sidchaudhary/gtm-skills / skills/brand-designer/onboarding-video
onboarding-video
UnverifiedProduces a short conversion-focused onboarding video in Remotion for an iOS app, Android app or website, embedded into an onboarding flow to move New Customers toward their activation moment, with each beat mapped to a conversion milestone that can be tracked. Use when asked to create, build or generate an onboarding video, app preview, activation demo, or any short video that demonstrates a product feature to drive activation. Boundary: `product-photography` directs still photography and `creative-brief` produces the messaging brief. This skill renders an actual video.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 29.8 KB |
| references/agent-security.md | markdown | 7.2 KB |
| references/funnel-benchmarks.md | markdown | 26.2 KB |
| references/house-rules.md | markdown | 13.3 KB |
| references/outbound-copy-standards.md | markdown | 15.5 KB |
| references/personalization-rules.md | markdown | 30.3 KB |
references/personalization-rules.md
View on GitHub# Personalization Rule Design
Reference for designing web and content personalization: types, rule structure, condition types, experience priorities, page zones, recommendation algorithms, measurement, and progressive personalization.
---
## Personalization Types
### Content Personalization
Modifying copy and images based on visitor attributes.
| Element | Example |
|---------|---------|
| Headline text | "The CRM built for [visitor's industry]" vs. generic headline |
| Body copy | Industry-specific pain points and solutions |
| Images | Role-appropriate hero images (marketer sees marketing dashboard, engineer sees API docs) |
| Case studies | Show case studies from the visitor's industry |
| Statistics | Display relevant benchmarks for the visitor's company size |
### Layout Personalization
Changing the order or visibility of page sections.
| Element | Example |
|---------|---------|
| Section order | Pricing section first for high-intent visitors; education first for new visitors |
| Section visibility | Hide "What is X?" section for returning visitors who already understand |
| Content density | Detailed view for technical evaluators; summary view for executives |
| Feature emphasis | Highlight API docs for developer personas; highlight dashboards for business users |
### Offer Personalization
Adjusting pricing, discounts, or promotional offers.
| Element | Example |
|---------|---------|
| Discount level | 20% for win-back segments, 10% for new customers |
| Plan recommendation | Highlight Enterprise for companies >500 employees, Starter for <50 |
| Trial length | 14-day standard, 30-day for enterprise evaluations |
| Bundle offers | Suggest add-ons based on current plan and usage patterns |
### Navigation Personalization
Modifying menu items and navigation paths.
| Element | Example |
|---------|---------|
| Menu items | Show "Upgrade" for free users, "Renew" for expiring contracts |
| Quick links | Surface most-used features in navigation |
| Onboarding flow | Skip completed steps, highlight next action |
| Documentation | Prioritize docs for features the user actually uses |
### CTA Personalization
Changing button text, color, and destination based on visitor state.
| Element | Example |
|---------|---------|
| Button text | "Start Free Trial" (new) vs. "Upgrade Plan" (free user) vs. "Talk to Sales" (enterprise) |
| Button destination | Link to appropriate signup flow, upgrade page, or demo booking |
| Urgency | Add countdown for time-limited offers to eligible segments |
| Secondary CTA | "Watch Demo" for new visitors, "See What's New" for returning users |
---
## Rule Structure
Every personalization rule follows the IF-THEN-ELSE pattern:
```
IF [segment condition] THEN [show experience variant] ELSE [default experience]
```
### Rule Schema
```json
{
"id": "rule-001",
"name": "Enterprise Hero Personalization",
"priority": 1,
"enabled": true,
"condition": {
"$and": [
{ "attribute": "company_size", "operator": ">=", "value": 500 },
{ "attribute": "lifecycle_stage", "operator": "in", "value": ["Promising", "Regulars"] }
]
},
"experience": {
"zone": "hero",
"variant": "enterprise-hero-v1",
"content": {
