ML-Powered TTV Tracking: Measure & Accelerate Customer Path to Value

A client signed a contract, 30 days passed, and they still don't understand why they are paying. Typical situation. We know how to measure and shorten the path to value. Over 15+ projects implementing TTV tracking in B2B SaaS products, we have accumulated practice that guarantees measurable results: average TTV reduction 25% faster than manual analysis.

Time-to-Value (TTV) is the time from contract signing to the moment when the customer first gets meaningful value from the product Wikipedia. In B2B SaaS, it's a leading indicator of long-term retention: customers who reach the Aha-moment within 14 days have 40% higher retention than those who never reach it. We automate this tracking with ML, enabling risk prediction and intervention before the customer gets stuck.

How to Define the Aha-moment?

Aha-moment = first achievement of a value milestone. Specific definitions depend on the product:

• CRM: first successfully closed deal through the system
• Analytics platform: first dashboard viewed with real data
• Automation: first successfully executed workflow
• Communications: first 10 messages from the team

Defining the milestone through data:

def identify_aha_moment(product_events, cohort):
    """
    Correlation analysis: which action predicts retention best?
    Method: find events after which 90-day retention is highest
    """
    event_types = product_events['event_type'].unique()

    correlations = {}
    for event in event_types:
        users_with_event = product_events[product_events['event_type'] == event]['user_id'].unique()
        retention_with = cohort[cohort['user_id'].isin(users_with_event)]['retained_90d'].mean()
        retention_without = cohort[~cohort['user_id'].isin(users_with_event)]['retained_90d'].mean()
        correlations[event] = retention_with - retention_without

    return sorted(correlations.items(), key=lambda x: x[1], reverse=True)[:5]

How ML Predicts TTV and Identifies Risks?

Early prediction of stuck customers. The onboarding journey is a sequence of steps. ML predicts the probability that a customer will never reach the Aha-moment:

def predict_at_risk_onboarding(account_id, days_since_signup):
    events = get_product_events(account_id, days=days_since_signup)

    features = {
        'setup_completion_pct': events['setup_steps_completed'] / total_setup_steps,
        'integrations_connected': events['integrations_count'],
        'users_invited': events['team_members_invited'],
        'login_frequency': events['unique_login_days'] / days_since_signup,
        'support_tickets_opened': events['support_tickets'],
        'training_modules_completed': events['training_completion'],
        'days_since_signup': days_since_signup,
        'plan_tier': account['plan'],
        'company_size': account['employee_count']
    }

    risk_score = onboarding_risk_model.predict_proba([features])[0][1]
    return risk_score

Time horizon: Day 3, day 7, day 14 — checkpoints. If on day 7 the risk prediction > 0.6 → intervention trigger.

TTV Segmentation: Table of Averages

By use case: Cluster customers by their goal. Different onboarding paths for different clusters — personalized next-step recommendations.

Intervention Engine

Automated vs. Human interventions:

def select_intervention(account, risk_score, bottleneck):
    if risk_score > 0.8:
        # Critical — CSM intervenes manually
        return {
            'type': 'human',
            'action': 'schedule_call',
            'owner': assign_csm(account),
            'message_template': 'high_risk_outreach'
        }
    elif risk_score > 0.5 and bottleneck == 'integration':
        # Automated — in-app tooltip + email
        return {
            'type': 'automated',
            'channel': ['in_app_tooltip', 'email'],
            'content': 'integration_setup_guide',
            'timing': 'next_login'
        }
    else:
        return {'type': 'monitor', 'next_check': 3}  # days

A/B testing interventions:

• Control group vs. group with nudge
• Metric: TTV (days to Aha), 30-day activation rate
• Bayesian A/B test: stop when posterior probability > 95%

Cohort Analytics

TTV Cohort Chart: Classic visualization: X-axis — days since registration, Y-axis — % of customers who reached Aha. Compare cohorts (different months, channels, plans).

