Marketing Analytics Agent AI

Invoked when: daily data refresh cycle runs — all ad platform APIs and the LOS funded loan feed are queried and dashboards updated

• –Calculates cost-per-funded-loan (CAC) for each active marketing channel — Google Search, Display, Meta, YouTube, organic search, and referral — by dividing the channel spend in the period by the funded loans attributed to that channel within the attribution window. CAC is calculated at the channel level, the campaign level, and the ad set level, providing a full decomposition of where within each channel funded loan volume is being generated efficiently and where it is not.
• –Calculates return on marketing investment (ROMI) per channel by comparing the net revenue from funded loans attributed to each channel against the channel spend. Net revenue uses the first-year interest income from the funded loan as the revenue proxy — a gross figure that excludes credit loss but gives a consistent basis for channel-level ROMI comparison. ROMI is presented alongside CAC on the daily dashboard.
• –Maintains a 4-week rolling average for CAC and ROMI per channel, providing the context for daily fluctuations. A single day's CAC spike is often noise; a 4-week rolling average that is trending upward is a signal. The dashboard presents both to prevent single-day anomalies from triggering unnecessary interventions.

Invoked when: weekly attribution model run cycles — all funded loans from the prior 7 days are traced back through their multi-touch journey

• –Traces every funded loan back through the borrower's pre-application touchpoints within a 14-day lookback window — the sequence of marketing exposures that preceded the loan application. Touchpoint data is assembled from UTM parameters in the LOS, CRM interaction records, and ad platform impression logs. Where the complete journey cannot be reconstructed, the funded loan is marked as unattributed and excluded from the attribution model for that cycle.
• –Applies a data-driven attribution model that distributes credit across touchpoints in proportion to each touchpoint's estimated contribution to the conversion — based on the statistical relationship between touchpoint presence and conversion rate across the historical dataset. The model is updated quarterly as the funded loan dataset grows.
• –Produces the attributed funded loan share by channel — the percentage of each week's funded loans that each channel receives credit for under the model. Attributed share is the input for the board pack's channel contribution narrative, particularly for upper-funnel channels whose contribution is undervalued by last-click attribution.

Invoked when: monthly board pack assembly cycle runs — triggered 7 days before the board meeting to allow 2 days for CMO review

• –Assembles the monthly board marketing pack: executive summary (funded loans acquired via marketing, total spend, blended CAC, and ROMI vs prior month and vs target), channel performance table (CAC, ROMI, funded loan volume and share by channel), attribution model output (data-driven vs last-click for comparison), top-performing and underperforming campaigns with the performance driver for each, and the marketing head's forward agenda framework.
• –Formats the pack for board presentation — not as a data appendix but as a narrative with data support. Each section leads with the key finding in plain English, followed by the supporting data. A board that receives a table of CAC figures without interpretive narrative will spend the meeting asking what the numbers mean rather than deciding what to do. The agent produces the narrative structure; the CMO reviews and adds strategic commentary.
• –Clearly states the attribution model, the lookback window, and the unattributed loan percentage in the pack header — so that board members who want to interrogate the methodology can do so, and so that the data is interpreted in the context of its limitations.