"headline": "Enterprise-grade engagement at scale",
"subheadline": "Trusted by Fortune 500 teams to manage millions of customer interactions.",
"cta_text": "Talk to Sales",
"cta_url": "/demo/enterprise",
"hero_image": "/images/enterprise-dashboard.webp"
}
},
"default": {
"variant": "default-hero",
"content": {
"headline": "The complete customer engagement platform",
"subheadline": "Grow revenue with personalized experiences across every channel.",
"cta_text": "Start Free Trial",
"cta_url": "/signup",
"hero_image": "/images/standard-dashboard.webp"
}
}
}
```
---
## Condition Types
### Lifecycle Stage
| Condition | Example Use |
|-----------|-------------|
| `lifecycle_stage = "Champions"` | VIP messaging, exclusive offers, referral prompts |
| `lifecycle_stage = "New Customers"` | Onboarding-focused content, getting started guides |
| `lifecycle_stage = "At Risk"` | Win-back offers, re-engagement messaging |
| `lifecycle_stage in ["Regulars", "Champions"]` | Loyalty rewards, advanced feature highlights |
### Behavioral Score
| Condition | Example Use |
|-----------|-------------|
| `activity_score > 80` | Power user content, advanced features, beta invitations |
| `activity_score < 20` | Re-activation prompts, simplified onboarding |
| `fit_score > 70` | Sales-ready messaging, demo CTAs |
| `intent_level = "High"` | Pricing-focused content, urgency messaging, direct CTAs |
### Referral Source
| Condition | Example Use |
|-----------|-------------|
| `utm_source = "google"` | Headline matching the search query intent |
| `utm_source = "linkedin"` | Professional tone, B2B-specific messaging |
| `utm_campaign = "competitor-x"` | Comparison content, switching incentives |
| `referrer contains "techcrunch"` | Press mention acknowledgment, credibility reinforcement |
### Device
| Condition | Example Use |
|-----------|-------------|
| `device = "mobile"` | Simplified layout, click-to-call CTAs, app download prompts |
| `device = "desktop"` | Full-featured demos, detailed comparisons |
| `device = "tablet"` | Medium-density content, touch-optimized interactions |
### Location
| Condition | Example Use |
|-----------|-------------|
| `country = "US"` | USD pricing, US-based case studies |
| `country = "DE"` | GDPR-focused messaging, EU data center mention, EUR pricing |
| `region = "California"` | CCPA compliance messaging |
| `timezone offset in [-8, -5]` | Americas-focused event times |
### Visit Count
| Condition | Example Use |
|-----------|-------------|
| `visit_count = 1` | First-time visitor welcome, brand introduction |
| `visit_count = 2-3` | Return visitor recognition, deeper content |
| `visit_count > 5` | Familiar visitor shortcuts, "ready to commit?" messaging |
| `visit_count > 10` | Direct conversion CTAs, reduce educational content |
### Time of Day
| Condition | Example Use |
|-----------|-------------|
| `hour >= 9 AND hour <= 17` | Business-hours messaging, live chat availability |
| `hour >= 18 OR hour <= 8` | After-hours messaging, async contact options |
| `day_of_week in ["Saturday", "Sunday"]` | Weekend-appropriate content, relaxed tone |
---
## Experience Priorities
Rules are evaluated **top-to-bottom by priority number** (lower number = higher priority). The first matching rule wins.
### Priority Ordering Best Practices
| Priority Range | Rule Type | Example |
|---------------|-----------|---------|
| 1-10 | Critical overrides | Maintenance announcements, compliance notices |
| 11-30 | High-value segments | Enterprise personalization, champion VIP experience |
| 31-50 | Behavioral targeting | High-intent visitors, returning visitors |
| 51-70 | Source-based targeting | Campaign-specific landing experiences |
| 71-90 | Demographic targeting | Industry or location-based content |
| 91-100 | Default / fallback | Catch-all experience for unmatched visitors |
### Conflict Resolution
- First matching rule wins (no stacking, no blending)
- A default fallback is **required** for every zone, visitors must always see something
- If no rules match for a zone, the default experience renders
- Rules can be toggled on/off without deletion (enabled/disabled flag)
---
## Page Zones
Zones are predefined areas on a page where personalized content can be injected.
