Why is "Applying this example's order to a different category" a common mistake when using the Shopper-Occasion Pack Matrix?

Symptom: Hierarchy accuracy reads 100% but shopper behavior in the actual category (yogurt, cereal, coffee) doesn't match the model's predictions. Fix: The Type > Brand > Flavor > Size > Price order is an ILLUSTRATIVE anchor for this chocolate-tablet example, and any real order is CATEGORY SPECIFIC. Yogurt research often shows Flavor first; coffee often shows Brand close second; cereal often shows Benefit / Dietary (gluten-free, high-protein) as the first gate. Always derive the correct hierarchy from category-specific shopper research, and commission the study if you don't have it. A wrong hierarchy with a 100% accuracy score against a wrong benchmark is the most dangerous possible outcome.

Why is "Treating the intra-brand dominance as real cannibalization instead of a hierarchy artefact" a common mistake when using the Shopper-Occasion Pack Matrix?

Symptom: A portfolio decision (delist the 'cannibalizing' SKU, consolidate to one size) is made based on the default hierarchy's top-competitor readout. Fix: Before any delist / consolidate decision, run the CDT at the hierarchy your category research supports (here, the illustrative target order). If the intra-brand dominance PERSISTS at that hierarchy, the cannibalization is real and a portfolio decision is defensible. If the intra-brand dominance DISAPPEARS at that hierarchy (as happens with p1 and p6 in the default scenario: 35% down to 18%), the cannibalization was always an artefact of the wrong framing; delisting would remove a legitimate shopper choice.

Why is "Assuming the 6-product model is a complete category representation" a common mistake when using the Shopper-Occasion Pack Matrix?

Symptom: A CDT-based strategy recommendation concludes "the category has only 6 meaningful SKUs" or "Aventra Filled Fruit and Nut has no competitors." Fix: The 6 products are a teaching sample. Real chocolate-tablet assortments run to dozens of SKUs in a large store. The p5 Aventra "no competitors at 2% across the board" finding is informative (the Filled branch is under-populated in this sample) but NOT a portfolio recommendation on its own. Extend the CDT analysis to the full category assortment via scanner-data pairwise cross-elasticity estimation before recommending any branch-expansion investment.

Why is "Running PPA tools (OBPPC, Incentive Curve, Pack Roles) before the CDT" a common mistake when using the Shopper-Occasion Pack Matrix?

Symptom: PPA analysis recommends a mid-tier pack gap that, after CDT research, turns out to sit in a branch shoppers don't actually navigate to; the new SKU launches and immediately underperforms. Fix: CDT is the DEMAND-SIDE root of every PPA decision. PPA Lesson 2 Pack Roles assumes the Size/Format gate; PPA Lesson 3 Incentive Curve operates WITHIN a CDT branch; PPA Lesson 4 OBPPC assumes a CDT hierarchy per cell. Always run the CDT first, then the PPA tools. The published PPA Lesson 7 Module Arc literally says "CDT is the demand-side root of all PPA work."

Why is "Confusing 'cross-price elasticity' with 'CDT switching probability'" a common mistake when using the Shopper-Occasion Pack Matrix?

Symptom: Scanner-data cross-XED analysis shows a 0.15 cross-elasticity between p1 and p3; CDT model shows 8% switching probability. Analyst concludes "the numbers don't match" and dismisses one or both models. Fix: They're different constructs. Cross-price elasticity measures volume response to PRICE changes; CDT switching probability measures the PICK rate under a free-choice decision. They correlate (products with high cross-XED usually have high CDT switching too) but they are not the same measurement. Use cross-XED for promo-priced competitive modeling (Pricing Lesson 6); use CDT for category structure / assortment / OBPPC decisions.

Free tool · no login required

Shopper-Occasion Pack Matrix

Build the Consumer Decision Tree hierarchy for a chocolate-tablet portfolio, score it against an illustrative target hierarchy, and watch the true competitive set reshape as you reorder the five decision gates. The same interactive model the full RGM Academy course uses for PPA Lesson 7, no auth, no paywall.

Updated 23 April 2026Extracted from the Price Pack Architecture module, lesson 7: Consumer Decision Tree

Loading calculator…

Guided walkthrough

Explore the tool, from setup to common mistakes

Five short sections explain the scenario, what each control does, how to read the output, and the mistakes to avoid. Open whichever helps. The tool above works without them.

