We build a flight arbitrage system and it was doing good, i think what contributed to the success (from MVP to profit) of that project was that we gave access of live information to buyers, so they can see which campaign is working better than other and they can make quick decisions like which campaigns they have to optimise etc. So when tech is stable and we are building new landers and buyers are seeing live data and optimising campaigns for profit, i think that’s the time we thought why not an automated system that should take care of campaign optimisation.
So we are talking about to run that expert system to do optimisation on already running campaign that can take quick decision for higher profit.
Algorithm: round robin with weights
Fig: with dummy data and is not related to any live scenario
Flow: So when user search for a flight route we open up a new tab and now we have 2 tabs where we can show listing, and when we open new tab user is focused on new tab.
System will take care of:
- which partner’s ads to show on focused tab
- which partners ad to show on previous tab (Leave Behind Tab)
- Drive traffic volume to which partner combination
Expectations: so just taking care of just these 3 things our system can drive more profit as we will drive volume to high paying RPC partners
Dry Run / Testing System:
So we created a test service that run on base of dynamic RPC for each partner through multiple weeks or months with a dynamic traffic (multiple tests with steady and dynamic traffic) and observed the results which we can see on these line charts and it looks promising. and then COVID flatten the travel industry and we never get a chance to get to a traffic volume where we could test this system.
This project and this service was a quite good learning for me and I learned a lot, from thinking about a system till delivering a solution and seeing expected results and above all accepting that this would never make it to production.