| Zone | Location | Content Types | Size Constraints |
|------|----------|---------------|-----------------|
| `hero` | Top of page, above fold | Headline, subheadline, image, CTA | Full width, 400-600px height |
| `banner` | Top bar, above header | Text + CTA link | Full width, 40-60px height |
| `sidebar` | Right or left column | Content recommendations, CTAs, forms | 250-350px wide |
| `popup` | Overlay, centered | Form, offer, announcement | 400-600px wide, auto height |
| `inline_cta` | Within body content | CTA button/banner between content blocks | Content width, 80-120px height |
| `exit_intent` | Overlay, triggered on exit intent | Last-chance offer, email capture | 500-700px wide |
| `sticky_bar` | Fixed at top or bottom of viewport | Persistent offer or announcement | Full width, 50-70px height |
| `slide_in` | Slides from corner | Secondary offer, chat prompt | 300-400px wide |
| `content_block` | Replaces a content section | Full content personalization | Section dimensions |
| `recommendation` | Below content or sidebar | Product/content recommendations | Variable |
### Zone Triggers
| Trigger | When It Fires |
|---------|--------------|
| Page load | Immediately on page render |
| Scroll depth | When user scrolls past a percentage (25%, 50%, 75%) |
| Time on page | After N seconds (typically 5-30 seconds) |
| Exit intent | Mouse moves toward browser close/back (desktop only) |
| Click | User clicks a specific element |
| Inactivity | No interaction for N seconds |
| Page count | After viewing N pages in session |
---
## Recommendation Algorithms
### Collaborative Filtering
**Method:** "Users who engaged with X also engaged with Y."
| Variant | Description | Best For |
|---------|-------------|----------|
| User-based | Find similar users, recommend what they consumed | Content recommendations |
| Item-based | Find similar items based on co-consumption patterns | Product recommendations |
**Strengths:** Discovers unexpected connections, works without content metadata.
**Weaknesses:** Cold start problem (needs interaction history), popularity bias.
### Content-Based
**Method:** Recommend items similar to what the user has already engaged with, based on item attributes.
| Signal | Description |
|--------|-------------|
| Category match | Same product category, content topic, or feature area |
| Attribute similarity | Similar price range, specifications, or tags |
| Text similarity | TF-IDF or embedding-based similarity of descriptions |
**Strengths:** No cold start for items (only needs metadata), transparent recommendations.
**Weaknesses:** Limited discovery (stays in the user's known preferences).
### Hybrid
Combines collaborative filtering and content-based approaches with configurable weights.
```
score = (weight_CF * collaborative_score) + (weight_CB * content_based_score) + (weight_popularity * popularity_score)
```
Default weights: CF 50%, CB 30%, Popularity 20%.
### Trending
**Method:** Surface items gaining traction across all users within a time window.
| Window | Use Case |
|--------|----------|
| Last 24 hours | "Trending today" |
| Last 7 days | "Popular this week" |
| Last 30 days | "Top this month" |
**Calculation:** Velocity of engagement (rate of change), not absolute volume.
### Recently Viewed
**Method:** Show items the user recently interacted with for easy re-access.
- Ordered by recency (most recent first)
- Deduplicated
- Exclude already-purchased items
- Typically limited to 4-8 items
---
## Personalization Measurement
### Lift Calculation
```
Lift = (variant conversion rate - default conversion rate) / default conversion rate × 100%
```
### Metrics to Track Per Experience
| Metric | Description |
|--------|-------------|
| Impression count | How many visitors saw this experience |
| Engagement rate | Clicks / impressions |
| Conversion rate | Conversions / impressions |
| Revenue per visitor | Total revenue / impressions |
| Time on page | Average time spent by visitors seeing this experience |
| Bounce rate | % who left without interaction |
### Segment-Level Performance
Break down performance by the segment that triggered the rule:
- Does the personalized experience actually outperform the default for this segment?
- Are there sub-segments where it underperforms?
- Is the sample size sufficient for confidence (minimum 200 impressions per variant)?
