Machine learning has been the subject of great hype over the years, both in the media and in science-fiction. It is a catch-all term that encompasses increasingly diverse fields of study, methodologies and processes. It engenders thoughts of near-human artificial intelligence that, despite recent developments in both theory and practical application, have so far not been able to be truly recreated. At its base, machine learning is simply teaching; teaching a ‘machine’ (program), using algorithms, to search for patterns in massive data sets. The desired outcome is a ‘machine’ that can look at a series of problems and generate a conclusion. Both the question and the answer are difficult to scope because, unlike teaching a person mathematics or a language, machine learning has to invent the brain first. Modern programs are asked to analyze more data, to be more agile in their conclusions and, in the case of unsupervised machine learning, to be fully autonomous in their ‘thinking’. HP is using machine learning to improve its partner portal. It has diagnosed some of the issues at hand for channel partners and has broken them down into four key areas, hoping to use machine learning to create a tailored partner portal experience.

HP develops ‘Navigator’

HP’s channel marketing and operations team is currently developing a machine learning system to improve its partner portal – called ‘Navigator’ – with a team of developers at a company spun off from Cambridge University. The code is being designed to teach the software to analyze the many different partner data sets including current profiles, sales, training data, marketing activities and customer information; and enable partners in four key areas:

  • Training (skills, competencies, how-to videos)
  • Marketing (assets, campaigns)
  • Product (pricing, updates and specializations etc)
  • Sales (deals unfulfilled, opportunities etc)

There are five general stages in the development of machine learning systems, from ‘historic’ (reporting) to ‘real-time’, ‘predictive’ (predictive analytics), ‘automated’ (machine learning) and the ultimate ‘intelligent’ (cognitive systems) stage. HP is working towards an ‘automated’ system that will allow its partner portal to deliver partner-specific, actionable information.

The team has used open source software for the development, which it estimates has cut core development time to six months, within a total project time to date of 18 months. Navigator is currently in its alpha testing stage, running in the background of HP’s partner portal system, gathering data, analyzing and returning results while the team stress-tests the coding. The goal is to continue this development over the next twelve months before launching fully in the second half of 2017.

The opportunity for Navigator lies in being able to give partners a strong, simple enablement tool within the current portal, which increases information on pricing, training and certifications; contributing to improved revenue and partner engagement. Importantly, it must show partners what they need to do to drive more business, whether it is taking on specific training or focusing their marketing campaigns on a different area. If Navigator can do this, it has potential ramifications for portals everywhere.

Enablement tools are the ultimate key for partners

Machine learning has evolved over many years, from early visualization theory to mathematical calculation and prediction in IBM’s ‘Deep Blue’; and then to ‘Deep Learning’ (a branch of machine learning that seeks to replicate further the ultimate goal of autonomous artificial intelligence). Even at its simplest levels, this is complex technology. It is attempting to replicate the brain’s ability to look at a problem involving large amounts of information and make a suggestion. The information, the suggestion and the process used for analysis vary depending on the branch of learning and the field it is tackling. However, considering the amount of learning a human goes through in a lifetime and the countless influences on the learning process, the scale of the task is immense, no matter the field.

To take ‘Deep Blue’ as an example, though strictly mathematical in nature (using probability trees to ‘look ahead’ to possible outcomes and evaluate all likely winning scenarios after each move) and based on a code written by a person, this kind of machine learning is only possible if the code is written by someone that understands the root of the problems the program will likely need to solve. This becomes further complicated in language-based machine learning. Being able to look into the language of channels, their structure and those of the individual partners is a complicated process, even along carefully structured lines.

It is unlikely HP is aiming to ‘teach’ the software autonomously to create fully tailored strategies for partners, mostly as this would require almost incalculably complex code and years of development for the creation of a neural network that is able to craft sales and marketing strategies in the way a human might. It will more likely be a simpler ‘call-and-response’ system, where the algorithm will sift through a partner’s data and decide, based on a set of signposts and guidelines, to pull certain pre-packaged action items and assets off a ‘shelf’ for the partner to use. However, there are issues with even this, more simple, approach.

If these pre-packaged ideas are not considered by partners to be enhancing or adding to their decision-making process, they risk being ignored or worse, turning partners off from their portals. Advice must be truly tailored and HP is targeting a threshold of no more than 30% of common information provided to each partner. In other words, if any of the information or recommendations seen by a partner are more than 30% similar to those seen by another, it cannot be deemed ‘tailored’. Some crossover is inevitable for partners within the same market space (and there are many in HP’s partner program) but this threshold appears manageable and realistic. As long as partner feedback is properly overseen after launch, this percentage level may even be reduced.

According to HP, more than 60% of partners use a standard text box field to search for required assets and information in their portals. This could be because some portals are seen as complex to navigate but also because partners understand what they need and prefer to look for it as and when they need it. With this in mind, the company faces two main options. HP can either make Navigator a small add-on that provides suggestions across the four key areas mentioned above and allows partners to choose how they engage with these, or it can use it to completely overhaul its partner portal. Partners may view an add-on as further complicating the portal environment, while an overhaul brings migration risks. However, it may be a risk the company needs to take to differentiate itself in its partner management. Portals are an important element of the partner-vendor relationship. If HP can add to this in a way that truly makes the lives of partners easier, it will be worth the time and resource investment. The company will almost certainly be looking to go further in its development and application of machine learning and this may be the right environment to test its current knowledge. However, this may also fundamentally shift the way partner portals are seen, in a way that tracks with wider market changes in technology and digitization, enabling partners to manage their positions in a competitive and rapidly changing landscape.

Receiving updates

Receive our latest PRs on emerging, enterprise and mobile tech delivered straight to your inbox.