ZenotiHelp

Review the explainability panel for a recommendation

ZTZenoti Docs Team Updated October 12, 2026 6 min readGuide

The explainability panel opens when you click a product row in the action grid, showing you the reasoning behind each replenishment suggestion β€” the inputs, step-by-step calculations, and the AI model's confidence in its forecast. From the panel, you can also override the suggested order quantity with a reason code and give direct feedback to improve future recommendations.

Audience: Inventory Coordinators who review and act on AI-generated replenishment suggestions.

Availability: AI Inventory Manager is in beta and requires CSM enablement. Contact your Zenoti customer success manager to activate it for your organization.

Before you begin

Permission: Add or Edit (on the Inventory security object) β€” required to override a suggested order quantity from the explainability panel.

  • AI Inventory Manager must be enabled for your center and the onboarding setup must be complete.
  • At least one recommendation row must exist in the action grid.
  • To override a suggested order quantity, you need Add or Edit permission on the Inventory security object.

Steps

Open the explainability panel

  1. Navigate to AI Inventory in the main navigation.
  2. In the action grid, click a product row.
  3. The explainability panel opens and displays exactly four parts: Summary, Inputs, Calculations, and Confidence.

Review the panel parts

The panel displays exactly four parts, each covered below: Summary, Inputs, Calculations, and Confidence.

Summary

  1. Read the Summary verdict. It shows one of three statements:
    • Replenishment recommended β€” the product's stock is at or below the reorder level and a replenishment order is suggested.
    • Stock level is adequate β€” current on-hand and incoming stock cover the reorder coverage period without ordering.
    • Stock integrity issue: on-hand is zero or negative β€” on-hand data may be inaccurate. Investigate before placing an order.

Inputs β€” fields and their sources

  1. In Inputs, review each value and its tagged source:

    Input Source
    Predicted daily usage AI model forecast
    Lead time One of: configured (manual override set by your team), learned-from-PO-history (derived from past purchase orders), or system-default
    Reorder coverage Applied setting (org, center, or product level)
    Safety buffer Applied setting (org, center, or product level)
    On-hand quantity Current inventory records for this center
    Incoming quantity Open purchase orders for this product

Calculations

  1. In Calculations, follow the step-by-step reorder arithmetic. The panel labels each line of the formula:

    • Daily usage: β€” the AI model's predicted daily usage for this product.
    • Γ— Lead time: β€” daily usage multiplied by lead time in days, showing the stock needed during the delivery window.
    • =Needed during delivery: β€” the total units required while waiting for stock to arrive.
    • + Buffer (X/day Γ— Y days): β€” safety buffer units added to guard against demand spikes or late deliveries.
    • =Reorder level: β€” the dynamic reorder point at which a replenishment is triggered.
    • Order quantity: β€” the suggested quantity after accounting for target stock, on-hand, and incoming stock.

    Note: Only Predicted daily usage and the confidence score come from the AI model. All values in Calculations are transparent, auditable arithmetic β€” not AI predictions. Given the same inputs, the calculations always produce the same result.

Confidence

  1. In Confidence, review the tier badge (T1 β€” High, T2 β€” Medium, T3 β€” Low, or T4 β€” Very Low) and its plain-language meaning for this product. A higher tier means the AI's demand forecast is more stable and reliable and the suggestion is a strong starting point; a lower tier means you should scrutinize the suggestion and verify on-hand and incoming quantities before ordering.

    The tier reflects how stable and consistent the AI's demand forecast is for this product β€” not a direct count of sales days. For the exact score range behind each tier and how to act on each one, see Check confidence scores and tiers.

Override a suggested quantity

  1. In the explainability panel, locate the Suggested order quantity field and enter the quantity you want to order.

  2. Select a reason code that best explains why you are changing the suggestion:

    • Promotion β€” a planned promotional event will drive above-average demand.
    • Seasonal β€” seasonal demand patterns differ from the AI's historical baseline.
    • Deal β€” a supplier deal makes ordering a larger or smaller quantity more cost-effective.
    • Too high β€” the suggested quantity exceeds what is needed for this period.
    • Too low β€” the suggested quantity does not cover expected demand.
    • Other β€” none of the above applies.
  3. Click Submit.

    • A non-blocking acknowledgment appears on screen and dismisses automatically.
    • The system records the recommended quantity, the quantity you entered, the difference between them, and your reason code.
    • Your override is included in future model training to improve recommendations for this product.

Note: To accept the suggested quantity without change, close the panel without clicking Submit. No override record is created.

Give feedback on a recommendation row

  1. In the action grid, hover over a product row.
  2. Click the thumbs-up icon to confirm the recommendation is accurate, or the thumbs-down icon to flag it as inaccurate.
  3. Feedback submitted appears on screen, confirming that the system has recorded your signal for future model training.

Tip: For richer feedback β€” such as explaining why a quantity is wrong β€” use the override flow (steps 8–10) rather than the thumbs-down icon. An override with a specific reason code provides a stronger training signal than a thumbs-down alone.

Worked example: Your action grid shows Aveda Shampoo with a Suggested qty of 24 units. You click the row. Summary reads "Replenishment recommended." Inputs shows Predicted daily usage of 3 units, lead time of 4 days (learned-from-PO-history), Reorder coverage of 30 days, and a safety buffer of 2 days. Calculations shows: Γ— Lead time gives 12 units (3 Γ— 4), buffer adds 6 units (3 Γ— 2), =Reorder level: 18 units, Order quantity: 24 units. Confidence shows T2 β€” Medium (score 0.62). A promotion is running next week, so you enter 48 in Suggested order quantity, select Promotion, and click Submit. The acknowledgment appears and the override is recorded.

Tips

  • Check the Lead time source tag before overriding. If lead time is tagged system-default, it may not match your actual supplier delivery time. Update the lead time setting at the center or product level so future calculations use accurate data.
  • Use T4 β€” Very Low confidence as a signal to verify data before ordering. Confirm that On-hand stock and Incoming stock are accurate in the Inventory module before acting on a very low-confidence suggestion.
  • Choose a specific reason code β€” Promotion, Seasonal, or Deal β€” rather than Other whenever possible. Specific codes give the AI model clearer feedback and improve future suggestions for similar demand patterns.
  • If Summary shows a stock integrity issue, correct the on-hand balance in the Inventory module before placing an order. Ordering against an inaccurate on-hand quantity can result in double-ordering.

Troubleshooting

  • Summary shows "Stock integrity issue: on-hand is zero or negative" but the product is physically in stock. On-hand quantity in Zenoti does not reflect actual stock levels. Perform a stock count in the Inventory module to correct the on-hand balance, then wait for the next nightly refresh for AI Inventory to recalculate.
  • Submit is unavailable when overriding. The quantity entered matches the suggested quantity, the quantity field is blank, or no reason code is selected. Enter a quantity different from the suggested quantity and select a reason code before clicking Submit.
  • Confidence shows T4 β€” Very Low for a product with regular sales. This reflects high forecast variance or insufficient recent sales data for this SKU at this center. Confirm the product has been sold consistently from this center. If the product was recently added or had a gap in sales, allow additional sales cycles for the forecast to stabilize.
  • Calculations values appear incorrect after you changed a setting. Settings changes apply at the next nightly refresh, not immediately. Wait until the following day for updated values to appear in the explainability panel.
  • The thumbs-up or thumbs-down icons are not visible. The cursor is not directly over the product row. Move the cursor over the product row to reveal the feedback icons.
Was this article helpful?
Your feedback helps us improve our documentation.