Parking Revenue Optimization for a Major Entertainment Venue

The situation

A multi-venue entertainment complex hosts concerts, sporting events, trade shows, and conventions across three distinct venues. The complex operates nine color-coded parking lots, each serving different areas and price points. Parking pricing had been set manually based on general event size and type, without a systematic way to predict demand per lot or to optimize prices against that demand.

The challenge was structural. Demand for any given lot depends on which venue is hosting, the type of event, expected attendance, day of week, time of day, and the prices of neighboring lots. Those variables interact: raising the price of one lot shifts demand to adjacent lots. Manual pricing could not account for those cross-lot dynamics, and the venue had no historical analysis connecting pricing decisions to revenue outcomes.

What we built

CorrDyn started with a data engineering effort to consolidate the venue’s event and parking data. Ticketing platform exports, parking system records, and internal event scheduling data were brought into a unified pipeline using dlt for ingestion and dbt for transformation in Databricks. The result was a clean, event-level dataset linking attendance, ticket sales, parking transactions, and pricing history across all nine lots and 136 historical events.

On that foundation, we built a demand prediction model for each lot using Random Forests. The algorithm was selected for its ability to handle non-linear feature interactions with a small training set while resisting overfitting. The model takes event characteristics and lot prices as inputs and predicts how many spots each lot will sell.

The pricing optimization layer sits on top of demand prediction. It varies prices within operator-defined bounds, predicts the resulting demand per lot, calculates total revenue, and iterates to find the combination that maximizes revenue across all lots simultaneously. The output is a recommended price per lot per event, delivered as a CSV that operators can review and adjust before applying.

We also built a timing analysis showing when parking purchases spike relative to event start times. The data consistently showed that the 2 to 4 hour window before an event’s start is when most purchasing happens, giving operators a concrete window for when price changes have the most impact.

What changed

Venue operators now receive data-backed price recommendations for every event instead of relying on rules of thumb. The feature importance analysis revealed which variables actually drive demand in each lot, replacing assumptions with evidence. The timing analysis gave the operations team a specific window for price adjustments rather than guessing when to act.

The model continues to learn. Each new event adds training data, and the demand predictions sharpen as the dataset grows. CorrDyn also built Ticketmaster reporting overlays and a parking data pipeline that feed the venue’s broader analytics efforts beyond pricing alone.