20 Jul, 2017

Enterprise Solutions

Do you need to maintain your edge over the competition or increase your profitability? IA will develop bespoke software to suit your needs and environment.

Intelligent Collections (Debit Orders)

Companies using debit orders usually have unpaid rates of 10% or higher; money that comes off the bottom line, as the resources spent collecting these amounts are usually expensive, and follow-up has a low success rate.  Many companies terminate the contract after a few consecutive unpaid instances.

IA’s PayLift predicts which clients have will not pay next month. The tool also calculates for which of those clients it would be profitable to request a Non-Authenticated Early Debit Order (NAEDO). These debit orders are processed a few hours earlier than others, and also allows a company to collect an outstanding premium for a number of days after the scheduled debit order date. NAEDO transactions are approximately ten times more expensive than normal debit orders, but with PayLift implemented at a well-known South African insurance company, targeted NAEDO has been shown to increase collections by 0.5-1.5%, after NAEDO fees. This translates into a 0.5-1.5% increase in turnover, simply by building intelligence into the collections process. Through the payment data, IA is also able to predict the expected client lifetime at your company, and the total lifetime value, allowing us to provide strategic collections intelligence: a real and really profitable application of advanced algorithms. We have also found a significant increase in retentions due to a drastic reduction in involuntary terminations due to non-collection, increasing the lifetime value of a client.

The debit order landscape in South Africa is changing, however, due to PASA’s new DebiCheck initiative. At IA we welcome this initiative for reducing debit order fraud and are looking forward to helping our clients navigate this change, increasing the fairness and transparency for both service providers and individual subscribers.

Saving Intelligently for Retirement

Investing optimally for our retirement is a really difficult problem. Although financial advisors can show that e.g. investing the full 27.5\% of your taxable income allowance in a Retirement Annuity. Pension funds also allow you to withdraw up to a third as a lump sum. Although financial advisors can demonstrate that saving 27.5% in an RA is not the most effective way to structure your savings, it is very difficult to find the optimal allocation, given the tax implications now and at retirement. For example, the more I put in my RA now, the lower my taxable income, and the more I can save on the same salary. But it does mean that during retirement I’ll be paying more income tax than I would if I had put the (admittedly less) money in Discretionary Investments such as Unit Trusts, Exchange Traded Funds (ETFs), etc. But DIs tend to grow better than RAs, if only due to lower fees. So how should one split one’s savings? At Invoke Analytics we solve problems like these by using mathematical algorithms, machine learning, and statistics, and decided to try our hand at it. The result is PyFin.

PyFin is a tool which helps individuals decide how to split their retirement savings given South African tax rates. It considers how much money is already in those accounts, how future savings can be allocated, and how they grow and are taxed now and at withdrawal. It then devises a plan that will maximise a person’s mean income after tax during retirement. The output of the tool is a report with a possible allocation plan for the rest of this tax year. It also provides the projections into the future. PyFin is an open source project which we hope will go live very soon. You can check the progress here.

Retail Solutions

Retail store managers need to decide how much of each product to order so that products are not over- or under stocked. To do so requires data from the shop floor and stock on hand files on database software, pdf files of incoming deliveries already ordered and bulk palette specifications, and the product’s sales history. The manager then tries to forecast future sales and places an order based on this estimate, the business’s projected cash flow, and when the order needs to be paid. This is usually done using spreadsheets. Even small retail stores carry thousands of different products. It requires constant focus and vigilance to ensure that nothing is missed. Errors inevitably arise and propagate through the process. The decisions based on them cost money in lost sales or overstocked lines that don’t sell but need to be paid for. Managers need a tool that can do this quickly and reliably for all their products.

We created an enterprise solution that relieves the manager’s administrative burden. We automate the process of drawing data from the database program and read pdf bank statements, invoices, delivery notes, and product specifications. We consolidate all this information, allocate cash flows to different accounts, find patterns in the sales history, forecast future sales accurately, and combine everything to supply the store manager with a suggested order list that maximises profit and while keeping the business cash-positive. He adds human insight and prints out the order form.

From the data automatically drawn from the database software, the software sifts through the thousands of products and provide daily updates of store health based on advanced AI algorithms. At a glance, the manager gets a daily update of problem areas. The result was that the administrative load decreased, profit increased, accounts were paid on time, stock health improved, and management had the time and mental space to focus on all the intangibles that make stores truly great.

Other projects

IA has done various other projects, from intelligent routing of delivery vehicles from multiple depots to multiple clients, Optimal demand management pricing of airline tickets, web scraping, increasing the competitiveness of clients on online platforms, fraud detection, designing and analysing real-world statistical experiments, and patented technology for condition monitoring on billion-Rand steam turbines.

What can we do for you?