ZenotiHelp

How the Retention Manager Works

ZTZenoti Docs Team Updated October 25, 2026 7 min readGuide

The How the Retention Manager works view in AI Retention Manager explains the signals the model uses when calculating churn risk for guests across your organization. Opening this view gives you a read-only explanation of those signals directly from the dashboard β€” no configuration or retraining is available from this entry point.

Audience: Employees who use the AI Retention Manager dashboard to monitor guest retention risk.

Before you begin

  • You must have access to the AI Retention Manager dashboard for your organization.

    Permission: AI Retention Manager security object β€” View to open the dashboard and read it; Edit to create target segments, refresh member lists, and send guests to AI Lead Manager. Set per role at Configuration > Security > Security roles.

  • The dashboard must have finished processing your organization's guest data. If setup is still in progress, the dashboard displays the setup view instead of risk data. See Set up onboarding and understand model training.

  • The How the Retention Manager works view is available regardless of whether you are viewing the dashboard in organization scope or center scope.

Note: The How the Retention Manager works view is read-only. You cannot toggle signals on or off, or trigger a model retrain, from this entry point.

Steps

Open the How the Retention Manager works view

  1. Open AI Retention Manager from the Guest menu in the Zenoti left navigation.
  2. Wait for the dashboard to finish loading. The header displays the main navigation once risk data is ready.
  3. In the dashboard tab row, select How the Retention Manager works β€” the rightmost entry in the navigation.
  4. The view opens, displaying the signals the model uses to calculate risk for guests in your organization.

Read the view

Once the view is open, you will see the following sections:

  • Prediction signals β€” The signals the model considers when scoring each guest's churn risk.
  • The bands β€” An explanation of how the model assigns guests to one of the four risk bands. The view describes this as: The tool weighs each guest's full history against their own normal pattern and lands on one score: how likely that guest is to stop visiting. The score sets the color, like a traffic light.
  • Top parameters impacting your business β€” The signals listed here are ranked by how much each one moves the risk score. The subheading reads: Ranked by how much each moves the risk score.
  • Prediction accuracy β€” Shows how often the model is right about which guests are at risk, along with whether accuracy has moved up or down since the last update. If the model has not yet completed its first build, the view displays Prediction accuracy: not measured yet and the message: We'll show this once your Retention Manager has finished building.
  • These signals are managed by Zenoti β€” A note confirming that signal management is handled by Zenoti, not by individual centers or administrators.

If signal data has not yet been prepared for your organization, the view displays: The live signal list for your business is being prepared. Everything above is up to date. Come back shortly to see which signals the Retention Manager uses.

Understand the org-wide scope note (center scope only)

If you are viewing the dashboard in center scope, the view displays the following note:

These signals apply to your whole organisation's predictions. They are the same across all centers, not specific to this one.

This note confirms that the model is trained per organization, not per center. The signals shown reflect organization-wide predictions regardless of which center context the dashboard is currently displaying.

Switch between First time Visitors and Repeat Guests cohorts

The dashboard splits the scored guest base into two cohorts, presented as tabs:

  1. First time Visitors (first-visit guests)
  2. Repeat Guests (returning guests)

The Repeat Guests tab is selected by default when you open the dashboard. To switch cohorts, select the tab for the cohort you want to view.

Tab labels and guest counts

Each tab label shows the guest count for that cohort once the data has loaded. Before the count is available, the label renders without a count β€” no placeholder zero is shown. A header badge displays the total number of scored guests across both cohorts at all times, regardless of which tab is currently selected.

Behavior during a cohort switch

When you select a different cohort tab:

  • The currently displayed panel stays visible but dims while the new cohort's data loads. The panel does not collapse to a loading state, so the page height and your scroll position do not change.
  • The cohort label and its figures update together once the new data arrives. A cohort label is never displayed alongside another cohort's figures during the transition.
  • If you switch cohorts repeatedly before data returns, only the data for the most recent switch is applied. Earlier in-flight requests are discarded.

