Why this matters
Funnel analysis shows where users drop off. Product improvement recommendations translate those insights into prioritized, testable actions that grow activation, conversion, and revenue. As a Product Analyst, you will be asked to identify bottlenecks, estimate impact, and suggest the next best experiment or change.
- Real task: Diagnose a 40% drop at Checkout and propose 2–3 changes with estimates.
- Real task: Compare two onboarding variants and recommend which to scale and what to improve next.
- Real task: Prioritize a backlog using a scoring model so the team knows what to ship first.
Concept explained simply
You turn a funnel problem into a small set of focused actions that are:
- Root-cause driven (based on data patterns, not guesses)
- Prioritized (using simple scoring such as ICE or RICE)
- Testable (clear success metric and expected impact)
Mental model
- Find stage loss: Largest absolute loss is often the biggest lever.
- Segment for signal: Break down by device, source, cohort to find concentrated problems.
- Match issue to fix: Friction → UX/content fix; intent gap → value clarity; trust issue → reassurance/social proof; performance issue → speed/reliability.
- Prioritize with constraints: Highest expected impact per unit of effort wins.
Quick glossary
- Drop-off: Users who do not move to the next step.
- Conversion rate (step): Next_step / Current_step.
- ICE score: Impact × Confidence × Ease (1–10 scales).
- RICE score: (Reach × Impact × Confidence) / Effort.
A simple 6-step process
- Define the goal metric (e.g., Checkout conversion, Activation rate). Mini task: Write the exact numerator/denominator.
- Map the funnel with counts and step conversion rates.
- Locate bottlenecks via biggest absolute losses and abnormal step rates.
- Segment by device, source, geography, and new/returning users to sharpen the diagnosis.
- Propose fixes that directly target the friction type; define success metric and guardrails.
- Prioritize using ICE/RICE and turn the top 1–2 into experiments or small releases.
Example guardrails
- No increase in refund rate
- No decrease in downstream retention
- Page latency remains under target
Worked examples
Example 1: E-commerce checkout
Funnel (weekly):
- Product view: 100,000
- Add to cart: 18,000 (18%)
- Checkout start: 9,000 (50%)
- Payment success: 5,400 (60%)
Bottleneck: Product view → Add to cart (largest absolute loss). Secondary: Checkout → Payment (trust/fees).
Recommendations:
- Add-to-cart button prominence and above-the-fold placement (A/B test).
- Checkout trust boosters: clear shipping/fees early, badges, guest checkout.
Impact estimate
If Add-to-cart rises from 18% to 20%, adds 2,000 carts → 1,000 checkout starts (50%) → 600 payments (60%). +600 orders (+11% vs 5,400). For trust boosters, raising 60% to 63% yields +270 orders. Combined potential: +870 orders. Treat as rough ranges.
Prioritization example (ICE)
- Button prominence: Impact 7, Confidence 7, Ease 8 → ICE 392
- Trust boosters: Impact 5, Confidence 6, Ease 6 → ICE 180
Ship button prominence first.
Example 2: SaaS onboarding
Funnel (monthly new users):
- Sign-ups: 20,000
- Email verified: 12,000 (60%)
- Completed setup: 7,200 (60%)
- Activation (first key action): 3,600 (50%)
Bottleneck: Sign-up → Email verified (deliverability or motivation). Secondary: Setup → Activation (value clarity).
Recommendations:
- Magic link login or delayed verification until first session value.
- In-product checklist with sample data to reach first value in 2 minutes.
Impact estimate (RICE)
- Magic link: Reach 20k, Impact 0.15, Confidence 0.6, Effort 2 → RICE = (20000×0.15×0.6)/2 = 900
- Checklist: Reach 7.2k, Impact 0.1, Confidence 0.7, Effort 3 → RICE = (7200×0.1×0.7)/3 ≈ 168
Prioritize magic link.
Example 3: Mobile app notification opt-in
Funnel:
- App installs: 50,000
- Prompt shown: 45,000 (90%)
- Allow notifications: 13,500 (30%)
Bottleneck: Low allow rate.
Recommendations:
- Pre-permission screen explaining value with non-intrusive timing (after first win).
- Granular controls and preview of content frequency.
Impact estimate
Raising allow from 30% to 36% yields +2,700 permissioned users monthly. If notifications increase 7-day retention by 4pp, expect material LTV lift. Varies by country/company; treat as rough ranges.
Exercises
Do the following exercise. You can check the sample solution below and in the Exercises section.
Exercise 1 — Prioritize fixes from a funnel
Funnel (last week):
- Landing visits: 80,000
- Signup started: 16,000
- Signup completed: 9,600
- First action done: 3,840
Segments: Mobile signup completion 52%; Desktop 68%.
Task:
- Identify the main bottleneck and the most affected segment.
- Propose two improvements and estimate impact using RICE (define your own scales and effort).
- Pick one top recommendation with a clear success metric and guardrails.
Submission checklist
- Bottleneck named with numbers
- Two ideas with RICE inputs and score
- Chosen idea with metric and guardrails
Show sample solution
Main bottleneck: Signup started → Signup completed (16,000 → 9,600, 60% step conv; mobile worse at 52%).
Ideas:
- Shorten mobile form (reduce required fields; support autofill). RICE: Reach 10k (mobile signups), Impact 0.12, Confidence 0.6, Effort 2 → 360
- Progress indicator + error copy improvements. RICE: Reach 16k, Impact 0.06, Confidence 0.7, Effort 2 → 336
Pick: Shorten mobile form.
Success metric: Mobile signup completion rate from 52% to 60%. Guardrails: No drop in data quality proxy (verification pass rate), latency unchanged.
Common mistakes and self-check
- Jumping to UI fixes without quantifying which step loses the most users. Self-check: Can you show absolute losses per step?
- Using averages only. Self-check: Did you segment by device/source/cohort?
- Vague recommendations. Self-check: Is there a success metric, target, and guardrails?
- No effort estimate. Self-check: Did you apply ICE/RICE with an explicit effort value?
- Ignoring downstream effects. Self-check: Did you consider retention/revenue guardrails?
Practical projects
- Audit a signup funnel: Collect 2 weeks of step counts, compute conversion by device, propose 3 fixes, and score with ICE.
- Checkout trust sprint: Run a content test (fees clarity vs control), define guardrails (refund rate, NPS), and write a 1-page readout.
- Onboarding fast track: Design a 3-step checklist to reach first value; simulate impact with a spreadsheet using scenario assumptions.
Quick test
Anyone can take the test for free; only logged-in users have progress saved.
How to use your results
- If you score below 70%, review the worked examples and redo Exercise 1.
- At or above 70%, move to the next subskill and start a small A/B test plan.
Who this is for
- Aspiring and current Product Analysts who turn funnel insights into action
- PMs and Growth analysts needing a lightweight prioritization approach
- Designers/Engineers partnering on conversion improvements
Prerequisites
- Basic funnel metrics: counts, conversion rates, drop-off
- Comfort with segmentation (device, source)
- Ability to estimate reach/effort roughly
Learning path
- Before: Funnel mapping and diagnostics; Segmentation basics
- Now: Product Improvement Recommendations (this page)
- Next: Experiment design and A/B testing; Impact sizing and forecasting
Next steps
- Complete Exercise 1 and write a 5-sentence recommendation memo.
- Pick one live funnel in your product; compute absolute losses and top 2 ICE/RICE ideas.
- Draft a one-week test plan with success metric and guardrails.
Mini challenge
In 10 minutes, list three improvements for your worst funnel step. Assign quick ICE scores (1–10) and pick one to test first. Write one sentence defining the success metric and target.