The business has grown largely through acquisition. Each venue is run with a local-first approach, tailored to its community rather than following a single template.
About the role
The business needs a demand forecasting model for the venues it has acquired in the last 12 months. You will own that build, from approach through to production.
The core challenge is limited history. These venues have only a short trading record in the company's systems, and earlier data may sit in different point-of-sale systems or be incomplete. You will need to forecast reliably by drawing on patterns from comparable venues across the wider network, then improve the model as each site's own data builds up.
Your forecasts will feed directly into how venues roster staff, order stock and plan trading.
Key responsibilities
- Design and build a demand forecasting model for recently acquired venues, covering revenue, covers and product demand at daily and intraday levels
- Solve the cold-start problem by using data from comparable venues, hierarchical approaches and pre-trained time series models where they help
- Bring in external drivers such as weather, public holidays, school holidays, local events, sport and promotions
- Clean, align and integrate venue data from different point-of-sale and operational systems
- Deploy the model to production with automated retraining, monitoring and drift detection
- Set clear accuracy measures and report performance by venue, region and venue type
- Work with operations, finance and venue managers so forecasts are understood, trusted and used
- Document the approach so the model can be maintained and extended to future acquisitions
- 5+ years in data science, with demand or sales forecasting models built and run in production
- Strong Python and SQL
- Hands-on experience with forecasting and ML methods such as LightGBM, XGBoost, Prophet or ARIMA
- Experience forecasting with sparse, short or patchy history, such as new stores or sites
- Experience deploying and monitoring models on a cloud platform (Azure, AWS or Databricks)
- Ability to explain model outputs and limitations to non-technical stakeholders
- Tertiary qualification in data science, statistics, mathematics, engineering or a related field
- Hospitality, retail, QSR or other multi-site business experience
- Hierarchical forecasting (for example network, region, venue, product)
- Experience with foundation time series models such as Chronos or TimeGPT
- Databricks, PySpark and MLflow
- Weather and event-driven demand modelling
- Promotional uplift or causal impact analysis
Halcyon Knights is an equal opportunities employment agency, valuing the contributions made by all people and welcomes applications from people with disability, Aboriginal Australians, people of all ages and from culturally diverse backgrounds.
