Demand Influencers helps you understand how products in an assortment may affect each other's demand. When similar products are sold during overlapping periods, adding, removing, or changing one can redistribute demand across the others — Demand Influencers identifies these relationships and estimates how strongly one product may substitute for another.

The feature combines normalized demand signals, sales behavior, and assortment context into two key metrics: Cannibalization Risk and Confidence. It's integrated into the Similar Choices tab in Assortment Planning (Choice Plan) and extends the existing Similarity Index framework from the Similar Choices tab.

Note: Demand Influencers is diagnostic. It surfaces a risk signal to help you make decisions — it doesn't change your forecast, and nothing downstream of it is adjusted automatically.

Demand Influencers requires two things to be turned on for your organization:

Demand Influencers evaluates products through three sequential layers, moving beyond simple similarity to determine which products are likely to compete for the same demand:

Raw sales can be affected by factors like promotions and seasonal demand patterns. Before comparing products, Demand Influencers removes these external effects so the comparison reflects true underlying demand.

The model accounts for:

The result is a normalized demand time series for each choice — a more consistent representation of how customers actually demand the product, and the foundation for the behavior and cannibalization calculations below.

After demand is normalized, Demand Influencers compares each product with its similar products to understand how their sales behaviors relate to each other.

The comparison focuses on products that have demonstrated similar demand behavior historically. When these relationships are shown within an assortment, the system uses the timeframe overlap between the products' assortment lifecycles to determine which similar products are relevant to the selected choice.

This extends the existing Similarity Index by incorporating actual historical sales behavior, rather than relying only on product attributes and price. Two signals feed it:

A pair is tagged one of four ways:

A percentage is available on hover next to the tag. It only reflects the cannibalization-evidence half of the signal, so Co-moving and Independent pairs can both show a value near 0% — the tag, not the percentage, is what tells them apart. A higher percentage means stronger measured evidence that the products compete for the same demand; a low or blank percentage doesn't mean the products are unrelated, only that no substitution signal was found in the available history.

Because the comparison uses normalized demand signals, it's less affected by temporary promotions or seasonal fluctuations, giving a more reliable view of the underlying relationship between products.

Similarity alone doesn't mean two products will compete for the same demand. Demand Influencers combines the behavioral relationship between a product and its similars with additional assortment context to estimate the likelihood and strength of demand substitution.

Cannibalization Risk represents the expected strength of demand substitution between two products, scaled by how much their planned timing and locations actually overlap:

Two products can have a strong underlying behavioral relationship and still show a modest Cannibalization Risk if they only share part of the plan's timeframe or locations — the plan overlaps are gates on the risk, not just contributing factors. Risk is blank when a choice has no known sale dates or no planned locations to compare.

The resulting score ranges from 0 to 100:

A higher score means the two products are more likely to compete for the same customer demand — but the relationship is one-directional. A high score is meaningful evidence; a low or blank score is inconclusive rather than a sign the products are safe to plan together, since the underlying model is more likely to miss a real relationship than to flag one that isn't there. Risk is also not a share of demand — a risk of 60 doesn't mean 60% of a product's demand would move to the other.

Confidence indicates how much sales evidence supports the Cannibalization Risk measurement — not how strong the relationship is. A high Cannibalization Risk paired with High Confidence means the model has substantial historical data behind that estimate; a high Risk with Low Confidence means the estimate is based on comparatively little shared sales history.

Confidence is based on how many weeks of usable sales history were available to measure the relationship — more overlapping, unmasked weeks of data between the two products means higher Confidence.

Confidence always displays as a Low / Medium / High badge, with the underlying percentage available on hover:

Below roughly 50%, the measurement leans more on price and historical lifecycle context than on directly measured sales behavior — treat those relationships as preliminary.

Note: Cannibalization Risk and Confidence answer different questions. Cannibalization Risk measures how strong the expected substitution is. Confidence measures how much evidence backs that estimate.

Demand Influencers builds on the existing Similarity Index, but the two concepts serve different purposes.

Similarity Index answers "Which products are most similar to this product?" It's based on descriptive similarity, attribute overlap, and price proximity — see Similar Choices Tab for details.

