Another spanner in the works was the ambiguity around having two engineering teams, led by two separate CTOs, building online video platforms, Delta and Sierra - a change that happened a year into my tenure after just landing the full-time CTO role (I'd stopped consulting, only to find myself in further uncertainty with a business merger and potentially more rounds of org-design changes). Delta was under my ownership, serving traditional TV customers with value-added internet services to access Live TV & Video on Demand (VOD) streaming. Sierra, was a pure internet, deep catalogue subscription video on demand service (like Netflix), built and run by a parallel engineering team, under different CTO, in another country, outside of South Africa.
So, with these many changes in executive leadership, at the top-level financial review - executives would naturally inspect why the business seems to be duplicating engineering efforts at building online video platforms, even though the two products serve a different set of customers, emerged at different timelines, using different set of technologies - but they both appear to "stream video" at the high level. So why is there duplication? Where are the costs going? How do we streamline engineering costs?
Enter the Financial Model & Technology Budget Commentary Document
The first order of business for me, being tasked with a turnaround challenge - was to review the existing cost components, identify the main cost drivers and derive a perpetual financial model based on the core business driver of growing monthly active users.
Starting in 2017, neither the technology team nor the finance managers could tell with a high level of confidence what the true costs of running the technology platform were, and ultimately what the cost of running a technology platform translated to a cost per user value, and ultimately how much from the profit margin this cost was eating into. My platform provided value-added-services for free, existing users were homed to a primary TV subscription package (satellite TV using a dish and decoder / Set Top Box (STB)) as main profile. Additional viewing profiles for online consumption through mobile phones, smart TVs, game consoles and web browsers, were offered at zero cost. Additionally, my platform served internet connected STBs with larger video catalogues for streaming on demand, this too, at zero cost to the subscriber.
Financial Model (brilliant, if I say so myself!)
So I created a detailed financial model that was able to show how to forecast a largely fixed cost investment, to operate in a variable cost domain. The largest cost components for video streaming is the infrastructure cost for the networking pipe (internet transit) and content delivery network (CDN) data costs. Since video consumes large amounts of data, these costs are covered by the streaming operator. CFOs prefer fixed cost accounting than variable costs because it helps manage their cashflows and forecasts better. In an on-demand video world, where customers come and go, or viewing behaviour varies by seasonality or during peak events, the experience is more bursty than constant. This is why video providers focus heavily on content personalisation and recommendation systems to drive stickiness and increase engagement. The more users watching video, the more data is consumed, the costs of data increases. The longer users remain on the platform, the more users are engaged and the more video data is consumed. All these point to increasing costs - but - at some point, the costs per user eventually decreases because the number of users increase, the cost per user eventually decreases. I created a financial model that showed how we could manage variable costs through a fixed cost model, by buying data wholesale upfront, using a minimum commit model with overflow-at-zero-cost into next year for unused data, and also negotiated year-on-year cost savings, since cost of data decreases year on year.
Using my financial model, I was able to show significant cost savings to the tune of R80m ($5-7m) in two years. I also, single-handedly negotiated significant costs savings with our primary transit links and CDN providers, as a result of my financial model forecasting. In year one, I negotiated a 1233% increase in CDN bandwidth and secured 71% decrease in out-of-band data costs, 50% decrease for inband per GB data costs - keeping my overall costs relatively flat to track a fixed expense. In year two, I negotiated additional cuts, reducing in-band data costs by a further 63%, out-of-band by 48%. In year three, I negotiated further deal reductions because of increased competition in both the transit-and-cdn space, landing on a further 20% reduction on inband and out-of-band costs. Over this period, I secured enough guaranteed capacity in the network that exceeded future growth targets, but kept most of the projected fixed costs inline. As a result, when I decided to leave the company, in my 3 years tenure, I drove internet costs down by 84% on a cost per user metric and 96%reduction on costs per GB per user metric. Additionally, working with the same transit and CDN partners, I secured more than 7X improvement on overall internet throughput, reaching close to 1 terabits/sec on network capacity, which for the African continent, is not bad going indeed! In terms of storage, from 2017, I increased the data commitment by 2000% cumulatively year-on-year.