5.1Scenario setup

The starting SKU, market, and assumptions the model makes.

You are the Category Manager on a chocolate‑tablet portfolio heading into the annual PPA review with a major grocery chain. (The category, brands, and numbers in this walkthrough are illustrative.) Sales have been flat for two quarters despite three line extensions and a price‑pack redesign. Your internal framing of the category has been brand‑first (Corvane vs Brindel vs Aventra vs Store Brand), and every commercial decision has been built on that assumption.

In this scenario, shopper research has just come back with a clear finding: shoppers make the decision in a different order than your team assumed. They choose Type first (milk vs dark vs filled), then Brand, then Flavor, then Size, with Price as the final gate. Your brand‑first framing is misidentifying the competitive set: you've been thinking your Corvane Milk Hazelnut competes with your Corvane Dark Hazelnut, when the scenario's research says it competes with Brindel Milk Hazelnut and Store Brand Milk Plain.

Your job by Friday: run the Consumer Decision Tree under both hierarchies, quantify the shift in competitive set, and bring a CDT‑aligned portfolio recommendation to the review.

Your objective

Use the CDT tool to compare the brand-first vs Type-first target hierarchies, quantify the shift in the competitive set for each of the 6 focus products, and identify which SKUs are miscategorised under the current framing.

Key assumptions

The switching‑probability model is coarse by design. Five bands [2%, 8%, 18%, 35%, 55%] indexed by count of shared top‑of‑tree attributes. Real conjoint studies produce continuous switching matrices with hundreds of cells; this teaching model uses 5 tiers to make the SHAPE of the competitive set legible at a glance. The teaching point is the hierarchy‑driven reshape, not the exact switching dollar amounts.
Hierarchy accuracy is scored position‑by‑position against the illustrative target order for this chocolate‑tablet example (Type > Brand > Flavor > Size > Price). The target is a teaching anchor, not a published research finding, and any real CDT order is category‑specific: a cereal or yogurt CDT will gate in a different order, and applying this example's order to another category would score as high accuracy but be operationally wrong. Always derive category‑specific CDT from category‑specific shopper research.
"Intra‑brand" vs "cross‑brand" top‑competitor is the single most important signal on the page. An intra‑brand top competitor means your own range is the biggest threat to a given SKU: classic cannibalization, typically a signal that the hierarchy is putting Brand ahead of a more decisive gate. A cross‑brand top competitor means you are competing in the true category; whether that is good depends on whether the current hierarchy matches research.
Default hierarchy is deliberately wrong for teaching. Most commercial teams arrive at the CDT tool already assuming Brand‑first; the default [Brand > Type > Price > Flavor > Size] mirrors that common wrong framing, so the first slider move (Type to position 1) already delivers a meaningful teaching shift in the competitive set.
The sample is intentionally small (6 products, 4 brands, 3 types). Real chocolate‑tablet assortments run to dozens of SKUs in a large store; this sandbox teaches the STRUCTURE and state‑machine logic, not exhaustive category modeling. For portfolio‑level work, extend the methodology to your full category assortment via scanner‑data cross‑elasticity analysis.

5.2Controls & toggles

Every input the calculator exposes, its range, and what it changes.

Control	Range	Default	What it changes
Up / Down reorder buttons (per level)	5 levels in any permutation (120 total orderings)	[Brand > Type > Price > Flavor > Size] (0% hierarchy accuracy, deliberately wrong)	Moves each decision gate up or down. Gate 1 explains the most decision variance (an illustrative 35 to 50%); gate 5 the least. Reordering reshapes the entire switching-probability matrix.
Focus Product dropdown	6 products (Corvane Milk Hazelnut/Dark Hazelnut/Milk Plain, Brindel Milk Hazelnut, Aventra Filled Fruit and Nut, Store Brand Milk Plain)	p1 Corvane Milk Hazelnut 100g	Sets the focus product for switching-probability analysis. The chart shows how likely a shopper leaving this product is to pick each of the other 5 competitors under the current hierarchy.
Hierarchy Accuracy tile	0%, 20%, 40%, 60%, or 100% (steps of 20 per correct position; 80% is unreachable, since fixing four of five positions forces the fifth)	0% at default (deliberately wrong teaching default)	Percentage of positions matching the illustrative target order for this chocolate-tablet example (Type > Brand > Flavor > Size > Price). Green 80% and above, amber 40 to 79%, red below 40%.
Top Competitor readout	Any of the 5 non-focus products	p6 Corvane Milk Plain 45g at 35% under default (INTRA-brand signal)	Single highest-probability competitor. Cross-brand dominance means real category competition; intra-brand dominance means a cannibalization artefact of hierarchy misordering.
Switching probability chart	5 bars per focus product, 2 to 55% each	Red 40% and above, gold 20 to 39%, grey below 20% at default (one gold at 35%, four grey)	Full competitor ranking. Color shifts as you reorder gates; the bar pattern reshape tells you which gates are load-bearing for this focus product.

