Maxwell Social Infrastructure

Attribution Foundation

Closing the UTM Gap From ~36% to Full Coverage

Two-thirds of ad-driven conversions were invisible to the analytics layer.

The Challenge
As ad spend scaled into late 2025, the attribution gap widened: Google Ads was reporting 118 conversions while the Hex analytics layer was only seeing 43 — a ~64% gap. UTMs weren't sticky across the Maxwell Social site, Reddit Ads and LinkedIn Ads pixels only fired when their specific URL parameters were present, and form submissions on a separate domain weren't carrying tracking data through. Without reliable attribution, the ad program couldn't be optimized, the channel mix couldn't be defended, and budget decisions were being made on incomplete data.

The Matchbox Solution
We rebuilt the entire client-side tracking stack across the Maxwell Social web property and the Tripleseat booking flow, making UTMs sticky session-wide, repairing pixel firing logic, and ensuring tracking persisted through the cross-domain form submission.

• Made UTM parameters **sticky session-wide** across the Maxwell Social site, eliminating mid-funnel attribution loss as users browsed before submitting.
• Updated **all internal button URLs** to dynamically append UTMs when linking to the booking form on a separate domain.
• Reconfigured **Reddit Ads and LinkedIn Ads pixels** to fire all events regardless of URL parameter presence, fixing silent attribution drop-off.
• Rebuilt LinkedIn Ads and Reddit Ads as **conversion-optimized campaigns** bidding on actual signups rather than awareness clicks.
• Validated the fix with **41.7% Google Ads attribution** in post-fix data, matching the independent statistical model within ~1 percentage point.
• Rebuilt tracking as a coherent layer — **sticky UTMs, cross-domain handoff, reconfigured pixel firing, and platform-native conversion campaigns** — rather than as a patchwork of partial fixes, so attribution holds end-to-end going forward and every channel optimizes on real conversion data.

‍

Statistical Attribution

Modeling True Ad Impact Before the Tracking Was Reliable

Statistical inference where UTM data couldn't yet be trusted.

The Challenge
With UTM data unreliable for most of 2025, Maxwell Social couldn't answer a fundamental question: how much of inbound booking demand was actually being driven by ads? Reporting only what UTM captured would understate true impact and risk budget cuts on a program that was working. The team needed a defensible answer before the tracking fix could prove it directly.

The Matchbox Solution
We ran a linear regression on 66 days of paired daily ad-spend and daily-booking-request data, modeling true ad contribution and then validating the model against the post-fix UTM data once it became available.

• Built a **two-variable linear regression** across **66 days of ad spend versus daily booking requests**, isolating ads' incremental effect from organic baseline demand.
• Achieved **p < 0.0001 statistical significance** with a 0.49 correlation coefficient, confirming the relationship was not random noise
• Modeled ads as driving **~43% of inbound booking requests**, with the model later validated by post-fix UTM data showing **41.7%** — matching within ~1.3 percentage points.
• Used the validated model to recover an otherwise-invisible **multi-million-dollar ad-attributed pipeline figure** that broken UTM tracking had been failing to count.
• Held the model to an **empirical-validation test** — once UTM tracking was fixed, the regression's 43% prediction was checked against the actual 41.7% and survived to within ~1.3 percentage points, confirming the model's predictive power rather than relying on directional confidence alone.

‍

Lead Operations

AI-Scored Leads and Automated Sales Sequences

Every inquiry triaged, scored, and nurtured before it reached a salesperson.

The Challenge
Maxwell Social's sales process couldn't keep up with ad-driven lead volume. Inquiries flowed into a single sales inbox, follow-ups were manual and inconsistent, and there was no objective way to tell a low-budget casual inquiry apart from a Fortune 500 corporate booking before the sales team had spent time on both. As paid volume grew, the bottleneck shifted from generating leads to triaging and nurturing them.

The Matchbox Solution
We designed and deployed a HubSpot-based lead operations system — capturing every form submission with its full context, scoring it with a multivariable AI model, enriching it with HubSpot Breeze, and routing each lead into an automated outbound sequence calibrated to its priority tier.

• Built a **multivariable AI lead scoring model** that classifies every submission as Unqualified, Qualified, or High Priority, with **captured reasoning** stored alongside each score for ongoing model refinement.
• Negotiated and integrated **HubSpot Breeze enrichment credits** to automatically clean and enrich every new contact at zero marginal cost.
• Deployed **automated nurture sequences sent from the sales rep's own Gmail outbox**, driving a **~50% meeting-booking rate**, **80–90% open rates**, and **70% reply rates**.
• Tiered the workflow so **High Priority leads received priority booking links**, Qualified leads received standard sequences, and Unqualified leads stayed out of the sales team's queue entirely.
• Engineered the lead-ops system to compound — **every scored lead refines the AI model, every enriched contact deepens segmentation accuracy, and every booked meeting feeds new signal** back to the ad platforms upstream.

‍

‍

Closed-Loop Attribution

Connecting HubSpot, Tripleseat, and the Ad Platforms

Deal status flowing back to the platforms that created the lead in the first place.

The Challenge
Even with attribution fixed and lead scoring live, the data path stopped at the meeting booking stage. Once leads moved into Tripleseat — Maxwell Social's venue CRM — for sales follow-up, their downstream deal status was disconnected from HubSpot and from the ad platforms that originally surfaced them. The ad platforms still didn't know which leads actually turned into revenue, which meant they couldn't be trained to find more like them.

The Matchbox Solution
We built a closed-loop attribution architecture mapping Tripleseat deal stages to HubSpot deal statuses, then pushed the qualified-lead and deal-value signal back to Google Ads, LinkedIn Ads, and Reddit Ads as new conversion events for the platforms to optimize against.

• Mapped **Tripleseat deal stages to HubSpot deal status** via custom integration, eliminating the manual reconciliation gap between sales operations and marketing attribution.
• Pushed **MQL conversion signal and deal value** back to Google Ads, LinkedIn Ads, and Reddit Ads, letting the platforms optimize for qualified-lead generation rather than raw form fills.
• Fed **retargeting and lookalike audiences** automatically from HubSpot back into the ad platforms based on lead-scoring outcomes, compounding acquisition efficiency over time.
• Layered **server-side tracking on top of client-side pixels** for redundancy and durable closed-loop reporting across all three ad platforms.
• Connected every layer of the revenue stack — **ad platform → form → HubSpot scoring → automated nurture → meeting booked → Tripleseat deal data → back to ad platform** — so paid spend optimizes against real sales outcomes rather than upstream proxies.

‍

‍

‍