The oil & gas business has seen its fair share of software vendors who claim magical results delivered via a black box approach. Pick a geoscience or engineering discipline and you’ll likely find a piece of technology or software that offers a compelling solution, but the methods that drive output and analysis are intentionally hidden from view. To a trained scientist or engineer, a black box software solution is unacceptable since it is imperative to know what assumptions govern business critical decisions. And there are no bigger marketers of black box solutions than many of today’s artificial intelligence and machine learning vendors.
Novi is challenging the oil & gas market to stop accepting the black box approach and start demanding more of software vendors. With the latest release of Novi’s AI-powered well planning software, the company is taking a giant leap forward in bringing transparency to predictive analytics. In a marketplace full of claims about AI but devoid of actual proof of how it works, Novi is the first to provide users with powerful tools to explore, visualize, and validate Novi’s methods for predicting well performance.
Novi version 19.3 offers a suite of new capabilities designed to give users deep insight into the model and its predictions, including:
- Explain a Prediction – At a glance see the features driving production for each prediction. Each contributing feature shows whether it positively or negatively affected the prediction, provided in units of barrels or mcf.
- Contributing Wells – Quickly view from the model’s perspective which wells are the largest contributors to the prediction. Visualize the similarity scores, weights, and impacts in map view and pie charts.
- Model Accuracy – Verify model performance by comparing model predictions vs. actual results. Filter the predictions to specific operators, well spacing, or any grouping of interest.
For example, want to know how Novi predicted 163,884 bbl. of oil for the well below at IP180? Just click “Explain this Prediction” for a detailed summary that shows the features that drove the prediction up from the model average of 85,309 bbl. with proppant/ft. as the most influential feature. You can also change initial production to see how the prediction evolved over time, from day 30 to 720.
In addition, Novi’s latest software release introduces Novi Cloud. All data used to build a model and all output data from Novi is hosted on the Novi Cloud. The data store is easily connected to analysis tools such as TIBCO Spotfire. Spotfire users also benefit from a Novi-built Insights Dashboard connected to the Novi cloud that highlights key results and use cases for model outputs, such as identifying points of diminishing returns for completions intensity, detecting underdeveloped acreage, and interrogating model perspective on spacing. Just another example of how Novi is setting the standard for transparency.
Want to see Novi’s new prediction explanations in action? Contact Novi today and we’ll walk you through all of these new product features.