Bison is an AI, machine learning, data science & analytics company that can optimise your operations, minimise your costs, maximise your profits and increase your revenue quickly with our products and services
Bison is a team that truly knows what it is doing. Their expertise in machine learning and Al provided us with valuable insights and effective solutions. They seamlessly integrated advanced algorithms and innovative approaches, helping us optimize our processes and achieve impressive results.
We highly recommend Bison for any challenging data science, machine learning or Al project
Ben Zyl
CEO, Waave
Implementing Bison’s Al-driven plan allocation solution has been a game-changer for our business. The accuracy and efficiency of their allocations have significantly improved our bottom line with some product savings getting close to 40%.
I highly recommended Bison to any organisation looking to leverage Al and ML for smarter planning and forecasting!
Daniel Duggan
CEO, Yomojo
Building a product? Ensuring the data aspects of your projects and products are rigorous, do exactly what you need them to do and have defensible integrity in the marketplace can be the thumb on the scale of success that we provide.
Have what looks like a numerical problem with no solution? Using our expertise in econometrics, optimisation and data science, we’ll get you across the line, either by bringing you up to speed with the techniques, products and partners available - or designing the solution for you.
Finding patterns in historical data with statistical modeling and machine learning to identify risks and opportunities is amazing - but knowing the potential impacts of the decisions and choices you make with that output can be the icing on the cake. We provide that.
You've got data, maybe a lot of data - but how do you know if it's even valuable and if so how do you go about selling it? Well, it's straight forward - we analyse it, we anonymise it, we leverage our networks to find buyers and you get paid.
Disparate data sources, multiple tools, spreadsheets for miles, databases left fallow, no actionable insights and missing metrics – wherever your reporting and data engineering is at, we’ll make it better, faster, more reliable and more useful – actionable insights are what we deliver.
Looking to build a data component of your organisation but not really sure how to evaluate candidates or what you really need? Well we’ll dive in with you, find the right people, bring them in and get them up and running so your organisation can fire on all data cylinders.
We scoped, prototyped and delivered a process to calculate subscription prices for hundreds of thousands of retail energy customers using various techniques including panel regressions, farthest neighbours, next-nearest neighbours, and imputations. With this approach we were highly successful as preliminary results from a large Midwestern [USA] utility that show customers saving 6% on their bills with subscription rates, as well as double-digit peak-demand usage reductions.
A PR company in the process of building a reputational risk score realised that, while they excelled in their domain, taking survey data and converting it to reputational risk score that was technically convincing was outside of their wheelhouse. Thus we came on board and used our product development and start-up experience to guide the scope and development. Applying various supervised machine learning techniques, such as k-means clustering and silhouette analysis, as well as sentiment analysis (which is computationally identifying and categorising opinions expressed in a piece of text) to the data, we transformed their idea into a robust, sellable product.
Telco resellers (MVNOs) can have tens or hundreds of thousands of customers on scores of different plans. These plans have wildly different configurations and all of this can change rather frequently. What this means is that invoices from the carriers to MVNOs can be extremely difficult to audit and are frequently incorrect. However, for our MVNO clients, we are often able generate what the invoice should be weeks before the invoice arrives, enabling much faster financial reporting. We have also lowered the original invoice amounts by up to 25% a year in some cases by successfully identifying and demonstrating errors to carriers.
A furniture company was finding it difficult to accurately quote their customers for shipping on their orders. After researching the problem and getting an understanding of the domain, we implemented an expected value algorithm for their sales. We then combined that with ‘The bin packing problem’, which is a combinatorial NP-hard problem in computational complexity theory to predict how many cargo crates might be needed for the expected sales completed by each shipping date. This allowed the company to quote for shipping in a far more sophisticated, accurate and competitive way when compared to their previous method of quoting to cover all eventualities.
We can’t say much about this project yet – but getting properly stuck into the payments ecosystem was rather fascinating – in the first case, it seems like most of the predictions for credit cards are fairly well solved problems that use algorithms such as gradient boosted machines. In the second, it looks like open banking will ring in some major changes in short order, and finally, the melding of econometric style analysis with data science proved to be effective and interesting.
A now listed company needed BI and reporting when they were back in their seed stage. The founders didn’t have either the time nor the expertise to do this important task justice. We came on board, spun up the urgent reporting and prepared a priority list for the rest. Then it was a case of setting about finding and hiring the right people to carry on and grow the work that we started.
Finding out what the normal peak flow of traffic through an intersection is traditionally involves hiring traffic surveyors for a day or two. This is a bit of a gamble – how do you know the surveys were done on representative days? However, where there are vehicle counters such as induction loops in the road, there are digital tomes of text files filled with data. What we did is take years worth of this data, built supervised learning algorithms that clean it up, then identify in milliseconds how big, how long and when the peak traffic flow is for each lane of each intersection. If you were to analyse all 42,000 intersections in the SCATS system, then our method compared to the sum of person-days to do the same analysis, is an astounding 38,000 times faster.
Bison Analytics exists to convert data into profits and performance with our technical skills in econometrics, machine learning, data science, statistics and numerical techniques. We have extensive experience in product development. Not only do we use Python, SQL and Stata, we’re also quite handy in Power Point. If you’re building something and need it fortified, or want something built, be in touch.
The Penthouse Level 2/28 Helwick Street, Wanaka 9305, New Zealand
Level 2/11 York St, Sydney NSW 2000, Australia
Level 2/11 York St, Sydney NSW 2000, Australia
Automated page speed optimizations for fast site performance