IPinYou, the market leading ad tech company in China, recently announced the launch of its bidding algorithm competition. Having invested heavily in the research & development of their bidding platform, it has now launched the global competition in an attempt to identify new methods, theories and formulas to refine the bidding algorithm. IPinYou hopes it will also stimulate further academic and research based interest in the space. This is likely to be the first time that ad technology components have been crowdsourced in such a way.
Democratising Data
Launching a competition to build the world’s greatest algorithm has a degree of geek cool attached to it, but the broader theme of data being democratised is infinitely more compelling. IPinYou is following hot-on-the-heels of Netflix, who also famously ran an algorithm competition. Beyond providing some valuable PR, it also gave them access to talented academics while also improving the product and user experience. This was potentially the first time that the hidden and often unknown art of data science really became popularised and brought to the fore. Netflix was the first but since we’ve also seen other companies get into crowd sourcing:
One of them is Kaggle. Kaggle can be best described as a type of Elance for Data Science and Analytics. It is a marketplace of sorts which connects companies with real-world data problems and real datasets with data science and statistical talent. Outsourcing your algorithm to some degree.
But another emerging theme is this concept of algorithmic exportability or having the algorithm as simply a component of sorts. Two interesting start-ups emerging in this space are Wise.io & Algorithm.io. Riding the crest of the big data wave these companies are effectively providing machine learning/algorithms in as-a-service type model.
The above examples further demonstrate the growing trend of not just data transparency, but data modelling transparency and ultimately transferability. Data Science no longer presents a black box. It provides a greater opportunity for the entire industry to have a greater understanding of how models are constructed, the underlying principles, which ultimately leads to the industry continuing to innovate collectively.
Open Sourcing Data Models?
Getting back to the bidding algo contest, it is not yet known how IPinYou will present the submissions or present the entries. But it is not inconceivable to think that IPinYou might expose the winning entry. And why not? Perhaps this is the next iteration of adtech becoming as open sourced as cloud computing and infrastructure has?
Datacratic were the first to make the bidder open sourced. Could IPinYou be about to make the algo open sourced? Where does the role of value creation reside if everything effectively becomes commoditised Questions would surely be asked, that in the instances of where ad tech companies are capitalised up to their eyeballs, how do they provide a significant point of difference when cheaper, open source kit is available to clients?
And also, who would be in the position to offer it all as an open source, free stack? Perhaps one of the larger media companies whose revenue is not dependent on SaaS deployments but rather where money is still made the old fashioned way? But if more of the middle layer became cheaper, then the transactional layer between buyer and seller gets more cost effective for both parties, which in turn perhaps drives increased investment?
How the market adapts to increased understanding of Data Science
What if data science becomes completely exposed, more accessible, how does the market react to this? This surely increases the focus and impetus on the likes of agency businesses to acquire more talent that understands how to read data and have proficiency in query languages and data analysis in tools like SQL and R?
These people won’t be confined to specialist, isolated teams which are marooned from the rest of the business. Over time, client service teams, planners, account management will need more data fluency. And the reason for this is because marketers will be more data fluent. The concept of data science being smashed wide open will also have telling effects on marketers and organisations. They will begin requiring their marketing managers, brand managers to understand how data influences all decision making within the business.
This provides an interesting challenge for service-based businesses, namely how will they re-architect their business to meet this challenge and requirement - and more importantly where do they source the talent? Perhaps as a basic requirement, all employees will need to learn how to use open source DBs like MySQL/PostgreSQL, and even the data scientist's programming language du jour, Python?
The concept of algorithm contests, machine-learning-as-a-service might seem obscure right now, but it has the potential to shake up the very foundation of digital marketing. We will be doing a series on the "data tools of our trade" in the coming months, as we build upto ATS London.