The idea at the core of PredictHQ is simple: You can’t prepare for what you don’t see coming.
Yet it takes enormously complex data science to achieve demand intelligence for forecasting through our API of verified and ranked events.
Today we’ve launched Aviation Rank™ out of beta, our first industry customized system. It identifies the impact of real-world events on flight bookings. It’s both a very exciting product release for us, as well as the perfect opportunity to introduce our approach to data science, machine learning and powering artificial intelligence.
Introducing Aviation Rank™
Imagine being able to know at a glance how much additional demand an event will create on your routes – that’s what Aviation Rank™ achieves.
No more googling for events, no more guessing, no more missing demand spikes you could have seen coming.
Aviation Rank™ enables airlines to instantly know which events are most relevant to them and worth focusing on so they can adjust their load factor, supply chain and marketing.
Built on top of our millions of verified events, Aviation Rank™ is a predicted impact score calculated by advanced machine learning models. It ranks each event on a logarithmic scale of 0 (no impact) to 100 (major impact). It identifies around 3000 high-impact events each month – each offering multi-million dollar opportunities.
And it’s not just about the biggest events – one of the most common pieces of feedback we get from our airline customers is the breakthrough value for them is when they uncover the clusters of smaller events they would otherwise have missed. Often these perfect storms of demand have just as much impact as the biggest events, if not more.
Identifying the most relevant events for airlines requires finding and verifying many more. PredictHQ’s systems find more than half a million new events each month. More than 20,000 of these are ranked as major impact by our general ranking system, PHQ Rank. These 20,000 or so have huge impact on many companies, but airlines are focused on a very particular behaviour – will people book plane tickets to attend? Aviation Rank sorts through all our events to identify the ones that will.
The potential of knowing your company’s high-impact events is huge. Aviation Rank™ reveals to an airline where they can focus their efforts to achieve more accurate load factoring. And it is powerfully useful for managing inventory, supply chains and minimizing surplus capacity on a route. Most airlines revenue analysts work off a list of HSE (holidays and special events) but these lists can’t provide insight into all the high-value events worth tracking.
We developed it working closely with three airlines including a major national carrier, as well as AmadeusIT Group. Our lead data scientist on this project, Dr Xuxu Wang, explains more on how it works in this article.
Launching it today is the result of years of hard work. While much of the work to build Aviation Rank™ took place in 2019, it is only possible through the deep knowledge graph we have been developing since we first started working on PredictHQ.
PredictHQ’s knowledge graph
PredictHQ’s foundations gives us a lot of unique high-quality data to work with. It’s one of the reasons data science teams at our customers get so excited about what we’re doing.
Building our knowledge graph is one of the most important priorities for our team. It includes many models such as our verification and enrichment processes, our unique entities system, NLP work, and our proprietary ranking systems. We don’t talk about these much publicly, as they’re both complex and key IP for us, but Aviation Rank™ begins to reveal what PredictHQ will achieve in the future.
Without years of verified event data to work with, our patent-pending ranking systems would not be possible. We have two general ranking systems that are available to all API subscribers that enable much smarter demand forecasting. Aviation Rank™ is our first of many industry customized ranking system.
To differentiate precisely between similar looking events that will generate very different levels of significant inbound bookings takes many layers of intelligence, and more processes than a human can perform at once. Any ranking system needs to sort millions of events constantly and is powered by a network of machine learning models.
We can’t name every model required, but here is a small subset to indicate how complex ranking events is. You need models to:
- Identify if an event is real
- Identify similar records and which is most trustworthy, even when they have different names or locations (or both)
- Identify which real events will drive people movement
- Recognise if an event is recurring
- Categorise events into what type they are to better identify volume of people movement
- NLP processing to identify what sub-type of event
- Understand the median impact of each sub category of event and how this differs depending on expected attendance totals
- Understands how each event differs from the median to accurately calculate impact
And this is only a tiny fraction of what our event processing pipeline and ranking systems do. All of these systems took time and testing to perfect. But now they are established, they continue to develop and refine as they keep ingesting more data and our data scientists keep adding to each model.
This enables us to not just provide the most intelligent event impact data, but also to create entirely new demand intelligence products that open up forecasting capabilities that people have never considered.
What comes next
With a knowledge graph growing so swiftly, the team has to stay very focused on our roadmap even though there are so many exciting projects in the future.
We are planning more industry customized ranking systems, as well as working on a range of other features that are useful to our customers regardless of industry.
Getting the balance right is important – we have enough large customers in aviation to focus exclusively on that industry and build a huge business. But we have similarly-sized customers in accommodation, in transport as well as in food retail. And those are just our first four focus industries.
We are on track to announce some highly requested features over the next few months, and work closely with our customers to better understand their needs and wishes, so we can keep delivering the highest impact product for them.
So today, we’re delighted to announce Aviation Rank™, which will transform airline demand forecasting.