Monty Python’s much-loved Graham Chapman used to cut some of the comedy troup’s zanier sketches short by jumping into scenes dressed as a British Army colonel with full military uniform regalia and a customary baton tucked under his arm.
“Stop this now, it’s far too silly! Move along now,” was The Colonel’s time-honoured line which he employed every time to the same superb comic effect.
If he were still with us, Chapman would no doubt be casting his eyes over the current Artificial Intelligence (AI) renaissance and the new algorithmically-enriched innovations the IT industry is driving into the creation of the new universe of Machine Learning (ML) engines.
One things is for sure, The Colonel would not be keen on silly ML, he’d prefer SensibleML surely.
Sensible Machine Learning (ML)
A specialist in so-called Corporate Performance Management (CPM) – which may be another term for plain old Business Intelligence (BI) or even plainer management consultancy reborn for the digital age – OneStream is a Michigan USA headquartered firm now working to expand its ML competencies.
The company’s amusingly-named Sensible ML is a software service that aims to increase ML forecasting transparency and explainability of key ‘business driver’ variables.
This is essentially a time-series ML product that integrates with existing planning processes to drive improved forecasting accuracy and efficiency.
Why time-series makes sense
With time-series data being that part of the total information stream where we tag, label, list, classify and denote any given datum, set of data, wider dataset or some other information form factor by its core value and its time stamp, it makes logical sense to focus on temporal time-based data for ML applications that are engineered for business goal forecasting.
So this, as per the above-suggested technology validation, is what OneStream appears to have done.
The technology’s newly expanded capabilities are said to provide additional transparency and trust for users while increasing agility, time to value and enterprise value associated with operational planning processes.
Planning alignment imperatives
The company says that this expanded release addresses OneStream customers’ rapidly expanding need to align financial planning with operational planning processes such as Sales Planning, Demand Planning and Sales & Operations Planning (S&OP). OneStream customers with existing planning solutions can use Sensible ML to drive improved forecasting accuracy and efficiency while seamlessly integrating it into their planning, reporting and dashboard processes.
“OneStream views the opportunity for AI no differently from core CPM processes – AI must be unified to core processes if it’s going succeed,” said Tom Shea, CEO at OneStream. “Traditional CPM tools offer the same challenges they always do. They are fragmented, require data movement and maintenance behind the scenes and do not contain actual ML modeling capabilities.”
Shea says that his firm’s expanded Sensible ML release allows users to align financial and operational planning across the enterprise, with enhanced capabilities for using external data to further drive transparency and understanding of key business drivers while seamlessly integrating it into their existing CPM processes.
Sensible ML helps organisations evolve enterprise planning processes by making ML forecasting easy and intuitive for FP&A teams and operational analysts. Early adopting customers highlight the speed to value and increased accuracy of sales and demand forecasting with Sensible ML, making it a standout solution among competitors.
Back-testing creates transparency
“OneStream is at the forefront in applying AI to provide a sophisticated but usable methodology to improve forecasting and planning,” said Ventana Research analyst Robert Kugel. “Its approach looks promising, especially in its ability to establish trust and demonstrating a lack of bias in recommendations, which are essential to acceptance of this technology. Transparency is a must so, for example, OneStream provides built-in techniques such as back-testing models to demonstrate to users how well a model would have performed in the past.”
Key capabilities of the expanded Sensible ML release include a feature library, which enables users to automatically graft external data sources into a Sensible ML project, including (and these are the somewhat esoteric sounding examples that the company has provided) maritime container indices, Covid-19 case rates, interest rates, consumer price index (CPI0 figures, stock prices, weather etc.
Users can quantify how influential every feature is during training and production for every target and model. Then, for prediction explanations, users can see that for any target, model, forecast date, they can quantify how much each feature negatively or positively influences the forecast from the model’s average forecast.
Views & visualisations
New data visualisations and advanced data views are hoped to provide deeper insights and transparency into forecast drivers to help improve model confidence, accuracy and explainability.
Key themes emanating from this story include explainable AI and the whole gamut of AI bias and practical real-world usefulness, the use of time-series data which aligns closely to real-time data and end-of-batch data processing, the use of citizen-data-scientist tools and the wider use of unified data management.
Come to think of it, OneStream doesn’t sound silly at all.
Image credit: Adrian Bridgwater – Pink pickled eggs, don’t be silly