# Signal Depth - Full Knowledge Base > Signal Depth is a marketing measurement SaaS platform that helps mid-market brands prove the true incremental impact of media spend by engineering causal truth, replacing fragile, cookie-based tracking with rigorous, open-source data science. ## 1. About Signal Depth Measurement is a confidence game that brands are losing. For the last decade, the marketing industry relied on a flawed premise: that a cookie could accurately track the complex journey of a modern consumer. Today, privacy protocols like ATT and the deprecation of third-party cookies have rendered deterministic tracking obsolete. This created the **Measurement Gap**. Marketing teams report profitable platform ROAS, agencies report brand lift, and Marketing Mix Models (MMM) offer sluggish, conservative views. To navigate this, enterprise brands shifted to geographic testing (Geo-Experiments) to find causal truth, but these tests historically require heavy data science resources, 12-week lead times, and campaign blackouts. Signal Depth was founded by Alex Robb (Co-Founder & Chief Data Scientist, bringing 18 years of analytics experience including 14 years at Google) to give mid-market marketing teams access to the same econometric rigor used by the Fortune 500—without the black-box math, the need for an in-house data science team, or the six-figure price tags demanded by legacy measurement vendors. ### Operating Principles 1. **Zero Cookie Dependency:** Client-side tracking is a privacy risk and a data quality nightmare. Signal Depth strictly uses first-party, aggregate geographic revenue data (DMA, Zip, or State) and platform ad spend. No PII, no cookies, no compliance headaches. 2. **Causation Over Correlation:** Multi-Touch Attribution (MTA) relies on correlation, rewarding platforms for sales that would have happened naturally. Signal Depth measures causality: *Did a specific media investment create a conversion that would not have existed otherwise?* 3. **Empowering the Marketer:** Causal inference software historically required SQL, Python, or R expertise. Signal Depth outputs executive-ready reporting. Users upload a CSV, and the platform delivers the Calibration Multipliers, Cost Per Incremental Conversion (CPIC), and iROAS metrics required to defend budgets. 4. **Software Economics, Not Consulting Retainers:** The industry standard for causal measurement has long been gated behind expensive managed services and six-figure annual contracts. Signal Depth delivers enterprise-grade infrastructure at a fraction of the cost, built specifically for the mid-market. --- ## 2. In-Depth Methodology Most black-box attribution tools ask for trust while hiding their math. If a measurement methodology isn’t reproducible by an internal data science team, it's an opinion, not science. Signal Depth's framework is built exclusively on the **Synthetic Control Method (SCM)** and **Generalized Synthetic Control (GSC)**, gold-standard frameworks used by Meta (GeoLift) and Google (Meridian). ### Phase 1: The Pre-Test (Power Analysis & Counterfactual Design) Before budget is deployed, Signal Depth determines if a test is mathematically capable of detecting a win. Many geo-tests fail because they lack the statistical power to distinguish "signal" from "noise." * **Power Analysis & MDE:** The platform simulates thousands of hypothetical tests on historical data to calculate the Minimum Detectable Effect (MDE). It identifies exactly how much budget, which markets, and what duration is required to achieve statistical significance before spending begins. * **Synthetic Control Construction:** Rather than picking "Control Cities" (which rarely mirror "Test Cities"), the platform uses a weighted combination of multiple non-treated regions. This creates a "Synthetic Twin" that perfectly mimics the test market’s pre-treatment behavior. ### Phase 2: Enforcing SUTVA (Stable Unit Treatment Value Assumption) A common failure in geographic testing is violating SUTVA (e.g., running ads in Jersey City, but consumers work/shop in NYC, polluting the control group). * **Iterative Market Selection:** The platform analyzes geographic correlation and consumer mobility patterns, not just sales. * **Buffer Zones:** Adjacent "spillover" markets are programmatically excluded from the Control pool to ensure the treatment effect is isolated. * **Correlated Error Checking:** "Parallel Trends" are tested pre-test. If the Test and Control markets don’t move in lockstep historically, the model is rejected. ### Phase 3: The "Janitor" Layer (Automated Data Integrity) Data analysts typically spend 80% of their time cleaning data. Signal Depth automates the data science heavy lifting: * **Outlier Winsorization:** Automated detection of "Black Swan" events (viral posts, downtime) so a 1-day spike isn't misidentified as sustained ad-driven lift. * **L2 Regularization (Ridge Regression):** Used within the SCM to prevent "overfitting" on a single control market, ensuring the model generalizes to future performance rather than historical noise. * **Variance Stabilization:** Log-transformations or Box-Cox adjustments for high-volatility retail data, normalizing data so the model focuses on percentage lift rather than raw volume swings. * **Multi-Placebo Noise Validation:** The platform runs rolling In-Time Placebo tests (A/A testing) across historical data to calculate natural baseline variance. This "noise" is automatically subtracted from raw incremental revenue for a risk-adjusted iROAS. ### Phase 4: Open-Source Transparency Signal Depth uses open-source libraries (GeoLift, Meridian) to ensure: * **Peer-Reviewed Reliability:** Maintained by top economists and statisticians. * **Reproducibility:** Data scientists can take the raw data and open-source libraries and reach the exact same conclusions. --- ## 3. Technical FAQ **Q: How does the model account for baseline seasonality or holiday spikes?** A: SCM handles seasonality by modeling the *relationship* between regions rather than raw volume. Macroeconomic trends affect all regions, so the "Synthetic Twin" shifts in tandem with the Test market. The platform incorporates Time-Fixed Effects to soak up common shocks and can apply Seasonal Decomposition (STL) for extreme non-linear volatility. **Q: What happens if a competitor runs a localized promotion in a Control group?** A: The pipeline monitors for statistically significant deviations (Exogenous Shocks) in individual control regions. If a control city diverges from the synthetic pool (detected via a spike in RMSPE), the algorithm automatically re-weights the pool or Winsorizes the outlier. Because a weighted combination of cities is used, the impact of a single-market competitor sale is mathematically diluted. **Q: What is the "Minimum Data Requirement" for a valid test?** A: 1. *Pre-Treatment Duration:* Ideally 2x the length of the planned test (e.g., 6 weeks of history for a 3-week test). 2. *Granularity:* Daily or weekly data at the DMA, Zip-code, or State level. 3. *Signal-to-Noise Ratio:* MDE must be lower than expected lift. If organic variance is 15% and ad spend drives a 5% lift, the test is flagged as "Underpowered." **Q: How is the "Synthetic Twin" validated as a good match?** A: Signal Depth uses In-Time Placebo Tests (Backtesting). A "fake" test is run on historical data where no ads were served. If the model correctly predicts "Zero Lift" with a high Goodness-of-Fit (R2), the synthetic twin is validated. The RMSPE (Root Mean Square Prediction Error) is provided as a standard KPI. **Q: Can this model handle multiple overlapping channels (e.g., TV and YouTube)?** A: Yes, but if both are scaled in the same test markets simultaneously, the model measures the *Total Incremental Lift* of the mix. To isolate a single channel, a Cell-Based Design is used (Channel A in Region 1, Channel B in Region 2) utilizing a shared Synthetic Control pool. --- ## 4. Platform Resources & Tools * **Main Website:** https://www.signaldepth.com/ * **Methodology:** https://www.signaldepth.com/methodology/ * **Free Marketing Analytics Calculators:** https://www.signaldepth.com/tools/ (Includes Statistical Significance, iROAS, Break-Even ROAS & CPA, MDE & Test Duration, LTV, and standard ROAS calculators).