### Statistical Significance
Use the same Bayesian methodology as experiment testing:
- P(variant > default) > 95% = statistically significant lift
- Monitor for at least 14 days to account for weekly patterns
- Account for novelty effect: performance may decline after initial exposure
---
## Personalization Lift, Directional Ranges
How much does personalization actually move the needle? Vendors and research firms publish very different numbers depending on methodology, industry mix, and how "personalization" is defined, so none of the figures below are an authoritative benchmark. Use these as directional planning ranges, not numbers to promise a customer, and treat any single-study stat you see elsewhere the same way unless you've verified the source yourself.
| Personalization Type | Directional Lift | Metric |
|---------------------|------|--------|
| Any personalization vs. generic | Roughly 5-15% | Revenue per session |
| Personalized homepage vs. generic | Roughly 10-15% | Conversion rate |
| Segmented email vs. unsegmented | Meaningfully higher, commonly cited in the 2-4x range | Revenue per recipient |
| Behavioral triggers vs. time-based sends | Consistently and substantially higher, exact multiple varies too widely to quote as fact | Conversion rate |
| Product recommendations (collaborative filter) | Roughly 10-35% | AOV / revenue per session |
| Personalized CTAs vs. generic CTAs | Meaningfully higher, wide range across sources | Click-to-convert |
| Loyalty-tier personalized messaging | Roughly 2-4x | Repeat purchase rate |
| Post-purchase upsell (1:1 product recs) | Roughly 15-30% | Add-on attach rate |
| ICP-matched personalization (B2B SaaS) | Directionally higher, no stable published figure | Demo request rate |
### Personalization Failure Points
Where personalization goes wrong in practice:
| Failure | Frequency | Cause | Fix |
|---------|-----------|-------|-----|
| Personalization based on stale data | 45% of cases | No data freshness rules | Require recency constraint on all conditions |
| Wrong persona segment (mis-attribution) | 30% of cases | Relying on job title alone | Use behavioral signals + title together |
| Showing the same personalization > 10x | 40% of cases | No impression cap or refresh | Cap impressions and rotate variants |
| Personalization that doesn't match expectation | 25% of cases | Cookie data vs. current session | Blend persistent + session-level signals |
| No holdout group → can't prove lift | 60% of companies | Never set up control | Always hold 10% as unexposed control group |
## Progressive Personalization
Personalization depth increases as you learn more about the visitor.
### Level 1: Anonymous Visitor
Available signals: referral source, device, location, time of day, UTM parameters, landing page.
| Signal | Personalization |
|--------|----------------|
| `utm_source = "linkedin"` | Professional tone, B2B messaging |
| `device = "mobile"` | Simplified layout, app download CTA |
| `country = "UK"` | GBP pricing, UK case studies |
| `referrer contains "vs-competitor"` | Comparison-focused content |
| `landing_page = "/pricing"` | High-intent experience, demo CTA |
### Level 2: Known Visitor (Identified)
Additional signals: lifecycle stage, activity score, fit score, intent level, company, role.
| Signal | Personalization |
|--------|----------------|
| `lifecycle_stage = "Promising"` | Nurture content, success stories |
| `activity_score > 70` | Advanced feature highlights, power user content |
| `intent_level = "High"` | Direct pricing, demo booking, urgency |
| `company_size > 500` | Enterprise-focused messaging |
| `role = "Engineer"` | Technical content, API documentation |
### Level 3: Deep Personalization (Rich Profile)
Additional signals: purchase history, feature usage, support history, preferences, NPS score.
| Signal | Personalization |
|--------|----------------|
| `purchased_products includes "Pro"` | Upsell to Enterprise, complementary features |
| `feature_usage["analytics"] > 80%` | Advanced analytics tips, new analytics features |
| `nps_score >= 9` | Referral program, advocacy opportunities |
| `support_tickets > 3` | Proactive help, dedicated support CTA |
| `contract_renewal < 30d` | Renewal messaging, loyalty offer |
### Personalization Maturity Path
```
Stage 1: Segment-based → 3-5 audience segments, rule-based targeting
Stage 2: Behavioral → Activity and intent-based dynamic targeting
Stage 3: Predictive → ML-based next-best-action recommendations
Stage 4: Real-time → In-session behavioral adaptation
```
---
## Lifecycle Copy Patterns by Segment
How to write to each customer segment - specific copy patterns that work, based on general lifecycle marketing practice rather than any single named study.
---
### Champions Segment (High RFM, Top 20% of Customers)
**What doesn't work:** Generic loyalty rewards (10% off emails wasted on customers already buying, just burns margin).
**What works:** Exclusivity and early access. These customers want to feel like insiders, not bargain hunters.
Copy tone: Peer-to-peer, not brand-to-customer. "You've been with us long enough that we want to show you this before anyone else."
