KAAL · Behavioural Engagement Engine · Portfolio Showcase

The Engagement Engine

A gamification-driven customer acquisition and retention system built for a South African e-commerce fashion brand. Applies game design principles — player progression, unlockable content, personalised quest systems — to drive repeat purchase and LTV growth.

Python SQLite JavaScript Shopify API Klaviyo API Wilson Score Algorithm Netlify Cohort Analysis

2,416

Customers enrolled

Active engagement profiles

Engagement tiers

Starter → Member → Elite

Locked range products

Gated by tier progression

94%

Recommendation coverage

Customers with predicted next purchase

System overview — what this engine does The Engagement Engine applies core game design mechanics to a fashion e-commerce platform. Customers earn tier progression through purchase behaviour, unlocking exclusive products and personalised treatment at each level. A behavioural graph tracks purchase sequences to predict each customer's next most likely product — delivered as personalised recommendations via automated marketing flows. Every element maps directly to a game design primitive: tiers = player levels, locked range = unlockable content, recommendations = personalised quests, sequence analysis = player journey mapping.

Tier Progression System

Player levels · spend thresholds · unlock gates

🥉

Starter

2,067 customers · 85.6%

Entry threshold

1+ order · any value

Full public product catalogue Standard email flows Welcome sequence Abandon cart recovery

Progress to MemberTrack orders + LTV

🥈

Member

301 customers · 12.5%

Threshold

2+ orders OR R8,000+ LTV

All Starter benefits Access to Member-only edits Early access to new drops Priority customer service lane Partial Locked Range access

Progress to EliteTrack repeat depth

🥇

Elite

48 customers · 2.0%

Threshold

4+ orders OR R25,000+ LTV

All Member benefits Full Locked Range unlocked Direct Klaviyo VIP flow First access to limited releases Personalised styling recommendations Named VIP tag in Shopify

Maximum tierRetention focus

Tier distribution over time

Locked Range — Unlockable Content Gate

66 products · tier-gated access · game mechanic: unlockable reward

Design pattern borrowed from games In games, players unlock new zones, weapons, or abilities by reaching progression thresholds. The Locked Range applies the same mechanic to product access: 66 curated hero products are hidden from the public catalogue and progressively unlocked as customers advance through tiers. This creates genuine aspiration, increases perceived exclusivity, and rewards loyal customers with tangible access that new customers cannot buy.

Gate mechanic — how access is determined

🔒 Silk Noir Chemise — Elite only

🔒 The Signature Corset — Elite only

🔐 Pearl Satin Set — Member +

🔐 The Luxe Edit Collection — Member +

✓ The Vine — Public

✓ The Soft Pearl — Public

→

Elite
Full access

Member
Partial access

Starter
Public only

Locked range product breakdown

Tier gate	Products locked	% of catalogue	Avg price
Elite only	24	36%	R3,850
Member +	42	64%	R2,640
Public	—	Full catalogue	R1,890

Next Chapter — Recommendation Engine

Personalised next purchase · behavioural graph · Klaviyo delivery

Design pattern borrowed from games: personalised quest assignment In games, the quest system surfaces the most relevant next objective for each player based on their current progress and play style. This engine analyses each customer's purchase sequence, hero product affinity, and co-purchase patterns to predict their most likely next product. Recommendations are pushed directly into Klaviyo as personalised flow triggers — each customer receives a different email with their specific predicted product, not a generic promotion.

Customer A — Elite

Customer B — Member

Customer C — Starter

👑

Customer A

Elite · 7 orders · R41,200 LTV

Hero product

Pearl Collection

38% of all-time spend

Days since last order

23 days

Above average recency

Purchase sequence depth

7 transitions

Highly mapped journey

Predicted next

Silk Nightwear

Directed from Pearl → Silk path

Recommendation

Silk Noir Chemise (Elite-gated)

Prediction confidence

87%

78% of Elite customers who mirror Customer A's Pearl → Bridal → Silk path purchased the Silk Noir Chemise within 35 days. Median days-to-purchase: 18. Triggered via Klaviyo flow "Next Chapter — Elite Silk" with personalised product block.

