As the name suggests, adtech and its success is firmly rooted in the infrastructure that supports it. Adtech companies are measured by their ability to deliver advertising services within the finest of margins and the minutest of milliseconds.
To put that into context, supply-side platforms (SSPs) have set the deadline of around 150 milliseconds to receive a response from demand-side platforms (DSPs) once a bid request has been sent.
Delivering that speed of delivery and response requires high performance, low latency infrastructure.
The adtech industry is relatively new. Since the first display ad appeared on HotWired in 1994, adtech has evolved along with the rise in internet users, the proliferation of social media, streaming platforms and ecommerce. Some companies picked up the opportunity early and the likes of The Trade Desk, Nielsen, WPP and Criteo are now large, established adtech companies. Others are just getting going.
As with almost any industry, when an adtech company is in its infancy, it doesn’t have the people power to manage the infrastructure it relies on. It’s why so many are drawn in by the attraction of the hyperscaler clouds. Offering a substantial amount of free infrastructure credits, it’s a no-brainer for new entries to the adtech space to work with a hyperscale cloud provider to get up and running quickly.
But what happens when those credits run out? At this point, hopefully, your company has grown, and your customer base has expanded. So, with free infrastructure at an end, where does this leave adtech companies?
This blog looks at the true costs of hyperscale cloud and the longer-term infrastructure options open to adtech companies.
There are three key service providers within the world of adtech: demand side platforms (DSPs), supply side platforms (SSPs) and ad exchanges.
A DSP is software that is used for programmatic media buying; automating the process of real-time bidding so that there is very little input required when securing placement for an advert. It will even match your campaign to the most relevant audience on the publisher’s website. A DSP aggregates inventory from multiple different resources including ad exchanges and SSPs, calculating prices and delivery based on data.
Sounds like a lot? It is and DSPs have the architecture to match. They are generally made up of a bidder, DSP ad server, data platform, UI, banker, campaign tracker, reporting database, user profile data base and third-party integrations. Together they provide real-time analytics, audience targeting, bidding, budgeting, and campaign management.
As such, DSP infrastructure requirements center around storage for inventory, as well as support for tools such as Kafka for analytics and Aerospike for databases. With the number of transactions taking place, alongside the data analytics tools, it doesn’t take long for storage to become a key challenge for DSPs from a cost, space, and configuration perspective.
SSPs on the other hand are most concerned about bandwidth. SSPs work with publishers - websites, video games, social media sites for example - to monetize their platform by managing, selling, and optimizing inventory or in other words, ad space. Like DSPs, for SSPs to provide these services requires a number of components. Including integrations with other adtech platforms (DSPs, ad servers and ad exchanges), trackers collecting data about a publisher’s website and audience, and a reporting database to generate reports and view campaign analytics.
Thanks to the sheer volume of information being sent from their SSP ad servers and platforms, supply side platforms need infrastructure that provides cost-effective and reliable levels of bandwidth.
Lastly, ad exchanges or networks are the transaction points where money is exchanged for an advert that matches a particular target audience on a publisher platform. At a high level, it's an exchange between a DSP’s ad inventory and an SSP’s ad space on a publisher’s medium. Most of the time a DSP or SSP will have an integrated ad exchange/network functions within their platform.
It’s here that ping - the time it takes for a request to be sent from an SSP to a DSP and return with the winning bid - which, remember, needs to be below 110 milliseconds, is most important. A low latency network is crucial to delivering the industry standard ping times.
The hyperscale cloud providers hold the largest market share of the adtech infrastructure space. A key reason for that success are the free credits that they provide to start-ups and young adtech companies to build their platforms on the hyperscalers infrastructure. And why not? Money is always tight at the inception of a new company and if adtech companies can save themselves significant infrastructure costs by taking advantage of the free credits, who can blame them?
The problem is that - as mentioned earlier - those free credits have a cut-off point. And as companies become more successful and start to grow, they naturally also start to require more bandwidth and storage.
These requirements aren’t always predictable or stable. Like the stock market, there can be predictable daily highs and lows, but during busy advertising periods like Christmas, Thanksgiving and the Super Bowl, peaks can occur at any time. These peaks require a substantial increase in infrastructure resources like CPU and bandwidth.
This growth in resource requirements, plus the unpredictable peaks of bids/requests that run through adtech platforms, mean that adtech companies need infrastructure that can scale easily. Hyperscale cloud providers can deliver. But it comes at a price and a premium price at that.
So, what does hosting your infrastructure long-term with a hyperscale cloud provider look like?
Paying-per-use becomes paying through the nose for what you need. It’s quite rare to have an application or service hosted on hyperscale cloud infrastructure that can be scaled down when demand isn’t as high or isn’t needed at all times.
If the adtech company is successful, demand for those services will more likely increase and the resources needed to support them also grows. While paying-per-use might sound like a good option at the beginning, when you consider the reality of your application’s availability and growing resource requirements, it likely very quickly becomes an expensive option.
Flexibility becomes vendor lock-in. If an adtech company starts out hosting their platform on a hyperscale cloud, that platform is therefore developed and built to work with that hosting provider and its proprietary services.
When it gets to the point that costs start to ramp up and the adtech company wants to switch, the money and resources involved in moving to another hosting provider can be very high. Some companies even end up having to bring on additional resources dedicated to simply managing their usage of hyperscale cloud infrastructure and ensuring that costs don’t spiral out of control.
This blog isn’t designed to discourage adtech companies from using hyperscale cloud providers because we recognise that the free credits are very attractive when money is limited. What we want to do instead is point out where ready-to-use proprietary services on hyperscale platforms can be far more inflexible and costly than perhaps originally believed.
Infrastructure-as-a-service provided by bespoke hosting providers with experience in the adtech space can be a great alternative and in the long run, much more cost efficient. Services are provided with fixed costs and more transparency around what’s included in those costs. When an adtech company receives an invoice at the end of the month from a dedicated hosting provider, there are no hidden or unexpected costs.
There are infrastructure providers that have had difficulty offering a support service that meets customer’s expectations. Whether that’s through technical support tickets being unanswered for 48+ hours or providing inflated bills to customers due to unplanned spikes in usage, or even just failing to provide access to someone to discuss these issues. Typically, a commercial cost is involved in order to receive human support on these issues. In contrast, bespoke infrastructure hosting companies provide support that is far more accessible, personal and fast.
When it comes to looking at the future of your infrastructure - particularly if you’ve heavily invested and integrated with hyperscale providers and are beginning to feel the costs and support challenges building - consider having conversations with more bespoke hosting providers about how they can help you optimize your costs and increase performance without a full migration.
To find out more about dedicated adtech server hosting, visit our industry page.