TotalApp Docs

AI Predictive Models

Configure and monitor the RFM-based churn-risk and predicted-LTV scoring engine that powers CRM scoring and Customer Segmentation.

Overview

AI Predictive Models is the configuration and monitoring console for TotalApp's RFM-based (Recency, Frequency, Monetary) prediction engine. The engine scores your CRM contacts/deals to produce a churn-risk score and a predicted lifetime value (LTV) per cluster, and this screen is where you configure how it runs, trigger recalculations, and review its run history.

Same Engine, Different Screen

This screen and the Customer Segmentation screen in the Marketing group both run against the same underlying ML Prediction Engine. AI Predictive Models is aimed at admins/ops configuring and monitoring the engine; Customer Segmentation is a marketer-facing screen for running ad-hoc segmentation with a simpler, campaign-oriented UI.

KPI Strip

Four tiles at the top of the screen summarize the engine's most recent activity:

Last Run

Timestamp of the most recent scoring run, or a placeholder if the engine has never run for this tenant.

Deals Scored

Number of CRM deals evaluated in the most recent run.

CRM Sync

Shows On or Off depending on the current CRM Auto-Sync setting, colored to draw attention when sync is active.

Revenue Pool

Total monetary value covered by the most recent run, formatted compactly (e.g. "$1.2M").

Engine Settings

CRM Auto-Sync

A toggle that controls whether scoring results are written back onto your CRM deal records. When on, each recalculation updates the churn-risk score and predicted LTV fields on the relevant deals in CRM. When off, scoring still runs and appears in the run history, but no CRM deal records are mutated — useful for previewing the effect of a settings change before committing it to live data.

Default RFM Weight Configuration

Three sliders let you set the default split between the three RFM axes used to compute each contact's composite score:

  • Recency — how recently the contact engaged (more recent = higher score)
  • Frequency — how often the contact has engaged, weighted by their CRM pipeline stage
  • Monetary — the total monetary value associated with the contact

Weights are automatically normalized so the three values always sum to 100%. Click Save Settings to persist your changes — a confirmation checkmark appears briefly after a successful save.

Recalculate Now

The Recalculate Now button in the header triggers an immediate scoring run using the current default RFM weights, across the full CRM contact base ("all" scope). While running, the button shows a spinner and is disabled to prevent duplicate runs. On completion, the KPI strip and run history table refresh with the new run's results.

Run History

A table of past scoring runs, most recent first, with one row per run:

ColumnMeaning
TimeWhen the run completed
ScopeWhich subset of contacts/deals the run covered
Deals ScoredNumber of deals evaluated in that run
CRM MutatedBadge showing whether that run's results were written back to CRM deal records (depends on the CRM Auto-Sync setting at the time the run executed)
RevenueTotal monetary value covered by that run's scored deals

If no runs have ever been triggered for your tenant, the table shows an empty state instead.

Relationship to Customer Segmentation

Both AI Predictive Models and Customer Segmentation (in the Marketing group) call the same underlying RFM scoring engine — the same recency/frequency/monetary math, the same clustering logic, and the same churn-risk and LTV calculations. The two screens exist for different users:

  • AI Predictive Models (this screen) — for admins/ops configuring the engine's defaults, monitoring whether CRM sync is happening, and auditing run history over time.
  • Customer Segmentation (Marketing) — for marketers who want to run an ad-hoc segmentation with custom weights for a specific campaign, without needing to touch the engine's tenant-wide default configuration.

Changing defaults here affects Customer Segmentation's starting point

The default RFM weights you configure on this screen are the starting values a marketer sees when they open Customer Segmentation, though they can still adjust the weights there for a specific run without changing your tenant-wide defaults.

AI Assistant Panel

This screen follows TotalApp's dashboard AI Assistant pattern: it explains the engine's current configuration and recent run history in plain language, flags anomalies (such as a run that scored zero deals, or CRM sync unexpectedly being off), and can recommend configuration changes backed by the actual run history numbers — but it never edits settings or triggers a run on your behalf. Open it from the tab on the right edge of the screen; it supports fullscreen mode and Save as Report.

Frequently Asked Questions

What does turning off CRM Auto-Sync actually change?
Scoring still runs and shows up in the run history and KPI strip, but the resulting churn-risk score and predicted LTV are not written onto your CRM deal records. It's a safe way to preview what a weight change would produce before letting it touch live CRM data.
Do the RFM weights have to add up to 100%?
You don't need to balance them manually — the engine normalizes whatever values you set so they always sum to 100% when a run executes.
How is this screen different from Customer Segmentation in Marketing?
They use the same scoring engine, but AI Predictive Models is for configuring tenant-wide defaults and monitoring the engine over time, while Customer Segmentation is a marketer-facing tool for running one-off segmentations with custom weights for a specific campaign.
What happens if I click Recalculate Now while a run is already in progress?
The button is disabled and shows a loading state while a run is in progress, preventing you from triggering a duplicate concurrent run.
Can I see who or what triggered a past run?
The run history table shows the timestamp, scope, deals scored, CRM sync status, and revenue for each run so you can audit engine activity over time.