Engagement score breakdown

/100

Elite · Top 2% of customers

Recency28/30

Purchase frequency30/30

Monetary value25/25

Sequence depth11/15

🔐

Customer B

Member · 3 orders · R12,800 LTV

Hero product

Bridal Collection

44% of all-time spend

Days since last order

51 days

Approaching churn window

Purchase sequence depth

3 transitions

Mid-journey mapping

Predicted next

Pearl Satin Set

Bridal → Pearl path (Member gate)

Recommendation

Pearl Satin Set (Member-unlocked)

Prediction confidence

71%

61% of Member customers with a Bridal entry SKU purchase the Pearl Satin Set within 60 days. Days since last order (51) is inside the median re-purchase window (58 days) — high probability window. Klaviyo flow triggered: "Next Chapter — Member Pearl" with Member-exclusive product access note.

Engagement score breakdown

/100

Member · Top 14% of customers

Recency19/30

Purchase frequency22/30

Monetary value18/25

Sequence depth12/15

🌱

Customer C

Starter · 1 order · R2,890 LTV

Hero product

The Soft Pearl

First + only purchase

Days since last order

14 days

In honeymoon window

Purchase sequence depth

1 transition

Journey just starting

Predicted next

The Vine

Top exit from Soft Pearl entry

Recommendation

The Vine (public catalogue)

Prediction confidence

58%

The Vine is the #1 sequential purchase for Soft Pearl entry customers (Wilson lower bound: 0.52). Honeymoon window (14 days post-purchase) is the optimal trigger window. Klaviyo flow: "Next Chapter — Starter · Your next piece" — includes teaser preview of Member-unlocked products as tier progression incentive.

Engagement score breakdown

/100

Starter · Early stage

Recency24/30

Purchase frequency4/30

Monetary value0/25

Sequence depth0/15

System Architecture

Data pipeline · algorithm stack · deployment

Data pipeline — 7 layers

Ingest

Pull all orders, customers, and products from Shopify Admin API. Pull email event data from Klaviyo API.

Python · REST API

Storage

Normalised SQLite database. Customer × order × product × event schema. Incremental refresh.

SQLite · Python

Tier score

RFM scoring per customer. Threshold classification into Starter / Member / Elite. Tag pushed to Shopify.

Python · SQL

Graph

Co-purchase pairs, directional sequences, Wilson lower bound confidence. Recency trend overlay (90d vs all-time).

SQLite · Statistics

Recommend

Per-customer next-purchase prediction. Entry SKU → sequence path matching. Confidence-ranked product list.

Python · SQLite

Deliver

Push personalised next-purchase predictions into Klaviyo customer profiles as custom properties. Trigger automated email flows.

Klaviyo API · Python

Surface

Live HTML dashboards (this system) built from SQLite data. Deployed to Netlify. No server required.

HTML · JS · Netlify

Core algorithm — Wilson score confidence

Product pair recommendations use the Wilson lower bound formula to rank pairs by statistical confidence, not raw co-purchase count. This prevents high-volume products from dominating recommendations regardless of actual affinity.

def wilson_lower(successes, total, z=1.96):
    """Wilson score lower bound for ranking."""
    if total == 0: return 0
    p = successes / total
    centre = p + z**2 / (2*total)
    margin = z * (p*(1-p)/total + z**2/(4*total**2))**0.5
    denom = 1 + z**2/total
    return (centre - margin) / denom

# Applied per product pair:
for a, b, buyers, total in pairs:
    score = wilson_lower(buyers, total)
    # score = minimum credible co-purchase rate
        

Engagement score formula

# Engagement Score (0-100)
# Recency:   30 pts — days since last order
# Frequency: 30 pts — order count + interval
# Monetary:  25 pts — LTV vs cohort percentile
# Sequence:  15 pts — graph depth + hero share

def engagement_score(customer):
    r = recency_score(customer.days_since_last)
    f = frequency_score(customer.order_count)
    m = monetary_score(customer.ltv, cohort_p90)
    s = sequence_score(customer.graph_depth,
                        customer.hero_share)
    return r + f + m + s  # max 100
        

Commercial Impact

Measured outcomes · repeat purchase · LTV uplift

Metric	Baseline (pre-engine)	With engagement engine	Delta
Repeat purchase rate	9.1%	14.4%	+5.3pp
Average LTV (repeat customers)	R5,820	R8,350	+43%
Email flow open rate	18.2%	34.6%	+90%
Locked range conversion (Elite)	n/a	22.4%	New mechanic
Recommendation acceptance rate	n/a	31%	New mechanic