Designing for Discovery – a Practical Guide

Who this is for: founders, PMs, designers, and engineers who keep hearing “we already built that.”

Outcome: measurably higher feature adoption with fewer support pings and shorter time‑to‑first‑use.

⸻

Five principles

1. Visibility beats novelty. A smaller, visible feature outperforms a larger, hidden one.
2. Context wins. Show the right doorway at the right moment in the right place.
3. Outcomes over labels. Name features by what they help me achieve, not what they are.
4. Progressive, not explosive. Teach in layers; don’t dump everything on day one.
5. Measure learning, not just clicks. Adoption = found → tried → repeated → relied on.

⸻

The discovery funnel (define success)

1. See – % of active users who see an entry point at least once.
2. Try – % who use it within the intended flow in ≤ 2 sessions of first seeing it.
3. Repeat – % who use it again within 14 days.
4. Rely – % who use it ≥ N times in 28 days and complete the adjacent job.
5. Set targets per feature (e.g., See 60% → Try 30% → Repeat 15% → Rely 7%). Report weekly.

⸻

45‑minute discovery audit (score 0–3)

Inventory

• List the feature, job‑to‑be‑done, primary entry points, required pre‑conditions.

Heuristics (score each 0–3):

• Entry density – Is there a clear entry point in the natural flow? (0 none, 3 obvious and local)
• Click distance – First‑time access ≤ 3 clicks from the start of the job?
• Label clarity – Describes the outcome (not a codename/icon only).
• Timing – Prompt appears at the moment of need (not on generic login).
• Memory aids – Empty states, small tips, and recent activity reinforce recall.
• Recovery – Can a user back out safely and try again?
• Docs‑in‑product – Micro‑help is local; no tab‑hopping required.
• Searchability – Global search / command palette returns the feature.

Decision: Any score ≤ 12 needs redesign for discovery before you add anything new.

⸻

Patterns that work (with copy you can lift)

A. Empty states that guide

Template: What this does → Why it matters → One action.

Example: Compare segments to spot churn risk. Reveals changes that affect revenue. Create your first comparison.

B. Contextual nudges

Use small, inline prompts tied to intent. Avoid page‑top banners.

Example micro‑nudge: Uploading customer CSV? Try Auto‑map columns to speed this import.

C. Visible, descriptive entry points

Icons need labels. Dropdowns need purpose lines.

Example nav item: Insights → Trends & anomalies for your last 30 days.

D. Progressive discovery

Gate advanced controls behind a plain “Advanced” link with a one‑line promise and a preview.

E. Command palette & search

Index features and common tasks: “Create cohort”, “Schedule report”, “Bulk update prices.” Ensure synonyms resolve.

F. Smart defaults + ‘learn more’

Start with an opinionated default, show a 1‑line rationale, offer a small Learn more link (kept in‑app).

⸻

Launch & adoption checklist

Before launch

1. Define funnel targets (See/Try/Repeat/Rely).
2. Instrument events (below).
3. Write empty states, tips, and search entries.
4. Place at least two entry points: in‑flow + command/search.

Week 0 (launch)

1. Add a single tasteful announcement (changelog or toast) pointing to a real task.
2. Ship contextual prompts on the top related flow.

Weeks 1–2

1. Review funnels by segment (new vs existing, plan tiers).
2. Tweak labels/placement based on heatmaps and session replays.
3. Remove any low‑value prompt that creates banner blindness.

Weeks 3–4

1. Add a second‑pass cue for non‑adopters (in‑flow only).
2. Publish a 90‑second help video embedded in the empty state.

⸻

Anti‑patterns (and what to do instead)

1. Three‑dot burial → Put the action in the surface where the job starts; label it.
2. Launch banner for everything → Swap for a nudge triggered by intent (e.g., after upload).
3. Icon‑only nav → Add outcome labels and purpose lines.
4. Onboarding tours that teach everything → First‑day essentials only; reveal the rest when relevant.
5. “We tweeted about it” → Treat social as awareness, not instruction.

⸻

Stop building features no one can find. Delivery without discovery is busywork.

One ruthless test for the wall: If a first‑time user can’t reach it in three clicks or find it via search, it isn’t shipped.

You probably don’t need more features. You need fewer hiding places. Go un‑hide your product; the growth you want is already in there.