In your opinion, what are the most important factors to consider when designing a data pipeline for scalability?