I retired from personal blogging in July 2008 but you can find me over at blog.xero.com
Follow @roddrury
Some of you may have noticed a bit of coverage on my next big thing: Xero. Idealog and Computerworld have covered us just recently and you can find out a bit more what we’re doing in my Welcome to Xero blog post.
Xero is an online accounting solution for small businesses, delivered on a Software as a Service (SaaS) basis. I believe it is one of the biggest market opportunities out there and a global opportunity we can credibly go after from New Zealand.
We’ve been heads down on this for some time. We have a number of beta customers on board and will soon move into a limited release. A challenge for SaaS offerings is building out operational capability. As the software is getting to where we need need it we are now beginning to ramp up that investment so we can deliver a customer experience in line with the quality of the software. For that reason we’re bringing in customers in small batches, learning where they need guidance, rinse and repeat.
Xero is all about people. Making a difference to the people that run small businesses and providing an opportunity for our best talent to build a world class company. The next step in my career is to build a long term company and culture that attracts and fosters talent, to earn export revenue.
Xero is design led. We recruited early some of the top interaction designers we could find Philip Fierlinger and Grant Robinson.
We’ve put an awesome technical team in place with Craig Walker, Kirk Jackson, Andrew Butel, Fletcher Brown, Adam Burmister and Jeff Wegesin.
We’ve invested heavily in product management and customer care with an exceptional group of customer advocates: Michelle Perera, Andy Leeb, Catherine Walker (Orange Girl), Catherine Robinson, Donna Wylie, Larissa Paris and Louise Roebuck.
We managed to attract Kate McLaughlin from the National Business Review to run Xero marketing and communications and the very talented Darryl Gray is in house brand guy.
Small business accounting identity Hamish Edwards brings the domain expertise to ensure what we deliver is not only technically world class but really does improve the performance of a sector that makes up 95%+ of all businesses.
It’s early days but we think we’re doing something special.
Through this blog I’ll try to share the experience. Hopefully you can enjoy some of the fun.

Rod starts to talk about Xero…
Rod,
I have noted that your Accounting System does forecast? I am no accountant, however I am familiar with forecasting algorithms, but I am wondering if the forecasting function adopted in your system an accounting forecasting algorithm or statistical forecasting algorithm ? You don’t have to tell me this, as I understand that it is confidential to your product development.
As a non-accountant myself, I think (I am not entirely sure about this) that i>accounting forecasting algorithm rely on calculations of Present Value , Future Value and Rate of Returns, etc, which are naive. This means that forecasting using accounting forecasting algorithm, there is no measure of statistics involved. Example, the expected future value at the end of the next 2 months is projected to be $200,000 . It doesn’t give the confidence, something as projected to be $200,000 at a confidence level of 90% or so. On the other hand, statistical forecasting algorithm not only give a future value projection but it also gives the statistical measures of probability confidence, model goodness of fit, etc,… Historical data is required as input for statistical forecasting algorithm while not necessary needed for accounting forecasting algorithm.
There are a number of statistical forecasting algorithms available from literature, open source, commercial tool today with different capabilities in terms of robustness and accuracy but ARMA (Auto-Regressive Moving Average) and its variants (ARIMA, VARIMA , SARIMA, FARIMA) are quite popular.
Statistical forecasting algorithm such as ARMA is a generalized time-series algorithm and not restricted to application in finance only but also to anything & everything that is time-series based, such as forecasting temperature, forecasting the quantity of consumption of a specific supermarket item, forecasting the number crimes committed, etc,…
I have also noted from the description of your producthere that it does use charting. Here is something new for you to try and implement it into your product if you think it is worth it. You can use some pattern recognition algorithms for detection of non-obvious but hidden patterns in the time-series data. Automatically alerting these hidden patterns to the user, IMO is important. Perhaps, you can start with Dynamic Time Warping (DTW), where there are lots of free codes available from the internet. DTW is a popular algorithm for DSP (digital signal processing) and primarily had been used for speech recognition of today. It has also been adopted recently for application in Bio-Informatics for detecting similar DNA sub-sequences. DTW has been around for over 20 years. Here is a brief of how to use DTW. The user can specify a time-interval that of interest to him/her. Perhaps this time interval was something dramatic happened to the company cash flows (or some similar event), and the user (may be the management) want to flag such time-sequence as to be alerted if it is likely to happen again in the future. This flag pattern time-series subsequence is then compare with the longer time-series that is available from the database to find other time-series subsequence that are the closest match. See, DTW not only able to find similar time-series subsequence pattern from the past, but it can look for a likely similar emerging time-series subsequence pattern in the future. DTW has been applied to automatic detection of the onset of seizures at critical care units at hospital these days. The seizure onset time-series subsequence (signal) is known as a prior by medical practitioners. The instrumentation for the embedded medical system is constantly checking in real time any sign of such pattern starting to be developed. The medical practitioner could then be alerted to attend the patient immediately therefore he/she could take some pre-emptive measures or treatments that would have avoided further complications.
Here is another variant of DTW where you can download and look at the algorithm which is described in there:
Derivative Dynamic Time Warping
Finally, the other pattern recognition algorithm that is worth exploring is the Fast Fourier Transform (FFT). The charting for your system shows red & blue regions. FFT can tell how often (ie, what is the frequency occurrences) an event takes place, ie, weekly, monthly, every 3 working days. This means that FFT detects the occurrences of regularity in a time-series. There are also lots of free FFT codes available on the internet.
Having had the privilege of meeting several of the Xero team at our own name launch, and having had a quick guided tour of the business and the product, there’s no doubt that Xero is buzzing. Yhe product looks great and should really appeal to the target market. Congrats on HC too.
Some of the URLs I have posted in my previous message are not properly link thru. Here are the correct links:
Dynamic Time Warping (Wikipedia)
Derivative Dynamic Time Warping (Algorithm Derivation paper)
Fast Fourier Transform (Wikipedia)
Falafulu Fisi: that was the most impressive use of acronyms I’ve ever seen in a comment - seems like you should send me your CV :)
Falafulu Fisi: the fact that you know how to warp time should secure your employment!
Good luck to the team at Xero, you certainly have the people capital to pull it off. Might have to get my CV in to as long as HC isnt in charge of team bonding activities.
I’ve seen the future… and it was xero. No really though I was privileged enough to be involved in the Beta testing and I was and I am still very impressed - not to mention looking forward to being an actual user.
What I am excited about is the Xero developers API so that I can integrate it with my own Web 2.0 e-commerce software for WordPress ;)
In fact I’m busy trying to find the right team / partners to help me market this part of my business now…
[...] http://www.drury.net.nz/2007/03/06/xero/ [...]