I learned a lot with Héctor Parra, a Google engineer, about how to leverage data and Machine Learning for business when I interviewed him a few days ago. I summarise below my takeaways from that conversation.

Start the project

It is of paramount importance to explain the value of data for the business to align all the stakeholders and set goals.

Then, we must assess the situation, including:

• Objectives,
• What data is available,
• Skills,
• Issues – E.g., silos, lack of skills.

Before starting the project, we must plan, architect and clearly communicate the plan:

1. How we plan to keep data quality,
2. Load all data in one single place (Data Warehouse, Data Lake),
3. Normalise data to combine in a stack, and
4. Analyse and combine for final consumption.

Data Science

Many companies don’t want a Data Scientist team because they think they would need it for just one project. Over time they realise that Data Science is a long-term investment.

Machine Learning

If we need to implement a Machine Learning model, let’s not reinvent the wheel – We have half-programmed or pre-trained ML models.

Even when re-using a pre-trained model, an ML project is complex. For example, the Integration (especially if we need to achieve near real-time) or API management.

Forecasting, which is what Machine Learning does, is an extension of studying the past:

1. We need enough data,
2. Quality data,
3. Real-world instability reduces the effectiveness of historical data (we can’t predict what never happened), and
4. We must weigh instability and normality. Then, we can merge both in your model.

As we have already mentioned, instability makes historical data lose its effectiveness. The consequence is that we must make Short-term predictions when it happens.

Goals

The key to a successful project is how to set up the objectives. Choose your battles wisely:

• What to predict,
• What for, and
• How to reduce bias

To better choose where to focus our efforts, we must consider:

• Type of company,
  • Current situation,
  • Latest changes

So analytics let us gain the insights we need about our organisation to find opportunities, identify risks, and set priorities.