Bottleneck Analysis: Onboarding step funnel: where do most customers get stuck? Step with highest drop-off → priority for UX improvement.

def funnel_analysis(onboarding_steps, cohort):
    funnel_rates = {}
    for i, step in enumerate(onboarding_steps):
        users_reached = cohort[cohort['max_step_reached'] >= i]['count']
        users_completed = cohort[cohort['max_step_reached'] >= i+1]['count']
        funnel_rates[step] = users_completed / users_reached
    return funnel_rates

Median TTV by segment: Weekly monitoring. Rise in median TTV → problem in onboarding (product regression, ICP change).

Comparison of Intervention Types

What's Included in the Work

• Documentation: model card with features and metrics, pipeline diagrams, ML service architecture
• Access: TTV cohort chart and funnel analysis dashboards for Product & CSM teams
• Training: workshop on interpreting risks and configuring interventions
• Support: code review and model updates for new events during the warranty period

Integration

Product Analytics: Amplitude, Mixpanel, Heap — event sources. Warehouse: Snowflake/BigQuery — feature store for the model.

CRM: Salesforce Custom Object "Onboarding Progress" — visibility for CSM. Health score combined with TTV progress.

In-app Messaging: Intercom, Pendo, Appcues — delivery channels for automated interventions based on ML triggers.

Timeline: Aha-moment definition + TTV cohort analytics + basic risk scoring — 3-4 weeks. ML prediction of at-risk accounts + intervention engine + A/B test framework — 6-8 weeks.

Want to measure and reduce your product's TTV? Contact us — we'll help set up the tracking system for your stack. Order an audit of your current onboarding to assess TTV reduction potential. Get a consultation on implementation — we'll explain how to integrate ML tracking into your product analytics.

When does a time series forecasting model fail in production?

The CFO requests a quarterly sales forecast. An analyst builds SARIMA on three years of data, achieves MAPE 8.3% on the test set, and deploys. Two months later, the metric in production jumps to 23%. The root cause: the model was trained on pre‑COVID data, tested on a stable period, but production hit a promotion and supply chain disruption. Data leakage plus distribution shift—perfect notebook numbers, a broken forecast in reality. We have seen this pattern dozens of times across retail, fintech, and IoT. Our team has delivered more than 50 forecasting projects over 5+ years.

Incorrect cross-validation. Standard train_test_split for time series creates data leakage: the model sees future values during training. The correct approach is TimeSeriesSplit or walk‑forward validation with an expanding window.

Multiple seasonality. Hourly electricity consumption has three seasonalities: daily (24h), weekly (168h), yearly (8760h). SARIMA handles only one. Prophet can handle multiple but scales poorly to thousands of series.

Missing values and anomalies. A missing sensor reading is information (the sensor turned off), not NaN. Linear interpolation destroys this signal. Proper handling depends on the missingness mechanism.

Cold start. A new SKU in a 50,000‑item assortment has no history, yet a forecast is needed. Standard approaches fail; cross‑learning or feature‑based methods are required.

Why is model selection critical for your data?

Prophet (Meta) – a solid start for business data with clear seasonality and holidays. Fast setup, interpretable, built‑in outlier detection. Fails on irregular patterns and does not scale beyond ~10k series without parallelization.

Gradient boosting on features (LightGBM, XGBoost) – often underestimated. Engineer lags (t‑1, t‑7, t‑28), rolling means, day‑of‑week, holidays. The model trains on all series simultaneously, solving cold start via transfer learning. MAPE in retail often beats neural nets with proper feature engineering.

TFT (Temporal Fusion Transformer) – a transformer designed for interpretable forecasting with covariates. Built‑in variable selection, temporal attention, quantile outputs. Available in pytorch‑forecasting. Requires ~10,000+ records per series for stable training.

PatchTST – splits the series into patches (like ViT for images), capturing local patterns better than classic transformers. Excellent for long‑horizon forecasting (96–720 steps ahead).

N‑HiTS, N‑BEATS – attention‑free neural architectures, faster than TFT, competitive accuracy. N‑BEATS won the M4/M5 benchmarks for tasks without covariates.

