Research Foundry · Prototype
Driver Intelligence Stack

Stop ranking drivers.
Start pulling levers.

Traditional driver analysis tells you what correlates with your outcome. It doesn't tell you which levers will actually move it — for which customers, under what conditions, and at what threshold urgency becomes critical.

Non-linear threshold detection
Segment-specific driver profiles
Wave stability tracking
Decision-ready output
Traditional analysis
Driver Intelligence Stack
DRIVER IMPORTANCE — CORRELATION COEFFICIENTS
Three problems with this

Everything looks sort of important. Which one do you fix first? The model doesn't say. The boardroom argues. Nothing changes.

High-level constructs are mixed with tangible specifics. "Value for money" is a broad attitude. "Digital ease" is an operational metric. Treating them the same hides completely different interventions.

Relationships between drivers are invisible. Regression assumes independence. In reality, low scores on two drivers together are catastrophically worse than either alone.

Layer 1 — Foundation
Honest baseline → exactly what's driving outcomes and by how much
Layer 2 — Non-linear
Tipping point detection → the danger zones standard analysis never sees
Layer 3 — Strategic
Strategic themes → Pricing Fairness, Relational Strength, Digital Friction
Layer 4 — Temporal
Signal Watch → which drivers are rising, fading, or heading toward crisis
Output: not a ranking — a decision system. Which lever, for which segment, with what urgency.
The architecture
Four layers. One intelligence system.

Each layer builds on the last — progressively revealing what your data already knows, but standard analysis can't see.

LAYER 01
📐
Foundation
Honest baseline intelligence
Cuts through the noise to tell you what's actually driving outcomes — and exactly how much each driver is contributing for any individual customer. No black boxes. No guesswork.
Unlocks: Directional driver ranking with individual-level explainability
LAYER 02
Tipping Points
Where the danger zones live
Reveals the thresholds your standard analysis never finds — the score below which a customer is almost certain to leave, and the driver combinations that make it catastrophically worse.
Unlocks: Tipping point detection, danger zone mapping, interaction effects
LAYER 03
🗺️
Strategic Themes
From drivers to decisions
Organises individual drivers into the strategic themes that matter to your board — Pricing Fairness, Relational Strength, Digital Friction — and shows how priorities differ by customer segment.
Unlocks: Board-ready framing, segment-specific action plans
LAYER 04
📈
Signal Watch
What's rising. What's fading.
Tracks how the importance of each driver shifts wave-to-wave — flagging which signals are accelerating toward crisis, which are fading in relevance, and which are dangerously volatile.
Unlocks: Early warning, urgency ranking, trend intelligence
Interactive prototypes
Three lenses on the same data.

Australian home loan customers under cost of living pressure. Each tool reveals something the others can't.

Layer 01 — Foundation
Risk Lens™
Why is this customer at risk? Select a customer profile to see exactly how each driver contributes to their switching probability — moving from the population baseline to their individual prediction.
Layer 01 Deep dive
Risk Lens™ — Customer Risk Decomposition
Context: Australian home loan customers · Outcome: Switching intent
Customer profile:
← protects loyalty  ·  base 38%  ·  increases risk →
Base rate (population avg): 38% switching intent
This customer: 74%
Increases risk Decreases risk Bar length = strength of effect
Customer snapshot
Key insight