5.3Step-by-step exploration

7-step guided exploration of the scenario.

Read the default (deliberately wrong) hierarchy
Leave every control at default. Read the hierarchy [Brand > Type > Price > Flavor > Size], the accuracy score (0%), and the top‑competitor readout.
Expected outcome: Hierarchy accuracy 0%. Not a single position matches the illustrative target Type > Brand > Flavor > Size > Price order. Focus p1 top competitor: p6 Corvane Milk Plain 45g at 35%. The signal: under a brand-first framing, your own Corvane Milk Plain is identified as the biggest threat to your Corvane Milk Hazelnut. That's textbook intra-brand cannibalization, but it's an artefact of the wrong hierarchy, not a real category signal.
Fix gate 1, move Type to the top
Click the Up button on the Type / Segment row until it is at position 1. New hierarchy: [Type > Brand > Price > Flavor > Size].
Expected outcome: Hierarchy accuracy 40% (2 of 5 positions match: Type at position 1 and Brand at position 2). Focus p1 switching probabilities partially reshape: p3 Brindel Milk Hazelnut climbs from 2% to 8% (Type matches, Brand differs, so 1 shared) and p4 Store Brand Milk Plain climbs from 2% to 8%. Cross-brand same-type competitors emerge. But p6 Corvane Milk Plain 45g STAYS at 35%: at hierarchy [Type, Brand, Price, Flavor, Size], p1 and p6 match Type, Brand and Price (both $2-$3), so 3 shared, so 35%. Teaching point: moving Type to gate 1 is NECESSARY but not SUFFICIENT. You need to finish the reordering (push Flavor ahead of Price) to see the intra-brand cannibalization signal drop.
Complete the target hierarchy
Continue reordering until the hierarchy reads Type > Brand > Flavor > Size > Price (move Flavor up, Size up, Price all the way down).
Expected outcome: Hierarchy accuracy 100%. Focus p1 top competitor: p6 Corvane Milk Plain 45g at 18% (down from 35% at default). The switching shift from 35% to 18% is the single most important number on the page: under the target hierarchy, p6 and p1 share Type and Brand but Flavor differs at gate 3, so 2 shared, so 18%. The cannibalization risk that looked dominant at the default just lost half its weight. Meanwhile p3 Brindel Milk Hazelnut sits at 8% (Type matches, Brand differs at gate 2, so 1 shared, so 8%) and p4 Store Brand Milk Plain at 8% (same), real cross-brand competitors, small but non-zero.
Test the price-first hypothesis
Reset, then reorder to [Price > Brand > Type > Flavor > Size].
Expected outcome: Hierarchy accuracy 20% (Brand lands in its target slot at gate 2). Focus p1 top competitor: p6 Corvane Milk Plain 45g at 35% AGAIN, but now because p1 and p6 share Price, Brand and Type (all three match at gates 1, 2, 3). The same 35% number as default hierarchy, different underlying attribute composition. Price-first reads the category as a 'price bucket shop' (shoppers pick price band first, then everything else), and in this example's illustrative order shoppers do not behave that way. The same top competitor number can come from very different hierarchies, and the accuracy score is what anchors interpretation to the assumed order.
Switch focus to p5, the orphan branch
Reset to the target hierarchy [Type > Brand > Flavor > Size > Price]. Change focus product to p5 Aventra Filled Fruit and Nut 180g.
Expected outcome: Every other product scores 2% switching probability, because p5 is the ONLY Filled product in the assortment, every other product's Type gate (gate 1) differs from p5's Filled type. Shared = 0, so base probability 2% across the board. The signal: under the target hierarchy, p5 has NO direct competitors in the modeled assortment. Two possible readings: (a) blue-ocean position, where Aventra owns the Filled branch; protect and grow it; (b) strategic orphan, where the range is missing the rest of the Filled branch (a Filled Caramel line, for example) that would compete with p5 but also expand the Filled sub-category. CDT separates the two readings by asking: does shopper research show Filled tablets as a significant share of category occasions? If yes, expand the branch; if no, Aventra is a niche play and mass-market investment in Filled is wasted.
Compare two focus products under the same hierarchy
Keep the target hierarchy. Switch focus back and forth between p1 (Corvane Milk Hazelnut) and p3 (Brindel Milk Hazelnut).
Expected outcome: p1 top competitor under the target hierarchy: p6 (Corvane Milk Plain 45g) at 18%. p3 top competitor: p1 at 8% (shared Type only; brand differs at gate 2). Read the asymmetry: Corvane's intra-brand cannibalization (p1 and p6) is higher than its cross-brand competition with Brindel (p1 and p3). For Brindel, p1 IS the biggest threat at 8%, so the brand-level competitive intensity looks different from each side. This kind of asymmetric competitive picture is a normal finding in CDT research; single-number cross-price elasticities hide it.
Map back to PPA tools and the rest of the portfolio architecture
Open the related‑concept links (OBPPC Framework, Pack Roles Framework, Pack‑Size Elasticity, Good/Better/Best). Cross‑reference the OBPPC Matrix Builder and the Pack-Size Elasticity Calculator.
Expected outcome: Understanding that the CDT is the DEMAND-SIDE root of every PPA decision. PPA Lesson 4 OBPPC (the [OBPPC Matrix Builder](/tools/obppc-matrix-builder)) assumes a CDT hierarchy when it maps Occasion by Channel cells to pack choices; PPA Lesson 3 Incentive Curve (the [Pack-Size Elasticity Calculator](/tools/pack-size-elasticity-calculator)) operates WITHIN a CDT branch (the 70-85 RSP/kg target is meaningless if the packs being compared are in different CDT branches); PPA Lesson 2 Pack Roles slot into the Size/Format gate of the CDT. All downstream PPA tools assume a specified CDT hierarchy; running any of them with the wrong hierarchy produces technically-correct numbers that mean the wrong thing. Always run the CDT FIRST.