How do we deploy TFT in production?

A typical pipeline via pytorch‑forecasting:

training = TimeSeriesDataSet(
    data,
    time_idx="time_idx",
    target="sales",
    group_ids=["store", "sku"],
    min_encoder_length=max_encoder_length // 2,
    max_encoder_length=max_encoder_length,  # 120 days
    min_prediction_length=1,
    max_prediction_length=max_prediction_length,  # 28 days
    static_categoricals=["store_type", "category"],
    time_varying_known_reals=["price", "promo_flag"],
    time_varying_unknown_reals=["sales"],
    target_normalizer=GroupNormalizer(groups=["store", "sku"], transformation="softplus"),
)

A common mistake: the default target_normalizer (StandardScaler) breaks predictions for series with zero values (no sales on weekends). GroupNormalizer with transformation="softplus" is the correct choice for count data.

Case study: retail demand forecasting

A chain of 120 stores, 8,000 SKUs, 28‑day forecast horizon. The original system: SARIMA per series, MAPE 18.4%, retraining cycle – 6 hours. We replaced it with TFT on PyTorch + pytorch‑forecasting: a single model for all series, MAPE 11.2%, retraining – 40 minutes on an A10G. Feature importance via variable selection revealed that day_before_holiday influences more than the holiday date itself. Annual savings on inference alone exceeded $50,000.

Step‑by‑step configuration

1. Data collection and preparation. Handle missing values (mark NaN, interpolate only for technical failures), aggregate to required frequency, engineer covariates (holidays, promotions, prices).
2. Create TimeSeriesDataSet. Set group_ids (store + SKU), time index, forecast horizon. Choose target_normalizer based on target distribution.
3. Train a baseline. Prophet or LightGBM first – to understand complexity.
4. Train TFT. Use TemporalFusionTransformer with loss=QuantileLoss(), tune learning rate and hidden layer sizes.
5. Validate and interpret. Walk‑forward test, analyze variable selection, build attention heatmaps.

How to properly evaluate forecast quality?

RMSE alone is misleading – it over‑penalizes large values. Our standard set:

• MAPE – interpretable, unstable near zero.
• sMAPE – symmetric, avoids division by small numbers.
• MASE (Mean Absolute Scaled Error) – normalized relative to a naive seasonal forecast, ideal for comparing series of different scales.
• Pinball loss – for probabilistic forecasting, inventory management.

We guarantee a model card with these metrics on the validation set and walk‑forward results on at least 6 months of history.

What deliverables do you receive?

• Documentation of chosen architecture and hyperparameter rationale.
• Reproducible training and inference pipeline (Docker + CI/CD + Airflow/Prefect).
• Committed code with unit tests for key components.
• Team training: retraining, output interpretation, deployment of new versions.
• 3 months of post‑delivery support (consultations, bug fixes, fine‑tuning).

The model is deployed via FastAPI or Triton Inference Server. Retraining is scheduled (e.g., weekly) via Airflow with drift validation and automatic rollback if metrics deteriorate.

Process and timeline

We start with EDA: visualization, ADF test, STL decomposition, analysis of missing values and outliers. This takes 2–3 days but often reveals systemic data issues that block forecasting. Then we build a baseline (naive seasonal, Prophet), engineer features for LightGBM, and select a neural architecture if needed. Walk‑forward validation with a realistic horizon. Deployment via API with automatic retraining scheduled via Airflow or Prefect.

Timeline: MVP forecast on one data type – 3–6 weeks. Hierarchical forecasting system with automation – 2–5 months. Cost is calculated individually based on data volume, number of series, and required accuracy.

Our team consists of certified ML engineers (AWS ML Specialty, GCP Professional ML Engineer) with 5+ years on the market and over 50 completed forecasting projects. Contact us for a free analysis of your data – we will assess the task and provide initial recommendations within 1–2 days. Request a consultation to ensure your forecasts work in production, not just in a notebook.