**Common patterns (this approach typically drives meaningfully more revenue than promotional campaigns sent to the same segment, though no single benchmark confirms an exact multiplier):**
| Copy Pattern | Example |
|-------------|---------|
| Early access before public | "Before we announce this publicly, [first name]..." |
| Behind-the-scenes | "Here's what we're working on next, you're one of the first to know." |
| Direct ask for feedback | "We want your opinion on [new product] before we launch it." |
| VIP event invitation | "You're invited to [exclusive event/session]", invitation-only framing |
**CTA:** "See it first" not "Shop now." The difference: one signals reward for loyalty, the other signals a push.
---
### At-Risk Segment (High Historical Value, Declining Recency)
**This is the highest-ROI personalization target.** These customers have proven LTV and are slipping away. Early intervention around the 90-day lapse mark tends to convert meaningfully better than waiting until a customer is fully lapsed and relying on a discount to win them back.
**What doesn't work:** "We miss you" generic, these customers don't respond to sentiment.
**What works:** Acknowledging the change without making it awkward + introducing what's new.
Copy structure:
```
1. Acknowledge the gap briefly: "It's been a while, things have changed since you were last here"
2. Introduce what's new/changed (new products, features, collections)
3. Soft offer (not your largest discount, that's for confirmed-churned)
4. Easy re-entry CTA
```
**Key finding:** At-risk customers who receive "here's what's new" emails (no discount) tend to convert better than fully lapsed customers targeted later with a win-back discount. Early intervention beats late discounting.
---
### New Customer Segment (First Purchase in Last 30 Days)
**Goal:** Drive second purchase. The second purchase is the LTV inflection point - customers who make a second purchase within 30 days tend to be meaningfully more likely to become long-term repeat buyers, though the exact multiple varies by category and isn't backed by one authoritative source.
**What works:** Product education + cross-category bridge.
```
Subject: "Getting the most out of [product they bought]"
Body:
- Genuine usage tip or care instruction for what they bought (builds trust, justifies the email)
- Cross-sell: 1-2 products positioned as "what people pair with [their purchase]"
- Framing: "what people pair with X" NOT "you might also like Y" (more specific, more trusted)
```
**Cross-sell specificity rule:** If they bought running shoes, suggest running socks or an insole. If they bought a face moisturizer, suggest SPF or a serum. Logical adjacency beats algorithmic recommendation at <$10M revenue, manual curation of top product pairings outperforms automated for smaller catalogs.
---
### Behavioral Trigger Copy Patterns
**"Viewed pricing page 3+ times", high-intent evaluation loop:**
What to say: acknowledge the decision they're trying to make, remove friction.
| CTA Type | Copy | CVR |
|----------|------|-----|
| Open conversation | "Have a question about [Product] pricing?" | Tends to convert well for demo/call bookings, though no verified benchmark exists for the exact rate |
| Value uncertainty | "What's included in [Plan name], the honest version" | Strong for comparison-stage visitors |
| Service-oriented | "Let's figure out the right plan for you" | Best for complex multi-plan products |
Email body: Short. "I noticed you've been looking at our pricing page. If you're trying to figure out which plan is right for [company type/use case], I'd be happy to walk you through it, or just send you a quick breakdown. What's your main use case?"
Generic promotional email to the same trigger typically converts in the low single digits. Conversational personalized email typically converts meaningfully higher.
**"Used feature X for the first time", positive momentum signal:**
What to say: One sentence acknowledging what they did. One sentence on what it unlocks. One CTA. Total: under 100 words. Goal is to maintain momentum, not to deliver a tutorial.
The next-step feature should be:
(a) Logically adjacent to what they just did
(b) Increases activation depth
(c) NOT behind a paywall, don't poison a success moment with an upsell
Example pattern: "You just created your first segment. Segments are the foundation of everything in [Product]. Next: send your first message to that segment." CTA: "Create a campaign."
**"Haven't logged in for 14 days", critical intervention window:**
Earlier intervention tends to outperform later intervention - the 14-day mark is a common trigger point, but treat the exact multiplier as illustrative, not a verified benchmark.
What not to say: "We haven't seen you in a while", passive-aggressive tone, clicks poorly.
