Analyse de sensibilite pour Amazon FBA : lire les diagrammes tornado

You have run a simulation Monte Carlo and you know the probability distribution of your product's profitability. The next question is: which variable matters most? If you could improve just one aspect of your product economics, where should you focus? Sensitivity analysis answers this question, and the tornado chart is its visual language.

Qu'est-ce que l'analyse de sensibilite ?

Sensitivity analysis measures how much the output (profit) changes when you vary one input at a time, holding everything else constant. It answers the question: "If prix de vente drops 10%, how much does my annual profit change? What if COGS increases 15%?"

The process is straightforward:

1. Start with your base-case values for all variables
2. Take one variable (e.g., prix de vente) and move it to its low estimate
3. Calculer profit with this low value, all else at base case
4. Move the same variable to its high estimate
5. Calculer profit with this high value, all else at base case
6. Record the range of profit created by this single variable
7. Repeat for every input variable
8. Sort variables by the size of their impact range

The result is a ranking of variables from most impactful to least impactful. The tornado chart visualizes this ranking.

Anatomie d'un diagramme tornado

A tornado chart is a horizontal bar chart where each bar represents one variable. The bars are sorted by width from widest (most impactful) to narrowest (least impactful), creating a shape that resembles a tornado -- wide at the top, narrow at the bottom.

Here is an example for a hypothetical stainless steel water bottle with a base-case annual profit of $14,200:

Comment lire le diagramme

La barre superieure merite toute votre attention

In this example, prix de vente is the widest bar. When price drops to its pessimistic estimate ($19.99 vs. $24.99 base), annual profit falls to $3,800. When price rises to its optimistic estimate ($28.99), profit jumps to $24,600. The total swing is $20,800 -- a massive range from a single variable.

This tells you that pricing strategy is the single most important factor in this product's success. Anything that affects your ability to maintain price -- marque differentiation, listing quality, patent protection, limited concurrence -- has an outsized impact on profitability.

Les unites vendues sont presque toujours dans le top deux

The second bar shows units sold per month creating a swing from $5,100 to $23,400. Demand volume is the other critical driver. Together, price and volume (which multiply to form revenus) typically account for 60-75% of total profit variance in Amazon FBA businesses.

This is consistent with what probabilistic analysis shows: the revenus line is where uncertainty lives, not the cost line.

Les barres inferieures sont la ou les efforts sont souvent gaspilles

Storage fees create a swing of only $1,400. Spending significant time optimizing storage (beyond basic gestion des stocks) is mathematically unlikely to move the needle. Yet many vendeurs obsess over storage fees while ignoring their pricing strategy or PPC efficiency.

The tornado chart redirects your attention where it matters most.

Construire votre propre analyse de sensibilite

Etape 1 : definir le scenario de base

Start with your best-estimate values for each variable. Use the P50 (median) from your probability analysis if available, or your most realistic single-point estimate.

Etape 2 : variation une a la fois

For each variable, calculate annual profit with that variable at its low value (all others at base), then at its high value (all others at base). Record both results.

Etape 3 : trier par amplitude d'impact

Calculer the range (high profit - low profit) for each variable. Sort from widest range to narrowest. This gives you the tornado chart ordering.

Etape 4 : identifier les variables actionnables

Not all high-impact variables are controllable. Separate the top variables into two catégories:

A quoi ressemble le tornado typique Amazon FBA

After analyzing hundreds of Amazon FBA products across catégories, clear patterns emerge in which variables dominate:

For products priced $15-$30: Selling price and units sold are almost always the top two variables, accounting for 55-70% of total profit variance. PPC cost per unit is typically third. COGS is fourth. Everything else is secondary.

For products priced $30-$60: The same ordering holds, but COGS and PPC become relatively more important because the absolute dollar amounts are larger. PPC sometimes moves to second place in highly competitive catégories.

For products priced $60+: Referral fees become more significant (15% of $75 = $11.25 per unit). Units sold becomes the dominant variable because volume uncertainty is typically high for premium-priced items.

Avance : tableaux de sensibilite bidirectionnels

Tornado charts show one variable at a time. But what about interactions? A two-way sensitivity table varies two variables simultaneously, creating a grid of outcomes:

The highlighted cell ($14,200) is the base case. The table reveals that if price drops to $19.99 AND volume drops to 200 units, you lose $2,400 -- a scenario that the one-at-a-time tornado chart would not show because it never varies both simultaneously.

Two-way tables are particularly useful for price-volume interactions, since these two variables are often negatively correlated (lower prices tend to produce higher volume, and vice versa).

Utiliser l'analyse de sensibilite pour ameliorer votre produit

Sensitivity analysis is not just diagnostic -- it is prescriptive. Once you know which variables matter most, you can take targeted action:

If price is the top driver: Invest in marque building, better photography, A+ content, and product differentiation. A $2 price premium across 400 units/month = $9,600/year in additional profit. That easily justifies $2,000 in better product images and $500 in A+ content creation.

If PPC is the top driver: Audit your keyword strategy. Eliminate bleeding keywords. Implement negative keyword lists. Test Sponsored Brands vs. Sponsored Products. A reduction in ACOS from 35% to 25% on $3,000/month ad spend saves $300/month = $3,600/year. Understanding the difference between ACOS and TACoS is critical here.

If COGS is the top driver: Renegotiate with your fournisseur at higher MOQs. Get quotes from alternative fournisseurs. Consider nearshoring if the frais de livraison component of your landed cost is dominant. A $0.80 reduction in landed COGS across 4,800 annual units = $3,840/year.

The tornado chart tells you where to spend your time. If storage fees are at the bottom of your tornado, do not spend 20 hours optimizing inventaire placement. Spend those 20 hours on your top two variables instead.

Limites de l'analyse de sensibilite

One-at-a-time sensitivity analysis has known limitations. It assumes variables are independent, which they are not. It does not capture the simultaneous interaction of multiple variables moving against you (the "perfect storm" scenario). And it assumes linear relationships between inputs and outputs, which holds approximately but not perfectly.

This is why sensitivity analysis works best alongside simulation Monte Carlo, which captures all variable interactions simultaneously. The tornado chart tells you which variable to focus on. Monte Carlo tells you the overall probability of success. Together, they provide a complete picture of your product's risk-reward profile.