Blog Article
April 28, 2022
The demand-side platform (DSP) you use to purchase campaign inventory will have a direct impact on your company's ability to drive sustainable growth. So how do you know if a DSP will deliver the results you need -- as measured by true business outcomes?
Here are 11 questions to ask any vendor whose DSP you are considering. Their answers will help you determine whether or not you should test the DSP, and ultimately add them to your ad-tech stack.
The purpose of a DSP is to enable you to purchase inventory efficiently, and machine learning (ML) drives efficiency.
More than that, ML is essential for delivering return on ad spend (ROAS). Every ad impression the DSP evaluates and buys for your campaigns should bring you closer to identifying and honing in on the ideal targeting criteria for your campaign.
If your DSP doesn’t leverage machine learning, keep looking.
If the DSP outsources its ML, it’s probably a white-label bidder, which is far less than ideal.
To begin, that white label bidder will be somewhat generic as it is designed to work with all brands that may use it. It will take considerable effort on your part to train it to your unique business needs. It will also take time for it to learn who your ideal audience is, and where to find them -- time you pay for in terms of media spend and ROAS.
Additionally, the digital advertising ecosystem has seen seismic changes as a result of emerging privacy regulations and changing consumer attitudes. A DSP that maintains a ML team can respond to market conditions faster.
ML is powered by neural networks -- i.e. networks of computers (aka neurons) that mimic the human brain. Like our brains, these networks capture signals, instantly assign a weight or level of importance, calculate decisions, and assess the outcomes in order to make a better decision the next time around. Unlike our brains, however, neural networks can receive and process more signals than a human brain can.
The more neurons a network has, the bigger the brain that’s assessing impressions and predicting which ones will deliver the business outcomes you want. Deep neural networks (DNNs) have many more neurons, enabling it to identify and exploit important nuances that lead to better and faster ROAS.
There are two reasons why deep neural networks drive better campaign performance. First, deep neural networks are designed to learn based on smaller training datasets -- a clear advantage in an age where privacy regulations are curtailing the use of third-party data sets. The ability to learn based on smaller training datasets will enable faster optimization, meaning you’ll see ROAS within two to three weeks, rather than the two months most DSPs need to fully train a model.
Second, DNNs can fire multiple inference models at every bid request. Those models cover a lot of ground -- assessing every aspect of an impression, including the operating system of the mobile device, time of data, IP data, and channel. The deep neural network knows the proper weight or importance of each answer to your particular campaign. It then combines all of this insight (on a sub-second level) to predict the value of the impression with regards to delivering actual business outcomes.
In the case of Moloco Cloud DSP, our ML sees 350 billion impressions per day. As a result, you get unprecedented scale, which, in turn, means you won’t exhaust your audience pool and won’t suffer as much from diminishing returns.
This question is a bit in the weeds, but it’s important for delivering the results you need. We want your ML to be as smart as possible, which is why Moloco asks all of our advertiser clients to share their postback data from all of their partners, not just us. If we can see all of your campaign results, we can deliver ROAS faster.
You can judge the efficacy of a DSP’s ML solution by asking the vendor about its “output rate” or how quickly its ML learns. If you’re planning a year-long campaign, waiting three months to see ROAS may seem like the industry norm, but it’s an unnecessary delay. ROAS should come in weeks, not months.
Here again, a deep neural network that uses your first-party data has an advantage. It won’t need to winnow out extraneous learnings from other brands. Your first party data -- the attributes and conditions that are likely to indicate a user has the potential to be a whale -- are the starting point, and deep learning ML never loses focus of it.
Tip: Most DSPs won’t share results of specific campaigns, but they should have CPA, ROAS and CPI metrics by industry vertical.
As mentioned above, the goal of a DSP is to let the brand purchase inventory efficiently, and a bid optimizer is essential to accomplishing that goal. Moloco’s bid optimizer maximizes ROI by better predicting the win probability, a feature that is extremely relevant as more and more auctions move towards first-price auctions. Factoring in market competition and win probability will be essential to secure the impressions and users brands need to grow their customer base.
Many DSPs will optimize the advertiser’s budget on a daily basis, but it’s equally important to be able to optimize it within a longer time frame (i.e. a weekly budget allocation). Rather than meet daily spending requirements only, the DSP should identify when users are most active and inclined to convert, regardless of the day of the week.
For instance, app installs occur when people have more time to explore and consider their value. In fact, iOS installs are 25% more on weekends than weekdays. Android installs are 6% higher.
Moloco’s Weekly Budget Optimizer looks at how your campaign performs, and optimizes ad spend for both time and day of week. Our ML engine will focus ad spend on the days and day parts when your users take the desired actions.
Brand safety and suitability are key concerns to all marketers. You don’t want your ad to appear next to questionable content, or content that simply doesn’t reflect your brand values. Nor do you want to purchase fraudulent impressions, even if you ultimately don’t pay for them. Only real people will convert, and the more real people who see your ads, the more you can build your customer base.
This is why IAB Gold Standard 2.0 certification is table stakes. To achieve certification, companies need to demonstrate their commitment to brand safety.
Industry awards are a recognition that a technology has been assessed and rated by independent evaluators, and deemed the best in the market.
For instance, Moloco has received a SMARTIES X award for its work with Korean shopping platform GS SHOP, which celebrates companies and technologies that result in significant business impact for brands, agencies, media companies, and technology providers.
Brand safety and brand suitability are top concerns for all marketers. As an advertiser, you should be able to see where every impression lands. Ask your DSP if they share log-level data, and in what formats it can be downloaded. Log-level data can help marketers develop strategies based on analysis of their creatives, unique impressions, and quality of traffic for example.
Does your DSP have a creative selection model and A/B testing capabilities? For instance, Moloco has a machine learning feature that automatically allocates budget to the best performing ad formats to help advertisers build high-performing creatives, and to analyze and refine the advertiser’ existing creative strategy.
Additionally, many app publishers prefer to do much of their ad creation in-house in order to drive cost efficiencies as well as iterate on ads quickly. A DSP that offers creative support and expertise will help you to achieve those goals.
Moloco Studio is powered by multi-disciplinary experts from around the globe who are focused on combining creatives, technology, and a surplus of data insights to ensure we outperform even the most demanding performance metrics.
Finally, we have a tip for comparing DSPs: all DSPs ask you to start with a running budget and campaign flight time. When testing a new DSP against your current one, be sure to use the same test budget and timeframe so you have an apples-to-apples comparison. If you spend more with one DSP you will get biased results, in that the one with the bigger ad spend will have more opportunity to deliver results.
Download the playbook to scale your fintech app beyond walled gardens by tapping into the full potential of the open internet, optimizing ad campaigns, and driving business growth.
Moloco’s internal data science team reveals 2024 industry trends and strategies for in-app purchase games and how top performers are driving growth through diversified acquisition and advanced monetization.
Learn how Moloco's in-app bidding SDK helps marketers achieve better ROI, enhanced targeting, and greater transparency in ad rendering on the open internet.
Moloco SDK offers a new direct path to supply for publishers and marketers, streamlining ad buying and enhancing returns. Officially certified by AppLovin’s MAX and Unity’s LevelPlay.