5.4Reading the output

Every KPI, the formula behind it, and how to interpret a positive or negative value.

KPI	Formula	How to read it
Hierarchy Accuracy	positions matching correct-order x 100 / 5	Green 80% and above, amber 40 to 79%, red below 40%. Your ordering scored against the illustrative target order for this chocolate-tablet example (Type > Brand > Flavor > Size > Price). Any real order is category-specific and comes from category-specific shopper research; this score is only valid inside this example.
Top Competitor	argmax over non-focus switching probabilities	The single highest-probability switch target from the focus product. Intra-brand means a cannibalization signal (often a hierarchy artefact); cross-brand means real category competition.
Switching Probability (per competitor)	[0.02, 0.08, 0.18, 0.35, 0.55][shared-gate count]	5-tier coarse model indexed by how many top-of-tree gates the focus and competitor share before the first mismatch. The bar color and length reshape as you reorder gates, and that reshape IS the teaching point.
Competitive Set Shape	bar pattern across all 5 competitors	The full bar chart. One warm bar (gold, or red at the top tier) over mostly grey means a tightly-clustered set dominated by one threat. Several gold or red bars mean a diffuse set with many real threats. All grey means an isolated product in its own CDT branch (blue ocean OR strategic orphan).

Read Hierarchy Accuracy first. Everything else is interpretable only in context of whether the current hierarchy matches reality. Then read Top Competitor and check intra‑brand vs cross‑brand. A high‑accuracy‑hierarchy cross‑brand top competitor is the normal, defensible finding; a low‑accuracy‑hierarchy intra‑brand top competitor is almost always a cannibalization artefact that disappears when you fix the hierarchy. The full competitive set shape (bar chart color distribution) gives you the texture of the category: dominated‑by‑one, diffuse‑competition, or isolated‑branch.