Sharing a cohort view as a link

The page address reflects the cohort you have selected, so you can share a direct link to a specific view:

  • The Repeat Guests tab is the default. When this tab is selected, no cohort value appears in the page address, keeping organization links clean.
  • To link directly to the First time Visitors tab, the page address must include ?cohort=first_visit. Any missing or unrecognized cohort value in the address defaults to the Repeat Guests tab.

Tip: To share a link that opens directly on the First time Visitors cohort, copy the page address after selecting the First time Visitors tab. Recipients who open the link will land on that view automatically.

Tips

  • If How the Retention Manager works is not visible in the dashboard tab row, the dashboard may still be loading or your organization's model may not yet have completed its first build. Wait for the dashboard to fully load and try again.
  • Prediction accuracy is a useful indicator of model health. If accuracy is trending down, it does not mean the model is broken β€” it means the model's estimates are less certain and you should treat risk bands as prompts to reach out, not as certainties.
  • The signals list loads asynchronously. If you open the view and see Loading signals…, wait a few seconds and the list will populate without requiring a page reload.
  • Risk data across the dashboard has a maximum delay of 24 hours. The signals shown in this view reflect the same data window β€” they are near real time, not instantaneous.
  • Use the How the Retention Manager works view alongside the Client Watchlist and the About this group panel beside Retention Groups to build a fuller picture of what is driving risk in your organization. The signals view explains which factors the model weighs globally; the Client Watchlist and About this group panel show how those factors play out for individual guests and risk bands.
  • When switching between the First time Visitors and Repeat Guests cohort tabs, the header badge always reflects the total scored guest count across both cohorts β€” use it to keep track of your full guest base size regardless of which tab is active.
  • If you switch cohorts quickly while data is still loading, the dashboard applies only the most recent selection. Wait for the dimmed panel to resolve before switching again if you want to confirm data for an intermediate cohort.

Troubleshooting

The How the Retention Manager works view is not visible in the dashboard tab row. The dashboard tab row appears once the dashboard has finished loading and your organization's model has completed its first build. If the tab row is not visible, the dashboard may still be initializing. Wait for loading to complete and refresh the page. If the tab row remains absent, see Set up onboarding and understand model training to confirm whether model training has finished.

The view shows "Loading signals…" and does not progress. This indicates the view is still fetching data. Wait a few seconds and check again. If the view does not load after a full page reload, the signal list for your organization may still be being prepared β€” the view will display a message confirming this. Come back shortly and try again.

The view displays "Prediction accuracy: not measured yet". This appears when your organization's first model has not yet finished building. It is expected during initial setup. Once the first model completes, the accuracy figure will appear automatically. See Set up onboarding and understand model training for context on how the model builds.

The org-wide scope note is not visible when viewing the dashboard at center scope. The note only appears when you are viewing the dashboard in center scope. If your session is scoped to the organization level, the note does not appear β€” this is expected behavior.

The view shows an unexpected error instead of signal data. If the dashboard displays Something went wrong loading this view, select Reload to attempt recovery. Your data is not affected. If the error persists, contact your Zenoti administrator.

A cohort tab shows no guest count after switching. Tab labels load their guest counts after the cohort data arrives. If a count does not appear, the data may still be loading β€” the panel will dim during the load and update once data returns. If the count remains absent after the panel fully resolves, select Reload and try again.

After switching cohorts quickly, the displayed data does not match the selected tab. If you switch tabs in rapid succession, the dashboard applies only the most recent selection and discards earlier in-flight requests. Wait for the panel to fully resolve after your last tab selection before concluding the data is incorrect.

A shared link does not open on the expected cohort. Links to the Repeat Guests cohort work with a plain organization URL β€” no cohort value is required. Links to the First time Visitors cohort require ?cohort=first_visit in the page address. If a link includes a missing or unrecognized cohort value, the dashboard opens on the Repeat Guests tab by default. Verify the link was copied after selecting the First time Visitors tab.

Was this article helpful?
Your feedback helps us improve our documentation.