Layer 02 — Non-linear
Tipping Point Explorer™
Where are the tipping points? Drag the slider to move through driver score values and watch switching intent respond — including the non-linear thresholds that regression completely misses.
Layer 02 Deep dive
Tipping Point Explorer™ — Where Loyalty Breaks Down
Context: Australian home loan customers · Outcome: Switching intent
Driver
Driver score: 5.0 / 10
1 10
Predicted switching intent
62%
Tipping point crossed. Below this score, switching intent accelerates non-linearly. 847 customers currently in this danger zone.
1.02.03.04.05.06.07.08.09.010.0
PDP (population average) ICE (individual customers) Danger zone
Layer 02 — ICE: Individual Conditional Expectation
Segment Divergence Map™
The population average hides the real story. ICE curves show how individual segments respond differently to the same driver — revealing that some cohorts hit a tipping point far earlier and far more sharply than others.
Segment Divergence Map™ — Same Driver, Different Customers
Driver: Value for money · Outcome: Switching intent · Segment analysis
Show segments:
1.02.03.04.05.06.07.08.09.010.0
Owner-occupiers
Tipping point at 6.4/10. Below this, switching intent rises steeply to 80%+. The sharpest inflection of any segment — mortgage stress amplifies price sensitivity non-linearly.
First home buyers
Relatively flat response to value perception — less price-sensitive. Their tipping point is proactive communication, not price. Treating them the same as owner-occupiers wastes intervention budget.
Layer 02 — Interaction effects
Force Field Analysis™
The most powerful insight the Driver Intelligence Stack produces. Two drivers at low scores together create switching risk that is catastrophically worse than either alone — and one high score can act as a buffer that protects loyalty even when another driver fails.
Layer 02 Deep dive
Interaction A — The danger combination
Low value + Low proactive comms → near-certain churn
SWITCHING INTENT BY COMBINATION
The finding

Value perception below threshold: switching intent 54%. Proactive comms below threshold alone: switching intent 48%. Both below threshold simultaneously: switching intent 81%. The combination is not additive — it's multiplicative. Regression completely misses this.

Interaction B — The buffer effect
High proactive comms buffers against declining value perception
SWITCHING INTENT: VALUE DECLINING, COMMS HIGH VS LOW
The finding

As value perception declines from 8.0 to 5.0, customers with high proactive comms scores show 32 percentage points lower switching intent than those with low comms scores. Proactive communication is the primary buffer against value-driven churn. Fix comms first — it buys time on value.

Layer 03 — Strategic
Signal Compass™
Individual drivers cluster into strategic themes. Hover a cluster to see which drivers belong, their combined importance, and the strategic action they point to.
Layer 03 Deep dive
Signal Compass™ — Strategic Theme Mapping
Intelligent clustering · 4 strategic themes identified · Click a theme to explore
Pricing Fairness
Relational Strength
Digital Friction
Service Quality
Click a cluster theme
Select a strategic theme from the legend above to see which drivers cluster together, their combined importance, and the recommended action.
Layer 03 continued
Same drivers. Different segments. Different story.

The stack produces segment-specific driver profiles. What matters most to an owner-occupier under mortgage stress is not what matters most to a first home buyer.

Owner-occupier
Refinancer
First home buyer
Investor
Driver importance ranking
Segment intelligence
Value perception is the threshold driver for this segment
Owner-occupiers under mortgage stress show non-linear switching behaviour around value perception. Below 6.4, switching intent escalates regardless of other driver performance. Proactive communication acts as a buffer — but only while value perception stays above threshold.
Driver importance ranking
Segment intelligence
Speed of resolution dominates for refinancers
Refinancers have already made an active decision to switch once. Their tolerance for friction is near zero. Speed of resolution and digital ease are the primary retention levers — not relationship or value. They need the process to work, not the relationship to feel warm.
Driver importance ranking
Segment intelligence
Proactive communication is uniquely protective here
First home buyers are in unfamiliar territory. They interpret silence as neglect. Proactive communication — explaining decisions, flagging upcoming rate changes, checking in — has twice the loyalty impact for this segment than any other. The relationship is being formed right now.
Driver importance ranking
Segment intelligence
Transparency on fees drives loyalty for investors
Investors manage lending as a business decision. Unexpected charges trigger disproportionate switching intent — not because of the amount, but because opacity signals poor partnership. Transparency on fees and rate clarity are the dominant retention levers for this commercially-minded segment.
Layer 04 — Temporal
Which drivers are on the move?

Driver importance isn't static. Under cost of living pressure, some drivers are rising in urgency while others are losing relevance. Stability tracking shows you where to watch.