5.55 common mistakes to avoid

Diagnostic patterns that catch most misuse of this calculator in practice.

Mistake 1Applying this example's order to a different category
Symptom: Hierarchy accuracy reads 100% but shopper behavior in the actual category (yogurt, cereal, coffee) doesn't match the model's predictions.
Fix: The Type > Brand > Flavor > Size > Price order is an ILLUSTRATIVE anchor for this chocolate-tablet example, and any real order is CATEGORY SPECIFIC. Yogurt research often shows Flavor first; coffee often shows Brand close second; cereal often shows Benefit / Dietary (gluten-free, high-protein) as the first gate. Always derive the correct hierarchy from category-specific shopper research, and commission the study if you don't have it. A wrong hierarchy with a 100% accuracy score against a wrong benchmark is the most dangerous possible outcome.
Mistake 2Treating the intra-brand dominance as real cannibalization instead of a hierarchy artefact
Symptom: A portfolio decision (delist the 'cannibalizing' SKU, consolidate to one size) is made based on the default hierarchy's top-competitor readout.
Fix: Before any delist / consolidate decision, run the CDT at the hierarchy your category research supports (here, the illustrative target order). If the intra-brand dominance PERSISTS at that hierarchy, the cannibalization is real and a portfolio decision is defensible. If the intra-brand dominance DISAPPEARS at that hierarchy (as happens with p1 and p6 in the default scenario: 35% down to 18%), the cannibalization was always an artefact of the wrong framing; delisting would remove a legitimate shopper choice.
Mistake 3Assuming the 6-product model is a complete category representation
Symptom: A CDT-based strategy recommendation concludes "the category has only 6 meaningful SKUs" or "Aventra Filled Fruit and Nut has no competitors."
Fix: The 6 products are a teaching sample. Real chocolate-tablet assortments run to dozens of SKUs in a large store. The p5 Aventra "no competitors at 2% across the board" finding is informative (the Filled branch is under-populated in this sample) but NOT a portfolio recommendation on its own. Extend the CDT analysis to the full category assortment via scanner-data pairwise cross-elasticity estimation before recommending any branch-expansion investment.
Mistake 4Running PPA tools (OBPPC, Incentive Curve, Pack Roles) before the CDT
Symptom: PPA analysis recommends a mid-tier pack gap that, after CDT research, turns out to sit in a branch shoppers don't actually navigate to; the new SKU launches and immediately underperforms.
Fix: CDT is the DEMAND-SIDE root of every PPA decision. PPA Lesson 2 Pack Roles assumes the Size/Format gate; PPA Lesson 3 Incentive Curve operates WITHIN a CDT branch; PPA Lesson 4 OBPPC assumes a CDT hierarchy per cell. Always run the CDT first, then the PPA tools. The published PPA Lesson 7 Module Arc literally says "CDT is the demand-side root of all PPA work."
Mistake 5Confusing 'cross-price elasticity' with 'CDT switching probability'
Symptom: Scanner-data cross-XED analysis shows a 0.15 cross-elasticity between p1 and p3; CDT model shows 8% switching probability. Analyst concludes "the numbers don't match" and dismisses one or both models.
Fix: They're different constructs. Cross-price elasticity measures volume response to PRICE changes; CDT switching probability measures the PICK rate under a free-choice decision. They correlate (products with high cross-XED usually have high CDT switching too) but they are not the same measurement. Use cross-XED for promo-priced competitive modeling (Pricing Lesson 6); use CDT for category structure / assortment / OBPPC decisions.

Related concepts

Go deeper on the theory

Continue with the lessons

Go further inside Price Pack Architecture

This calculator is the sandbox slice of Lesson 7: Consumer Decision Tree. Each of the other 6 Price Pack Architecture lessons teaches a complementary concept that sharpens how you read the output above.

See Shopper-Occasion Pack Matrix inside the full lesson

RGM Academy lets you pull every commercial lever yourself inside a senior-practitioner simulator, with the AI RGM Strategist coaching every decision you make.

Unlock the full Consumer Decision Tree lesson

Or see what a team rollout includes