TotalApp Docs

Customer Segmentation

Group your CRM contacts into RFM-based segments (Recency, Frequency, Monetary value) and sync any segment straight to an email campaign.

ML-Powered Customer Segmentation

The TotalApp Customer Segmentation module goes beyond traditional rule-based segmentation, using machine learning algorithms to predict customer behavior. Hidden patterns discovered automatically by the system let you target audiences with greater precision.

Automatic Clustering

Automatically split your customers into meaningful groups using k-means and hierarchical clustering algorithms.

Churn Prediction

Identify customers at high risk of churn in advance and launch proactive retention campaigns.

LTV Scoring

Use models that predict customer lifetime value (LTV) to identify your most valuable customers ahead of time.

Purchase Prediction

Predict customers' interest in specific product categories and their likelihood to purchase.

Available ML Models

Model Description Use Case Minimum Data
Churn Prediction Identifies customers at high risk of churn Customer retention campaigns 500+ customers, 3+ months of data
LTV Score Predicts customer lifetime value Investment focused on high-value customers 1000+ customers, 6+ months of data
Purchase Prediction Predicts the timing of a customer's next purchase Time-triggered campaigns 200+ customers, 2+ months of data
RFM Segmentation Recency, frequency, monetary value analysis Customer value segmentation 100+ transaction records

Model Training

Models are automatically retrained every week. Initial training can take 24-48 hours. Model accuracy can be tracked from the dashboard.

Getting Started with an ML Segment

1. Choose a Model
2. Set a Threshold
3. Create the Segment
4. Attach to a Campaign

Threshold Values

You can set a threshold value for each ML model. For example, for churn prediction, you could select customers with "over 70% churn risk." This value directly affects the number of customers and the quality of the segment.

Tip

Very high threshold values (e.g. 90% churn risk) shrink the segment size. Try a range of 60-75% for a reasonable balance and observe the results.

Overview

Customer Segmentation reads deal activity from your CRM pipelines and scores every contact on three axes — how recently they engaged, how often, and how much revenue they represent. Those scores classify each contact into one of five segments: Champions, Loyal Customers, Potential Growth, At Risk, and Inactive. The result is a live segment list you can inspect, drill into, and push to a campaign — no manual list-building required.

The screen is built as three panels: Segmentation Settings on the left (weights and pipeline scope), Segments in the center (the classified cluster list), and the selected segment's Customer List on the right.

Adjustable Priorities

Three sliders — Recency, Frequency, Monetary — let you decide what matters most for this run. Weights are normalized automatically so they always add up to 100%.

Pipeline Scope

Run segmentation across every CRM pipeline, or restrict it to a single pipeline when you only care about one product line or region.

Five Segments

Champions, Loyal Customers, Potential Growth, At Risk, and Inactive — each with member count, average lifetime value, churn risk score, and revenue contribution.

Sync to Campaign

Select a segment and sync its customer list directly into an email campaign queue — no export/import step.

Marketing Mode vs. Data Mode

The screen supports two views of the exact same underlying data, switchable at any time from the toggle in the top-right corner. Your choice is remembered the next time you open the screen.

Area Marketing Mode (default) Data Mode
Screen title Customer Segmentation Audience ML Segmentation
Left panel Segmentation Settings — plain-language slider hint ML Model & Tuning (RFM) — normalization/clamping language
Run button Run Segmentation Execute ML Re-Clustering
Raw vector readout Hidden Visible — shows the normalized { r, f, m, scope } JSON payload sent to the clustering engine
Center / right panel labels Segments / Customer List, columns read "Segment" / "Customer" Cluster Ledger Matrix / Contact Manifest, columns read "Cluster" / "Contact"

Who uses which mode?

Marketing Mode is the default for growth marketers and campaign owners who just need to build and act on a segment. Data Mode is for analysts or data engineers who want to see the exact weight vector and scope payload behind a run before trusting it.

Building a Segment

1. Adjust priorities
2. Pick a pipeline scope
3. Run Segmentation
4. Select a segment & sync

1. Adjust priorities

Drag the Recency, Frequency, and Monetary sliders to reflect what should drive this segmentation. For a re-engagement campaign, weight Recency higher; for a VIP/loyalty campaign, weight Monetary higher. The three values are normalized automatically, so you don't need them to add up to 100% yourself.

2. Pick a pipeline scope

Choose "All Pipelines" to segment your entire CRM contact base, or pick a single pipeline to scope the run to one product line, region, or team.

3. Run Segmentation

Click Run Segmentation. A short live log (loading customer data → analyzing purchase activity → scoring → building segments) shows progress, then the Segments panel populates with all five segments and their metrics.

4. Select a segment and sync

Click a segment row to load its Customer List on the right. From there, click Sync to Email Campaign to push that segment's contacts into the campaign queue. Segments with zero matching customers cannot be synced.

Empty segments

If a segment matches zero customers under the current priorities and scope, syncing is blocked. Adjust the sliders or widen the pipeline scope and run segmentation again.

The Five Segments

Segment Profile
Champions High recency, high frequency, high monetary value — your top-tier, currently active customers.
Loyal Customers Frequently active, steady mid-spend customers with regular touchpoints.
Potential Growth Recently engaged but with lighter frequency or spend — good nurture-campaign candidates.
At Risk Meaningful historical value but fading engagement — prioritize for retention outreach.
Inactive Dormant customers with minimal recent activity and spend.

Each segment row shows member count, revenue contribution %, and a churn risk score. Selecting a segment shows every customer in it with their pipeline stage and individual score.

AI Assistant

Every segmentation run can be discussed with the built-in AI Assistant, opened from the tab on the right edge of the screen. It has full context on the current segment ledger and the customer list for whichever segment is selected, so you can ask things like:

  • "Summarize the current segment matrix"
  • "Which segments carry the highest churn risk?"
  • "Which segment should I sync to a campaign next?"
  • "Highlight the top customers in the selected segment"

Any reply can be saved as a report with one click, so you can share the analysis or revisit it later from My Reports.

Frequently Asked Questions

Where does the segmentation data come from?
From your CRM deals across the pipelines you scope the run to — each deal's stage and last-contact date drive the recency and frequency scores, and its value drives the monetary score.
Do I need to re-run segmentation after every new deal?
Segments reflect the data as of the last run. Re-run segmentation whenever you want an up-to-date picture — there's no automatic background recompute.
What's the difference between Marketing Mode and Data Mode?
They show the exact same segmentation run with different terminology and detail level — Marketing Mode uses plain-language labels and hides the raw weight vector; Data Mode shows ML/RFM terminology and the normalized vector payload. See the "Marketing Mode vs. Data Mode" section above.
Can I sync the same segment to a campaign twice?
Yes. Each sync creates a new queue entry — syncing again after new customers enter the segment is the normal way to keep a campaign's audience current.
Why is a segment empty?
No contact in the current pipeline scope matched that segment's recency/frequency/monetary thresholds. Try widening the pipeline scope or adjusting the priority sliders and running segmentation again.