Cannibalization Index answers "Which similar products are most likely to compete with this product for demand?" It's the first metric in this article to incorporate actual historical sales behavior, and adds assortment context on top of it.

Similar products can be identified independently of whether they're currently active in the same assortment timeframe. But when Demand Influencers is shown within Assortment Planning, assortment timeframe overlap is used to determine which similar products are relevant to show for the selected choice — so you can focus on the similar products that matter right now, rather than reviewing every similar product.

For a selected choice, the Demand Influencers view shows a ranked list of similar products, with relevant assortment timeframe overlap, that may be influenced by the selected choice:

Products are ranked by Cannibalization Risk, highest first, so you can quickly spot the strongest potential demand relationships.

Imagine an assortment contains Product A, Product B, and Product C.

Product A and Product B are identified as similar products with overlapping assortment timeframes. Their historical sales behavior shows meaningful co-movement in opposite directions over a solid stretch of shared sales history, similar prices, and their planned lifecycles and locations overlap heavily.

Their Demand Influencers relationship might show:

Even a strong, well-evidenced relationship rarely lands near the top of the 0–100 scale, because Risk is scaled down by how much the two products' planned timing and locations actually overlap on top of the underlying behavioral signal. A relationship in the 50s or 60s with High Confidence should be read as a strong signal, not a weak one.

Product C may also be similar to Product A, but if its assortment timeframe doesn't overlap with Product A's, it may not be surfaced as a relevant Demand Influencer for the current assortment view.

When adding a new product, use Demand Influencers to identify similar products with overlapping assortment timeframes that may experience demand substitution. This can help answer: "If I add this product, which existing similar products could lose demand?"

When removing a product, Demand Influencers can help identify similar products with relevant timeframe overlap that may absorb some of its demand. This can help answer: "If I remove this product, which remaining similar products are likely to benefit?"

Changing the launch date changes how much two products overlap within the assortment. Use Demand Influencers to understand how different launch timing can affect potential demand redistribution. Save the plan after changing a date — Demand Influencers reflects the saved plan, not unsaved edits, and shows a plan-edited notice until you save.

Extending or shortening a product's lifecycle can increase or reduce its overlap with similar products. Demand Influencers helps identify relationships that may change as assortment timeframe overlap changes. As with launch dates, save the plan before re-checking the numbers.

Selling similar products in overlapping locations can increase the potential for demand substitution. When changing or excluding selling locations through Manage Exclusions, use Demand Influencers to understand how the resulting location overlap may affect cannibalization. Save the plan before re-checking Location Overlap (Plan) — it won't reflect unsaved location changes.

Use the ranked Demand Influencers list to identify the strongest substitution relationships among similar products with relevant assortment timeframe overlap. This helps you distinguish between:

Before approving an assortment plan, review products with high Cannibalization Risk to identify potentially redundant or highly substitutable choices. This can support better assortment composition decisions and reduce unexpected demand redistribution after launch.

Demand Influencers helps you:

A few reasons a similar product might not appear:

A missing product isn't the same as a cleared one — it may be a real competitor the model simply couldn't measure yet, rather than one that scored low.

No. Similarity alone doesn't mean two products will compete for the same demand. Cannibalization Risk layers additional assortment context — price proximity, historical lifecycle overlap, and how much the products' planned timing and locations actually overlap in the current plan — on top of the Sales Behavior signal, so two products flagged as cannibalistic can still score low on Cannibalization Risk if their plan overlap is small.

They answer different questions. Cannibalization Risk measures how strong the expected demand substitution is; Confidence measures how much sales evidence backs that estimate.

Similarity Index answers "which products are most similar to this product?" based on characteristics and price. Demand Influencers (Cannibalization Index) adds actual historical sales behavior and assortment context to answer "which of those similar products are most likely to compete with this one for demand?"

In Choice Plan, select a plan row, open the Offerings panel, go to the Similar Choices tab, and click View Demand Influencers. It requires Similarity Index and Cannibalization enabled in Settings → ML Configuration, and a populated cannibalization dataset for your organization.