Applied GenAI Curriculum for AI PMs  ·  Depth Layer  ·  Eval Thread 5/5

Unit 16 — Evals Reference Shelf

The FAQ distilled from teaching hundreds of engineers and PMs — plus the sharp idea that product design and evaluability are the same problem. A lookup, not a linear read.

Tags
evals
Layer
DEPTH — ongoing reference, not linear reading
Objective
A place to look up trustworthy answers when a specific eval question arises mid-project.
Depends on
Unit 06. Closes the eval thread (06 → 07 → 09 → 14 → 15 → 16).

How to use this unit

The other eval units teach a method start-to-finish. This one is a shelf you return to when a specific question comes up: should I buy an off-the-shelf metric? build a tool? write an eval for this? put a check inline? Skim it now so you know what's here; come back when you need the answer. It closes with the single most useful reframe in the whole thread — that a product being "hard to eval" is usually a product problem, not an eval problem.

Metrics & tooling

Should I use ready-to-use / generic metrics? No

Generic scores — helpfulness, coherence, quality, BERTScore, ROUGE, cosine similarity — measure abstract qualities that may not matter for your task. Good scores don't mean your system works; they mostly create false confidence. They won't catch that a real-estate bot suggested an unavailable showing (an assertion catches that) or confused two client personas (a validated judge catches that).

The one legitimate use: as exploration signals to surface interesting traces worth reading — never as the measure of quality itself.

Should I build my own annotation / trace-viewing tool? Usually yes

The highest-leverage investment is a friction-free, domain-specific viewer that puts everything you need to judge a trace on one screen. Off-the-shelf dashboards nudge you toward generic metrics; a custom viewer (buildable in under a day with a notebook or AI coding assistant) keeps the team actually looking at data. Don't buy fancy tools before you've used what you have.

When to write an eval — and when not to

Should I practice eval-driven development (write evals before features)? Generally no

Unlike normal software, LLMs have infinite failure surface — you can't anticipate what breaks. Writing evaluators for imagined failures wastes effort and blocks you. Start with error analysis; write evals for failures you actually discover. Narrow exception: a crisp, known constraint like "never mention competitors" can be written up front.

Should I build an automated evaluator for every failure I find? No

First fix obvious gaps (you often just never specified "be concise" or a format). Then apply the cost hierarchy: cheap code checks (regex, schema, execution tests) for anything rule-shaped; an LLM judge only for subjective qualities and only for problems you'll iterate on repeatedly — because a judge costs 100+ labeled examples plus ongoing weekly maintenance and cross-team coordination. Always run a cost-benefit check: does this failure justify the investment?

How much should I fixate on model selection? Less than you think

Swapping models is the reflex many teams reach for first. Don't — without evidence. Do error analysis and ask whether the data actually says the model is the problem. Usually the fix is context, prompt, or a bug, not a new model.

CI evals vs. production evals

These are complementary systems; the key difference is the data they run on.

CI / CDProduction
DatasetSmall, curated (~100+): core features, regression tests for past bugs, known edge cases.Sampled live traces — usually no reference answer available.
MethodsFavor deterministic assertions (run often, cost matters).Lean on reference-free LLM judges, run async; track confidence intervals.
TriggerEvery change.Continuous sampling; alert if the CI lower-bound crosses threshold.

The loop that ties them: when production error analysis reveals a new failure pattern, add representative examples to the CI set so it can't regress.

Guardrails vs. evaluators (don't confuse them)

Guardrails — inline, before the user sees it

Fast, deterministic, explainable (regex, schema, block-lists, light classifiers). Target clear-cut, high-impact failures (PII, injection, malformed JSON). A false positive is a production bug, so keep them conservative. You'd almost never put a slow LLM judge here.

Evaluators — after the fact, async

Measure nuanced qualities rules can't (factual correctness, completeness). Feed dashboards and regression tests; they don't block the answer. Heavier compute (LLM judge) is fine because they run in the background.

An evaluator can double as a guardrail only if it's fast/cheap enough for the request path and the false-positive/false-negative trade favors blocking — high-stakes domains may accept blocking, creative ones may not.

The reframe: "it's hard to eval" is a product smell

The most common objection to evals is "our product is hard to evaluate." That's usually a product problem, not an eval problem: an output that's hard for you to verify is hard for your users too — in the worst case they must redo the work from scratch to trust it. Designing for verifiability comes before building evals.

✕ Answer-only

"Net revenue was $4.21M." Nothing to check — the user must redo the analysis to trust it. A 50-page report the doctor must re-read the whole chart to verify.

✓ Built for verification

Show provenance (which vetted source, who authored it), the assumptions made, intermediate numbers, and a flagged list of what couldn't be verified — each as a scoped unit to accept, edit, or reject.

Four questions to design a product for verification

  1. What does the user actually need to check?
  2. What trusted thing can they compare it against?
  3. What signals or heuristics do experts use to verify?
  4. What smaller units can they accept, edit, or reject?
Why this closes the eval thread Provenance and progressive disclosure make outputs checkable; breaking a workflow into scoped units (is this contradiction real? does this citation support the claim?) makes both users and evals able to grade the work. Evaluability and good product design are the same discipline. Before AI, verification happened incidentally while doing the work; with AI, verification is the bottleneck — so design for it on purpose.
Going deeper When your organization is building real eval infrastructure, the paid cohort course "AI Evals for Engineers & PMs" is a justified next step — but only once you're past the basics in this thread and actually shipping eval systems.

What this shelf gives you

Quick answers you can now cite mid-project:

That completes the eval thread. From here the remaining DEPTH units cover architecture internals, data, monitoring, and safety — pulled in whenever a live decision calls for them.