What Caused the Changein my Website's Performance
Traditional analytics tells you what changed. Gorilla Works uses causal inference modeling to tell you what your campaigns caused — isolating true incremental lift from noise, seasonality, and coincidence.
From Correlation to Causation
Traditional analytics answers one question: "What happened after we sent that campaign?" That's a useful starting point, but it leaves the most important question unanswered: "Would it have happened anyway?"
Gorilla Works uses causal inference modeling to answer that harder question. By reconstructing what would have happened to your website traffic, search visibility, and conversions without the campaign, we isolate the true incremental lift your marketing actually caused. The result is a number you can act on — not a correlation you have to guess at.
Counterfactual Reasoning
Models what would have happened without your campaign to isolate true lift
Multi-Signal Fusion
Combines traffic, search, ad, and mail data into a single coherent picture
Confounder Control
Filters out seasonal trends, algorithm updates, and concurrent campaigns
Four Integrations. One Coherent Picture.
Each integration contributes a distinct class of signals that, in combination, allow the model to triangulate causation rather than inferring it from a single stream.
Google Analytics
Signals Ingested
- Sessions and unique users
- Engaged session rate
- Bounce rate and depth
- Conversion events and goals
- Traffic source breakdown
What It Powers
- Baseline traffic modelling
- Post-campaign lift measurement
- Engagement quality scoring
- Conversion attribution window
Google Search Console
Signals Ingested
- Organic impressions by query
- Click-through rate (CTR)
- Average position per keyword
- Page-level performance data
What It Powers
- Organic visibility lift detection
- Branded vs. non-branded search split
- CTR shift attribution
- Search demand correlation
Google Ads
Signals Ingested
- Ad spend by campaign
- Impressions and clicks
- Conversion tracking events
- Campaign dates and pauses
What It Powers
- Paid / organic interaction modelling
- Budget-change confounder control
- Cross-channel interference detection
- Incremental ROAS estimation
Direct Mail
Signals Ingested
- Mail drop dates per campaign
- Estimated delivery windows
- Piece volume and geography
- Response rates (where tracked)
What It Powers
- Mail-to-web response window anchoring
- Geographic lift segmentation
- Multi-wave campaign modelling
- Offline-to-online bridge attribution
Six Numbers You Can Act On
Each is a confidence-scored estimate of the true incremental effect your campaign had — not a raw before/after comparison.
Attributed Revenue Lift
Estimated incremental revenue generated by the campaign, holding all other factors constant. Derived from conversion data and historical revenue-per-session baselines.
Incremental Traffic
The number of sessions caused by the campaign above the counterfactual baseline — what your traffic would have looked like if the mail had never dropped.
CTR Impact
Whether the campaign produced a measurable shift in how often people click your listing when it appears in organic search — a leading indicator of brand recall.
Cross-Channel Interference
Detection of concurrent campaigns that may inflate or suppress results. When two campaigns overlap, the model flags the interaction and adjusts confidence scores accordingly.
Baseline Drift
Slow, structural changes to your organic performance unrelated to any single campaign — algorithm shifts, seasonality, or brand-equity growth — separated from campaign effects.
Direct Mail Lift
The share of observed web lift attributable specifically to the physical mail piece, controlling for paid search activity and seasonal trends active during the same window.
The Correlation Trap
The problem with last-click
Last-click attribution assigns 100% credit to the final touchpoint before a conversion — usually a branded search triggered by someone who already decided to buy. Multi-touch models spread credit across channels but still assume every observed touchpoint contributed. Neither approach asks the critical question: would this conversion have happened anyway?
The counterfactual approach
Causal modeling builds a prediction of the world without your campaign — the counterfactual baseline — and compares actual outcomes against it. The gap between what happened and what the model predicted would have happened is the campaign's true causal effect. This is the same logic used in randomized controlled trials, applied to your observational data without requiring a physical control group.
Why it changes decisions
If you're measuring correlation, you'll keep spending on channels that look good but are actually riding organic growth. If you're measuring causation, you know which spend is actually moving the needle — and you can allocate your next budget accordingly.
Specific problems causal modeling solves
From Raw Data to a Decision
Every insight in Gorilla Works follows the same pipeline from raw signal to an attribution score you can act on.
Data Collection
Your connected integrations continuously sync signals from Google Analytics, Search Console, Google Ads, and your direct mail campaigns. Data is normalised, cleaned, and aligned to a common timeline.
Counterfactual Construction
Before each campaign window is analysed, the model builds a prediction of what your metrics would have looked like in the absence of any campaign activity. This prediction controls for known confounders: seasonality, trend, and concurrent ad spend.
Causal Estimation
The model compares actual observed performance against the counterfactual baseline across all integrated data sources. Each outcome — traffic lift, CTR change, conversion impact — is assigned a point estimate and a confidence score.
Confidence Scoring
Results are classified into confidence bands (Strong, Solid, Suggestive, Inconclusive) based on the size and consistency of the estimated effect relative to baseline variance. You always know how much to trust each number.
Actionable Insights
Scores and lift estimates appear in your Campaign Attribution dashboard. You can compare campaigns head-to-head, track performance trends over time, and use the data to plan your next send with confidence rather than guesswork.
The bottom line
Gorilla Works doesn't just tell you what your numbers are — it tells you which numbers you caused. Every campaign you run feeds the model, sharpening its baseline and making the next attribution more precise. The longer you're on the platform, the clearer your edge becomes.