1. AI Model Overview & Data Inputs

Q: What data sources and fields are used as inputs into RevSure’s ML model?
A: RevSure’s ML engine ingests data from the entire GTM tech stack, including CRM, marketing automation, sales automation, website analytics, ad platforms, and ABM vendors. These systems power the GTM Data Graph™, enabling a unified view of the customer journey across all touchpoints.

Q: Does the model account for touch sequence, time decay, or velocity?
A: Yes. The model incorporates:

Recency, frequency, and sequence of touches
Age and stage duration of leads, accounts, and opportunities
Time until quarter end
Campaign-level attributes like spend, source, team, channel, pipeline value, etc.
LinkedIn engagement (if connected)

Q: How does the model manage missing or inconsistent data?
A: RevSure uses a combination of ML, heuristics, and data enrichment for robust handling of incomplete or noisy data, including imputation techniques and flexible sample-size handling.

2. How RevSure AI Compares to Other Models

Q: How should we think about RevSure’s AI model alongside traditional attribution models?
A: The AI model:

Is ideal for multi-touch attribution, outperforming rules-based models (e.g., U-shape, linear) by assessing each touch’s true contribution.
Considers counterfactual scenarios—e.g., "what if this touch didn’t happen?"
Yields more holistic and accurate outputs, particularly when journey depth increases.

Q: How do I interpret attribution differences across models?
A: Variance is expected. For example, the AI model may credit a mid-funnel campaign that gets underweighted in a U-shaped model. This is due to the AI model’s use of Markov chain probabilistic logic for conversion path analysis.

3. Model Accuracy, Validation & Benchmarking

Q: How does RevSure validate model performance?
A: Each model undergoes:

Rigorous backtesting with metrics like F1 Score, MAPE, and cross-validation
Quarterly accuracy reports comparing day-level projections to actuals

Q: What kind of accuracy can we expect from RevSure models?
A: Typical accuracy ranges (as observed in Q1–2025) are:

Pipeline Volume: 81% – 93%
Pipeline Value: 86% – 95%
Closed Won Volume: 83% – 92%
Closed Won Value: 80% – 91%

Q: Do you have accuracy benchmarks ?
A: Yes:

Pipeline Gen Volume: 84.23%
Pipeline Gen Value: 86.68%
Booking Volume: 82.72%
Booking Value: 86.59%
Note: A funnel reconfiguration in March required model retraining, causing brief fluctuations

4. Interpretability & Attribution Transparency

Q: Can I see what factors influenced model predictions?
A: Yes. You can:

View feature importance and campaign weights in the Funnel Conversion Attribution (FCA) module
See top conversion paths for each stage (e.g., Visitor → MQL, Pre-Lead → Booking)
Use stage-specific configs in FCA that align with the filters used in AI Attribution for consistent comparisons

Q: Is there visibility into why a campaign was credited?
A: Yes. The model shows aggregate campaign contributions to conversions by name, channel, and type. Individual opportunity timelines can also be reviewed in-app.

5. Customization & Flexibility

Q: Can we customize the model to reflect our business logic?
A: Absolutely. RevSure supports:

Custom signal ingestion
Feature configuration to include or exclude attributes
Filters that align with specific views or models (e.g., stage-to-stage vs. generation attribution)

6. Limitations & Roadmap

Q: What are current limitations or known blind spots?
A:

Sudden changes in GTM or product strategy can temporarily reduce accuracy until the model adapts.
External macro events are factored in, but major internal GTM shifts may take time to reflect in projections.

Q: What improvements are planned?
A: The roadmap includes:

Continuous model performance monitoring
Feature expansion based on real-time data behavior
Enhanced interpretability and cohort-specific attribution logic

7. Additional Definitions & Clarifications

Q: What is the purpose of the "campaign channel" field?
A: It serves as an umbrella classification that aggregates campaigns across types, source systems, or content themes.

Q: How are campaign types determined?
A: These are collaboratively defined with your team and can be revised.

Examples of Campaign Type Classifications:

Direct – Webpage visits
SEARCH – Google Ads
Website – SFDC campaign types
Referrer – Redirected traffic
Organic Posts / Sponsored Updates – LinkedIn content
Interval – Outreach sequences

8. Practical Insights & Engagement Metrics

Q: How many campaign touches typically drive a successful opportunity?
A: As of the latest analysis, 6.64 campaign touches per opportunity is the observed average to reach gross pipeline (GP). Repeat touches are included in this count.

Q: Can I track or compare this behavior year-over-year?
A: Yes. Filters within the RevSure platform enable time-based comparisons, including cohort-specific touch analysis and campaign timelines.