Momentum Tracker™ — 8 Wave Signal Intelligence
Australian home loan customers · Waves 5–12
W5W6W7W8W9W10W11W12
Rising urgency ↑
Value for money
Importance up 18pts across 8 waves under cost of living pressure. Now the #1 driver where it was #3.
Volatile — watch closely ⚡
Proactive communication
High variance wave to wave. Strong when present; catastrophic when absent. Delivery consistency is the issue.
Declining relevance ↓
Branch / service access
Structural decline across all segments as digital adoption increases. Still matters for 65+ segment.
The combined output
Where complexity becomes
decision intelligence.

Each layer adds a dimension. Together they produce something none could generate alone — a specific, defensible, board-ready action.

1
Layer 01 — SHAP
Identify the at-risk cohort
SHAP decomposition identifies that value perception and proactive communication are the two dominant contributors to switching risk — but their interaction is what matters, not either in isolation.
2
Layer 02 — ICE/PDP
Find the threshold
ICE curves reveal a non-linear tipping point: below 6.4 on value perception, switching intent accelerates regardless of other driver performance. This threshold is invisible to regression.
3
Layer 03 — Clustering
Map to strategic theme
Value perception and fee transparency cluster together under Pricing Fairness. The intervention isn't a price cut — it's a communications strategy that reframes value before the threshold is crossed.
4
Layer 04 — Stability
Quantify the urgency
Stability tracking shows value perception importance has risen 18 points in 8 waves. This isn't a stable driver to monitor — it's an accelerating risk that demands action this quarter, not next year.
Decision Intelligence Output
Board-ready
SHAP finding

Of the 4,200 owner-occupier home loan customers analysed, 847 show a SHAP switching probability above 60%. The top two contributors in 91% of high-risk cases: value perception (low) and proactive communication (absent).

Traditional analysis: "Value for money is the top driver." No action specified.
ICE/PDP finding

Non-linear threshold identified at value perception score 6.4/10. Above this: switching intent is manageable (32–48%). Below it: switching intent escalates to 68–84% regardless of other driver performance. The threshold is the intervention point.

312 customers are currently below threshold. Each 0.1pt drop adds approximately 23 customers to the danger zone.
Cluster finding

Value perception and fee transparency cluster tightly under Pricing Fairness — and this cluster's combined SHAP importance has risen to 0.41 (vs 0.23 for Relational Strength). The intervention isn't a rate cut. It's a proactive value narrative that reframes what customers already pay.

Integrated decision

312 owner-occupier mortgage holders are below the 6.4 value perception threshold, of which 218 also score low on proactive communication. This is the highest-leverage cohort. Proactive outreach to this specific group — explaining value, not defending price — is projected to retain 68–74% of at-risk customers. Value perception importance is rising 18pts over 8 waves. The window is this quarter.

This recommendation couldn't exist without all four layers working together.
SiMon™ packages this output into a real-time decision system — connecting driver intelligence to live customer cohorts.
Explore SiMon™ →
The difference
Traditional analysis vs Driver Intelligence Stack.

Same data. Completely different decisions.

Traditional driver analysis
📊
Produces a ranked bar chart. Value for money: most important. Digital ease: second. Proactive communication: third.
🤷
Every driver looks roughly equally important. The differences are small. The model doesn't say which to fix first.
🏛️
The boardroom argues about which finding to prioritise. Nothing gets decided. The insight dies in the room.
📅
Six months later: same drivers, same discussion. The ranking hasn't changed. Nobody knows if the actions worked.
The report answers the question that was asked, not the question that needed answering.
Driver Intelligence Stack
🎯
Identifies 312 customers below the 6.4 threshold — the specific cohort where value perception has crossed into non-linear switching territory.
Shows that proactive communication acts as a buffer — and that 218 of the 312 at-risk customers are missing both. That's the intersection to act on.
📈
Value perception importance has risen 18pts in 8 waves. Stability tracking shows this is accelerating — not a stable signal to monitor later.
Board decision: proactive outreach to 218 customers this quarter. Retention projected at 68–74%. ROI calculable. Accountability clear.
The stack answers the question that matters: which lever, for which customers, with what urgency, and at what ROI.
Get early access
See it on your data.

The Driver Intelligence Stack is in active development. If you run driver analysis and want to understand what the ML layer reveals that regression misses — let's talk.

  • Start the conversation → ← Research Foundry