Prompt
How do I recommend features inside a SaaS product?
Latest observation
A good in-product feature recommendation system usually does three things well:
- Understands the user
- Suggests the right feature at the right time
- Shows clear value with minimal friction
Here’s a practical way to do it.
1) Start with the user’s context
Don’t recommend features generically. Base recommendations on signals like:
- Role/persona: admin, manager, analyst, creator, etc.
- Plan/tier: free, pro, enterprise
- Lifecycle stage: new user, power user, at-risk user
- Behavior: features used, features ignored, workflows completed
- Intent: what they’re trying to do right now
- Company context: team size, industry, maturity
Example:
If a user exports reports frequently, recommend automated scheduling or dashboard sharing.
2) Tie recommendations to a specific moment
Feature recommendations work best when they appear during a relevant workflow, not as random popups.
Good moments:
- After completing a related task
- When the user hits a limitation
- When usage patterns suggest an upgrade in capability
- In empty states
- In dashboards or side panels
- In onboarding or post-onboarding flows
Example:
After a user manually repeats the same action three times, recommend an automation feature.
3) Make the recommendation clearly useful
A recommendation should answer:
- What is it?
- Why should I care?
- What do I get?
- What do I do next?
A strong pattern is:
- Short headline
- One-line benefit
- Clear CTA
Example copy:
- Automate this report
- Save 3 hours a week by scheduling recurring exports.
- Set up automation
4) Use personalization, but keep it explainable
Users are more likely to accept recommendations if they feel relevant and transparent.
Good:
- “Recommended because you’ve exported this report 5 times this month”
- “Suggested for teams with 10+ users”
Avoid:
- Vague “You may also like”
- Recommendations that feel creepy or overly inferred
5) Rank recommendations by value, not just availability
Not every feature should be recommended equally. Score each recommendation based on:
- Relevance
- Likelihood of adoption
- Business value
- User benefit
- Timing
- Risk of annoyance
You can think of a simple score like:
Recommendation score = relevance + impact + timing - annoyance
Only surface the top 1–3 recommendations at a time.
6) Don’t overwhelm the user
Feature recommendations can hurt UX if they become noisy.
Best practices:
- Limit frequency
- Avoid showing too many at once
- Let users dismiss or snooze
- Don’t repeat dismissed recommendations too aggressively
- Use progressive disclosure
7) Match the format to the recommendation
Different formats work for different kinds of suggestions:
- Tooltip: quick, contextual nudge
- Banner: important but non-blocking
- Empty state card: great for discovery
- Side panel: persistent recommendations
- Modal: only for high-value, rare moments
- Checklist/onboarding: good for guided adoption
- Email/in-app message: for delayed or off-flow recommendations
8) Make the next step easy
A feature recommendation should lead directly to action.
Examples:
- “Try it now”
- “Enable this”
- “See demo”
- “Learn more”
- “Start setup”
If setup is required, reduce friction with:
- prefilled settings
- one-click activation
- guided setup flow
- sample data or preview
9) Measure effectiveness
Track whether recommendations actually help.
Useful metrics:
- Impression-to-click rate
- Click-to-activation rate
- Feature adoption rate
- Retention impact
- Task completion improvement
- Revenue uplift
- Dismissal rate
- Repeat exposure fatigue
Run A/B tests to compare:
- recommendation timing
- copy
- placement
- personalization logic
- number of recommendations shown
10) Common recommendation types in SaaS
Some common use cases:
- Next best action
- Cross-sell related feature
- Upsell to higher plan
- Workflow completion suggestion
- Automation opportunity
- Feature discovery
- Best practice recommendation
- Integration suggestion
A simple framework you can use
For each feature recommendation, define:
- Trigger — when it appears
- Audience — who sees it
- Logic — why it is shown
- Message — what it says
- Action — what user can do
- Success metric — how you’ll know it worked
Example
Scenario: User manually creates weekly reports.
- Trigger: After the 3rd report creation in 7 days
- Audience: Managers on paid plans
- Logic: Repeated manual behavior indicates automation value
- Message: “Save time by scheduling this report weekly”
- Action: “Set up schedule”
- Metric: % of users who enable scheduled reports
If you want, I can also give you:
- a feature recommendation product strategy,
- UI pattern examples, or
- a simple recommendation engine design for SaaS.