Time has come to speak about DSPs aka Programmatic aka the revenue machine ! Even if programmatic sounds a bit like the word “problematic” trust me it should make you think about magic.
Wait of what are you speaking about?
Well, I am speaking about programmatic. This amazing channel fully automated with a fancy Machine learning that provide you faster than any UA manager on earth the best revenues thanks to an access into the vendor-neutral real time bidding ecosystem !
Programmatic is splited by:
- Fully managed Agency ( I love a lot of folks there such as Aarki, Liftoff, Chartboost etc.)
- Managed partner self-service (my fav. Kayzen because they are amazing)
- Bidder as a service platform ( Beeswax for instance)
- Fully in house DSP ( any amazing UA director that managed to get learning from the folks above and then that pushed to management this crazy good idea and get it approved)
Ok but i have google why do I need a DSP?
Having Google is like beeing Parisian ; It is performance oriented, super crowded and everyone knows it. The reach is limited but still when you speak about it you usually reach your target. However… having a DSPs is wider. If you have a DSP you are french you-are-not parisian. You open the silo and just access a crazy long tail audience that you never reached before with crazy potential.
Wow, sounds great but how does It works?
So usually, at the start it does not work. And this is where a lot of people struggle; during this exploration phase. The Phase 1 is from far one of the most hard to go threw especially when you launch your first test. The exploration phase is not even going to optimize tower install but impression across a wide network of SSPs. As you pay to get impression, be strong ! It is not the time to pause or change things ; just time to learn.
Once you have learnt, Phase 2 is the install phase. At this phase you still won’t get this magic revenue but at least you will be able to say to your manager your CPI decreased yay. This install optimisation will reduce the amount of SSPs your are spending in and be the first light at the end of the funnel ; that here represent the revenue.
Last Phase, my favorit is the event optimisation phase. This phase is based on dynamic prediction based on the cost per event or purchase goal. When you reach this phase in you are usually happy.
First launch what can I do?
Except psychologically make your manager ready to see 0 in front of your ROAS ; I think you can do a few tweaks to hack the algorithm.
First, share your data. Your CPA is calculated with attributed and non attributed data and the more data you share the better to get smart so If you want to make all chances yours ; you will need to trust your partner or just avoid working with them.
Second, choose your best Tier Geos but with a significant audience size. Make sure the first impression you give to the algorithm is the good one and choose geos where you are more likely to have faster positive postback events.
Third, choose a strong creative concept. Do not do a creative testing or a seasonal creative on your first DSP campaign. Keep in mind you do not know how long is going to be your learning phase and the main principle will be to not touch until you do not end to learn. So, if your winter creative is the top performer and you launch it in december you maybe won’t be able to pause it for a long time (hopefully before summer).
Last but not least, do not change your budget ! Until you are learning, keep it flat and stay strong 🙂
But should I bid high or low then?
Well, ask first yourself if you are first or second price auction because this will change for sure the way you by on DSPs. Mopub switched few months ago and It can rapidly be more expensive that what it use to be by the past on the ua side.
Second, ask yourself who is in front of you. If you know your competitor is rich and the ARPI of your users are crazy high ; better bid high to be sure to win the bid.
Keep in mind, your bid price depends on your target CPI
multiply by your predicted CVR ( that’s why your creative is so important) by the ML factor. IF the ecpi < tcpi then the Factor is > 1 and your bid should be more aggressive. If the opposit, your factor is <1 and automatically it won’t be aggressive.
Last but not least ….
Programmatic is a matter of time and usually cost quite a lot. If you want to invest in It, I would advice to give it a try with a manage service (that worked with your competitors) BUT that is transparent. I can not tell It enough ; there is no point to work with DSPs that are blackbox. First, because long term your goal should be to use this data to build your own DSP ; second because if you trust a partner the partner should trust you especially when the investment is so high. I hope you are convinced and about to invest in this amazing channel !
