Challenge
A leading SaaS platform operating in a competitive and volatile market sought to improve how marketing performance was measured and forecasted. Existing analytics focused primarily on short-term subscription response, which made it difficult to distinguish temporary fluctuations from meaningful changes in underlying demand.
As a result:
-
Short-term performance volatility triggered reactive budget decisions
-
Upper-funnel and brand investments appeared inefficient
-
Pricing changes showed little short-term impact despite concerns about long-term demand erosion
-
Marketing and Finance lacked a shared, forward-looking view of growth drivers
The client needed a measurement framework that could simultaneously support near-term decision making while capturing the long-term effects of brand, pricing, and structural demand shifts.
Solution
Leveraging its proprietary multi-layered Dynamic modeling methodolgy, Marketscience developed an integrated, multi-horizon modeling framework designed to manage both short-term volatility and long-term growth.
Short-term forecasting and anomaly detection
A Dynamic Unobserved Components Model (UCM) was constructed to separate incremental sales from the evolving baseline while continuously generating expected Gross New Subscriber (GNS) forecasts at weekly and quarterly intervals.
- The model ingested weekly sales, media, and macroeconomic inputs.
- “Baseline” was treated as a living series that could deteriorate without sustained marketing support.
- Forecasted vs. actual GNS was tracked to measure deviations, forming an automated early-warning system for performance anomalies.
Long-term demand and brand modeling
To capture structural evolution in the client’s base sales, Marketscience applied a Vector Error Correction Model (VECM) explicitly linking brand metrics, pricing, and macro factors to long-run sales trends. This allowed the client to quantify cumulative brand effects, pricing elasticity, and the persistence of marketing impact beyond short-term campaign windows.
By integrating both views, the framework provided a single, consistent understanding of marketing effectiveness across time horizons.
Key Insights
The integrated analysis revealed several critical findings that were not visible in short-term models alone:
-
Including long-term effects increased total media-attributed volume by approximately 10-15% versus short-term MMM alone
-
This surfaced approximately 2,500-4,000 incremental subscriptions, representing $30-50M in additional lifetime revenue that had previously gone unattributed
-
Upper-funnel media, particularly television, was materially undervalued. While TV initially accounted for only 3-4% of short-term sales, long-term modeling increased its total contribution by 25%+, adding $10-15M in incremental lifetime value over a multi-year horizon
-
Pricing actions that appeared neutral in short-term analyses were shown to reduce long-term base demand by close to 10%, equating to a $50M+ lifetime revenue impact
The anomaly detection layer further clarified when short-term deviations reflected execution issues versus genuine structural shifts, allowing the client teams to respond to true performance issues without overcorrecting for short-term noise.
Business Impact
By explicitly modeling both short-term performance deviations and long-term demand dynamics, the client gained:
-
A more complete and accurate view of marketing value creation
-
Improved forecast stability and confidence in budget decisions
-
Better alignment between Marketing, Finance, and Strategy teams
-
The ability to rebalance investment toward channels driving both near-term acquisition and sustained growth
The result was a shift from reactive optimization to predictive, multi-horizon growth management without increasing total marketing spend.
What It Matters
In SaaS and other subscription-based markets, sustainable growth depends on managing immediate performance while continuously building future demand. Measurement systems that focus on only one time horizon risk misattributing value and distorting decisions.
By integrating anomaly detection with long-term demand modeling, Marketscience enabled a unified, forward-looking view of marketing effectiveness that supported both short-term control and long-term growth planning.
