individuals who code are blessed beings. they speak both the language of humans and machines. they ‘make’ things. they are artists who solve problems, many a times some of the biggest questions posed to man kind. they’re taking over the world with their ways of life and hacks. they are taking over the world with their deep learning and agi dreams. they mostly love simpler ways of life. they believe in what they do. they believe in lean and integrated systems. recently came these unicorns aka 10x developers who with their workflows were said to be productive by multiples of their clones; as a dear friend/mentor @alixedi says on his about page “I write frameworks that 5X developer productivity — reducing technical debt and increasing bus factor”. some say such unicorns don’t exist. i’m an idealist who strives to become one!
another class of individuals who do something similar to the above are the financial analysts aka suits who are sadly stuck in older ways of excel. to better understand the similarity one needs to synthesise the day-to-day activities of a financial analyst: they collect information from public and portfolio/client companies being from investment advisory/private equity shops and use this information to make models. the process of making these models is no different from writing code minus the pre-historic toolchain they use: excel, email, powerpoint, google, capitaliq, pitchbook, capgemini, bloomberg and more excel. they are also as finicky about the structure of their models and break them up into operating models (which also follow convention of revenue drivers, cost levers, etc) financial statements, returns analysis and the most important dashboards. these interactive dashboards with buttons, dropdown, interactive graphs and scrollers help decision makers base their decisions on what they understand or as another dear mentor/boss Asfi says: “believe they understand”! interestingly these financial analysts are equally committed to the readability of their models, hence use macros like tts turbo and wst as coders use linters. they make frameworks like the KADMOS developed by Asfi which ensures that their throughput increases multi-fold. Like coders write unit tests and get qa engineers to test their code, financial modellers have peers/seniors audit their models and ensure it passes the qc. they have their revolver code snippets and boilerplate financial models like coders. infact, they are as obsessive about their keyboard shortcuts and minimal use of mouse/trackpad as coders are about their text editors and debuggers.
a minor error on the part of both classes of artists can impact millions of lives through [financial crisis])(http://en.wikipedia.org/wiki/Model_risk) and disasters.
the above just opens up a huge space for innovation in improving the ways of financial analysts. this space can be broken down into three categories: information aggregators, workflow/productivity enhancers and knowledge retainers aka knowledge management systems (which are somehow not progressing beyond wikis/document tagging, more on this towards the end).
the first category of products, these information rather data aggregators are platforms like quandl, petrofeed and enerknol and regulations to enforce technologies like xrbl come into play. imagine the gain in productivity if a financial analyst can just pull in most updated, verified numbers (macroeconomic and market?) and policy summary (power policies and fiscal regimes of oil concessions?) into their models. citing sources will be easier and single sources of truth shall prevail. this in my opinion is more of a laborious challenge till ai evolves (yes evolves! by learning its way up to becoming a good research analyst) to a state to start identifying data sources, verify their accuracy and hopefully identify linear relationships between them to populating the databases of the quandl’s and enerknol’s. this effort as being done by the above platforms requires knocking on the doors of each data source and integrating with them or writing web-scrapers at times ;) this effort will be as pain-staking as the toughest challenge but foundation of all predictive analytics efforts: building robust data warehouses.
the second category of products are tools which reduce the collaboration time for financial analysts. which help them use the technologies developed by coders for making their industry decks and financial models collaboratively. this is where i find thinknum which is still in beta to be a game changer as techcrunch also recently published. but it is still a very small piece in the toolchain for 10x financial modellers. while excel compatibility, version control, publishing dashboards (to some extent) and concurrent multi-user edits is covered they need basic features like the tts/wst macros and more keyboard shortcuts. one feature that i would love to see/build on top of the thinknum platform would be a dashboarding platform so that analysts can make fancy looking model outputs with buttons, scrolls, input fields and graphs such that the user can play around with inputs and get updated numbers on the dashboard without having to go through the financial model if the model is rated well by other users lets say (we can figure out the gamification/credential setting in detail later). this way those users who understand irr’s and dividend yield’s but not necessarily modelling or excel (or thinknum in this case) will be able to play around with a few parameters and see their impact on a if the dashboards are designed well enough by the analyst (again, the analyst can get points/tips/monies for her dashboard, we can work it out later). not only this but an enhanced feature would be to be able to fork these models and embed a chat/discussion on the same page so that deals can be negotiated through one reference model creating real value instead of people (generally in suits) trying to outsmart each other or the governments (and vice versa!). analytical applications are being seen as the next big thing where the new rockstars aka data scientists analyse a problem which can be solved using running ai broadly (dm/ml specifically) algorithms to ‘predict’ outcomes for present/future cases. the above form of dashboards will be analytical-apps-2.0 as the user will be working around parameters on runtime which will run financial models which can further be backed by dm/ml algorithms and generate insights for them. this will be what-if-analysis++ of sorts. we still haven’t talked about migrating unit testing and qa like workflows from the coder’s world to the financial analyst’s world. what about continuous integration? what about integrations with slack where the analysts collaborate?
the third category of products is knowledge bases aka kms. somehow this category which is the most important for enterprises/institutions/individuals of all shapes and sizes has been stagnant for some time. the world seems to be awed and struck with wikis and tagging and taxonomy. the biggest names in the game offer document management, workflows (which are basically 1-step 2-step approvals, basic forms and some data transform+push/pull from structured systems type of functionality), wiki pages, project sites and funniest of all social networks. why will i want to use a company social network? really man? i’ll just fb/tweet if i have so much time and add my colleagues there. we need workflow management systems which understand user roles, their current task at hand and should be able to provide them with related information from within the corporate database and the internet on their screen. much like a personal assistant or agent as we call them. like the ones aigents are trying to build along with the famous efforts by the big 5 (msft, goog, appl, fb, amzn).
it is believed that 5x/10x developers need the appropriate cultures and environment to deliver with such high productivity.
under such circumstances an account of the ideal day at office of a 5x/10x financial analyst is presented below:
the analyst starts the day with a desktop based virtual check-in where the ai enabled desktop assistant pulls her objectives for the day from a goal alignment software like betterworks (which is an amazing tool in itself!). as the analyst picks the first task up, an overlay checklist is presented to her to ensure that she has taken all necessary steps as defined in the procedure/ organisational best practices. it may also help her contextualise her current task by providing relevant past work from within the company. incase the analyst needs to conduct a research, the ai agent pulls readily pulls data off quandl, enerknol, petrofeed and the likes. the analyst can do further research and recommend newer sources of information which are in turn presented to these data providers for addition into their services if found appropriate. as the analyst moves on and develops a financial model based on her research on a platform like thinknum, she continuously collaborates with experts in lbo modelling located in another geographic location maybe as contractors to the project through slack and thinknum. once her dashboards take shape, she has checked off all procedural checks on the task and her work is qc’ed and audited, she can send the same to decision makers. who are not only presented with the dashboard but their agent also provides them with the history of the case at hand, any precedents and potential impact of decisions. once a decision has been made, it is recorded in the knowledge base and the company can work towards a culture of validated learning and use knowledge as a competitive advantage while ensuring analysts only generate insights instead of collecting and cleaning data, productively using their time and wasting lesser brain cells on things which can be automated.
a wise man once told me that i can only be as smart as the average of 5 people i spend most of my time with. i think he was absolutely right! i changed my 5 people and keep looking for a better 5 everyday. if you feel we can collaborate on the above ideas do ping me on twitter.