What works: Barrier removal + value reminder.
| Copy Approach | Example Subject | Notes |
|--------------|----------------|-------|
| Outcome-focused | "Did [Product] solve your [problem] yet?" | Re-anchors on why they signed up. Tends to outperform feature-focused re-engagement, though the exact lift varies by product |
| Curiosity | "Quick question about [Product]" | Low-commitment, high-open-rate |
| New value | "Something new in [Product] since your last visit" | Gives reason to return not tied to their "failure" |
Body: Open with curiosity ("How's [use case] going for you?"). Remind them of one specific value they had access to but may not have used. Offer a resource or human touchpoint. Under 120 words.
---
### Personalization Failure Modes in Copy
**1. Creepy vs. Helpful, where the line is**
A useful framework: personalization feels helpful when it reduces effort. It feels creepy when it reveals surveillance the customer wasn't aware of.
| Helpful (reduces effort) | Creepy (reveals surveillance) |
|--------------------------|-------------------------------|
| "Based on your recent order, you might need [X]" | "We saw you looking at ankle boots for 12 minutes on Tuesday" |
| "You've browsed [category] a few times, here are the bestsellers" | "Based on your location near [store], you should visit tomorrow" |
| "Your account is almost full" | "We noticed you were comparison shopping" |
The rule: personalization is welcomed when customers believe you're using data they intentionally gave you. It becomes creepy when it reveals data they didn't consciously share.
Purchase-based personalization tends to outperform browse-based in brand perception - even when browse-based gets higher CTR - because purchase is "given" data.
**2. The Segment-of-One Fallacy**
Personalization has diminishing returns and can eventually hurt performance past a certain point.
Why maximum personalization fails:
- Coherence breaks down when personalizing on 10+ variables simultaneously, the customer gets something that feels random
- Sample sizes become meaningless, can't A/B test anything
- Personalization signals conflict, a customer who bought luxury, browsed sale items, and is in a mid-income zip shows contradictory signals. Weighting all equally produces a message that fits nothing.
**The practical limit:** Personalize on 2-3 variables maximum per campaign. Behavioral (recency/engagement) + one other signal. Adding a third variable rarely improves performance much, while adding significant complexity.
**3. Job Title vs. Behavior Mismatch**
The most common B2B SaaS personalization mistake: a VP of Marketing signs up. Team tags them as "executive persona" and sends ROI/strategy content. But they personally signed up to test an automation feature, they're in hands-on evaluation mode.
Behavior-triggered emails tend to significantly outperform segment-by-demographic emails on conversion to paid, though the exact multiple isn't backed by a verified benchmark.
Fix: use job title/persona to inform default content, but override immediately when behavioral data contradicts the assumption. If the VP of Marketing is clicking developer documentation, update their segment.
**4. Stale Data Personalization**
Using behavioral data older than 90 days for ecommerce, or 30 days for SaaS, actively hurts performance. It doesn't just miss, it signals that your personalization is lazy.
**Data freshness rules:**
| Data Type | Use for | Beyond this, use for suppression only |
|-----------|---------|--------------------------------------|
| Purchase history | Last 6 months | Older data signals you're not paying attention |
| Browsing/engagement | Last 30-45 days | Stale browsing preferences feel like surveillance |
| SaaS product behavior | Last 14-30 days | Old usage patterns may not reflect current goals |
| NPS/survey data | Last 6 months | Life circumstances change |
The worst stale data mistake: using a category preference from a one-time purchase occasion (bought baby products) to permanently tag someone in a "parent" segment. This pattern is a commonly cited reason for unsubscribes - "they kept sending me baby stuff years after my kid grew up."
---
### Subject Line Personalization: What the Data Shows
**First-name personalization, declining returns:**
| Context | Lift | Notes |
|---------|------|-------|
| Transactional emails (order, shipping) | 2-3% open rate lift | Already high open rate from relevance |
| Promotional emails (sale, new arrivals) | 6-8% open rate lift | Appears to be declining year over year, though the exact trend isn't backed by a single verified source |
| Re-engagement emails | 12-15% open rate lift | Still significant - these benefit most from feeling personal |
| B2B SaaS activation/re-engagement | 7-9% open rate lift | Directional estimate, not tied to a specific verified source |
**Warning: "[FIRST_NAME]" or "Hey ," in subject lines are among the highest unsubscribe triggers in the industry.** Data hygiene is a prerequisite for name personalization.
