equals (benefits minus cost) over cost; benefits from time saved, defect reduction, and faster onboarding; gather data via analytics and tickets. Whether you can frame the system as an investment.

Design Systems1 min read

Quantifying migration cost for a major release

Pre-release, scan codebases for affected usages and estimate engineer-hours; post-release, track actual upgrade lead time and effort. Whether you can estimate and measure migration effort.

Design Systems1 min read

Designing a UI consistency score

Combine signals like token vs hard-coded values, design-system vs off-system components, and override rates, then automate via AST scans and CI. Whether you can quantify consistency objectively.

Design Systems1 min read

Tracking design system adoption technically

Scan codebases for component imports via AST, compute adoption versus off-system usage, and aggregate into a dashboard over time. Whether you can instrument adoption with real data.

Design Systems1 min read

Reversing stagnating design system adoption

Find friction via interviews, reduce it with DX, codemods, and docs, then prove value with adoption and velocity metrics. Whether you treat adoption as a product problem.

Design Systems1 min read

Building a design system quarterly roadmap

Combine usage and adoption data, support tickets, stakeholder interviews, and a debt ledger into a scored backlog with a deliberate debt allocation. Whether you prioritize with evidence, not opinion.

Design Systems1 min read

Measuring design system documentation effectiveness

Search-with-no-results, time-to-first-component, support-ticket deflection, tied to analytics and registry data. outcome metrics over vanity page views. citing only page views or NPS with no instrumentation.

Databases & Architecture1 min read

Choosing a time-series database for metrics

High-ingest timestamped writes, time-window queries, retention and downsampling, time-optimized compression. fit of TSDBs to append-heavy time data.

Data Science & Analytics1 min read

Reconcile rising sign-ups with falling revenue per user

Reconcile the metrics via total revenue and segment mix, frame the tradeoff, recommend an action. business judgment and data storytelling. declaring one team wrong instead of synthesizing both.

Data Science & Analytics2 min read

Framing ad-load tradeoffs: revenue versus retention

Define revenue plus guardrail engagement metrics, run a long-enough experiment to see retention effects, and weigh short-term lift against lifetime-value erosion. balancing competing metrics over time.

CSS & Design Systems2 min read

Driving and measuring design system adoption

Lower friction with migration paths and great docs, incentivize over mandate, and measure adoption via package usage, component coverage, and code analysis. Strategy plus instrumentation for adoption.

Computer Vision1 min read

How is IoU computed and why prefer mIoU?

IoU is intersection over union of predicted and true pixels; mIoU averages per class; pixel accuracy is dominated by background. segmentation metrics under imbalance. equating accuracy with IoU.

Computer Vision2 min read

Evaluating generative models with FID versus IS

FID compares Inception feature distributions of real and fake images via Frechet distance between two Gaussians; it uses real data as reference and detects diversity issues, unlike IS which uses no real… rigorous generative evaluation.

Cloud Platforms1 min read

Diagnosing slow auto-scaled PaaS workloads

Application metrics like request latency, throughput, and DB query time; infrastructure metrics like CPU, memory, and scaling lag. layered debugging under load. jumping to add instances without isolating the real bottleneck.

Cloud Platforms1 min read

Difference between metrics and logs

Metrics are aggregated numeric time series good for trends and alerting; logs are discrete timestamped event records good for detailed root-cause analysis. observability fundamentals.

Analytics & Metrics2 min read

Calculate MRR with SQL including annual plans

Sum monthly_price for subscriptions active this month, filter on start and end dates, and normalize annual plans by dividing annual price by 12. correct MRR definition and date filtering.

Analytics & Metrics1 min read

Calculating Daily Active Users in SQL

Need per-event user_id and timestamp and a clear active definition; count distinct user_id within the day in a fixed timezone. Metric definition plus dedup SQL. Counting rows or fuzzy date-boundary and timezone handling.

Analytics & Metrics2 min read

Cohort analysis for an onboarding change

A cohort groups users by a shared start trait; compare pre and post Jan-1 signup cohorts on retention by age. Cohort reasoning and clean framing. Comparing calendar periods instead of cohort age, confounding seasonality.

Growth & Experimentation1 min read

Components of a testable A/B hypothesis

A specific change, a predicted directional effect on one primary metric, a rationale, and a measurable success threshold. experimental rigor before launch.

Docker & Kubernetes2 min read

Scaling on queue length with the HPA

Expose queue length through an external metrics adapter behind the metrics API, point the HPA at that external metric with a target per pod; KEDA packages this. external-metric autoscaling.

Docker & Kubernetes1 min read

Diagnose a Prometheus cardinality explosion

Find offenders via TSDB stats and topk count by __name__, identify unbounded labels, then drop or aggregate them with relabeling. operating Prometheus at scale. just scaling memory without fixing label design.

Get Metrics bites daily.

Five a day, five minutes, offline. With quizzes so it sticks.

Open testing — you’ll join as an early tester.