**Behavioral subject line personalization (highest-performing category):**
| Pattern | Example | Open Rate |
|---------|---------|-----------|
| Specific item count | "You left 3 items in your cart" | 18-22% (vs. 12-15% for "Something in your cart") |
| Specific product name | "The [product name] is still in your cart" | 24-28% typical range |
| Browse abandonment | "Still thinking about [specific product]?" | 14-18% |
| Social proof + scarcity | "10 people viewed [product] today" | 16-20% (limited-stock products) |
**Predictive subject lines, high reward, high risk:**
"Picked for you" framing vs. generic: tends to lift CTR noticeably, though the exact percentage isn't backed by a verified benchmark.
"Based on your last purchase" framing: lifts CTR when logically connected, but hurts CTR when irrelevant - the attribution to their specific behavior makes irrelevance more jarring than generic irrelevance.
Rule: only use predictive subject lines when your recommendation accuracy is above 70% relevance. Below that, generic editorial framing ("Our picks for this season") outperforms predictive for the same product set.
---
## What Personalization Actually Returns, and What It Requires First
Personalization does work, with reported revenue lifts around **5-15%** and bounce-rate reductions of
**20-45%** on personalized homepages for mid-market and enterprise B2B traffic. But three findings
determine whether a programme gets any of that.
### Identification comes before personalization
**Personalization without identification is guessing.** The teams reporting real lift invested in
visitor-identification infrastructure first. Without it, a "personalized" experience is a rule firing on
a signal that may not describe the visitor at all, and the failure is invisible: the page renders, the
variant is served, and nobody knows it was served to the wrong person.
So before designing variants, establish **how a visitor is identified**, what share of traffic can be
identified at all, and what the fallback is for the rest. If most traffic cannot be identified, the
honest answer is that the default experience matters more than the variants, and effort belongs there.
### Token-swapping is not personalization
Changing "we help companies" to "we help healthcare companies" with no supporting proof delivers minimal
lift. The reader notices the noun and nothing behind it changed: same case studies, same claims, same
CTA. It reads as a mail-merge, which is the web equivalent of "I noticed you're the [title] at
[company]".
A variant earns its place when the **evidence** changes with it, the case study is from their industry,
the objection addressed is the one their segment actually raises, the CTA matches how they buy. Swapping
a word while leaving the proof generic is worse than not personalising, because it advertises that
someone tried.
### Account-level beats one-to-one
The highest-return approach is **account-level**, not segment-of-one: adapting hero copy, case studies
and CTAs to the visiting company's industry or profile. One-to-one personalization costs far more to
build and maintain and does not reliably outperform it.
The approaches with real adoption are narrower than the pitch suggests: firmographic homepage changes,
intent-triggered content sequencing, account-specific landing pages, personalized chat openers, and
pricing-page adaptation. Anything beyond those is usually a maintenance liability.
### Measure per experience, never as one global number
Attributing a single figure to "personalization" is not achievable, and asking for one produces a number
nobody can defend. **Measure per experience and per segment.** A programme reporting "personalization
drove +8%" is reporting an average across variants that individually range from strongly positive to
negative, and the negative ones stay live because nothing separates them.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| reverse-engineer-anythingmorluto/reaReverse engineer native, managed, Electron/JavaScript, packaged, firmware, and browser targets with REA, and build agent environments (AgentEnv bundles) of web applications from a live instance or a capture, and of Linux desktop apps with a task from a request. Use shipped-artifact or requested runtime evidence to explain features, compare versions, decompile code, or guide a reconstruction. Skip REA for ordinary source-repository architecture analysis. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| diagnosing-superpowersobra/superpowersUse when a superpowers session went wrong and your human partner wants to know why — repeated work, ignored plans, stumbles, poor results, a skill that didn't fire, "it took too long", "why is it so expensive", "what is it doing" — or wants to build a bug report for the superpowers maintainers, for the current session or a past one identified by id or path, on any harness. | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| dispatching-parallel-agentsobra/superpowersUse when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |
| executing-plansobra/superpowersUse when executing an implementation plan in the current session as the implementer yourself — your human partner chose inline execution, or no subagent tool is available | Stars: 322k | Updated | Claude CodeCodex | Verified | Install |