Mobile Campaign - Goal break even & learn (3)
Hi Guys,
I've carefully read a lot of the guidance material on STM and also followed campaigns and posts over the past week. I've put together a rough campaign "skeleton" that I want to share with everyone and get the experts' view, feedback and advice on before I implement this. I have also inserted a few questions in relation to a few things I am not 100% sure of.
Take your time to provide your feedback as this is quite detailed and please let me know where I am going wrong...I hope it helps more people than just myself! Enjoy...
Basic criteria
• Vertical – choose 2: mobile – sweepstakes; app download; game download; grocery/shopping coupon
• Offers: 2 top offers in selected vertical
• Offer type: CPA/CPI; SOI
• Market: South East Asia, Eastern Europe, Latin America
• Traffic source: Decisive
• Tracker: Prosper202
• Affiliate networks: use 3 different networks that have same offer
• Budget: $1,500 to $2,000 - Q1 is this budget adequate for the scale of my approach or is it not enough?
Approach
A. Set-up
1. Offers/Verticals/markets/traffic source: ask affiliate manager of each of the 3 networks for best offers given basic criteria and select best 2 offers from provided options that exist on all networks
2. Campaign set-up: Set up 5 totally different campaigns for each offer (total of 10 campaigns) in Decisive and in Prosper202 and on 3 different affiliate networks
2.1 Each campaign to have different angles, targets and LPs, and markets
2.2 Let each campaign reach $15 spend; check frequently if there is an app or site which has more than a third of total spend of total campaign spend since the campaign began, has not had a conversion and has spend more than 3x times offer payout blacklist it” - Q2: Not quite clear on what is meant by “if there is an app or site which has more than a third of the total campaign spend” – what is an app or site in this context? Is this an offer, a campaign setting?
3. Banner set-up: Create 5 totally different 320x50 sized banners for each campaign (total of 50 banners) and set up in Decisive and Prosper202 - Q3: Just want to confirm my understanding with respect to the banners. The banners are linked to each campaign right? So when you develop banners you develop different banners for each campaign rather than using the same banners for each campaign?
4. Bidding set-up: set-up Smart CPA bidding at 1.5 times offer payout and with a limit of $8 per day, unlimited budget
5.Clone campaigns 4 ways (wifi; mobile; wifi_site; mobile_site): so I will end up with a total of 40 campaigns then (5 campaigns X 2 offers X 4 cloned campaigns) Q4: Why are we doing this? What are the different cloned campaigns? Do you have to set up each cloned campaign in Prosper202 also?
6.Landing page set-up: create 3 different landers for each campaign (so a total of 30 landing pages) and set them up in Prosper202 only. Q5: Do you create LPs for campaign and then use that same LP for all the banners in that campaign or do you create an LP for each banner? And, is it correct to only have to set up the LPs in Prosper202 and not in Decisive
B. Optimization
1.Campaign optimization (after campaign has reached $15 spend)
Kill campaigns with more than -30% ROI
Kill campaigns that have made less than $7 in payouts as they are not promising enough to optimize
Launch 1 new campaign for each offer every day
2.Carriers, Handsets, Placements, OS Optimization
Blacklist any carriers, handsets, placements, and OS that have not had any conversions and have spent 3 times the offer payout
3. Banner optimization (after campaign banner is part of has reached $15 spend)
3.1 Kill any banners that have had 700 impressions but no clicks
3.2 Kill any banners that have had 3000 impressions and less than 0.2% CTR
3.3 Kill any banners that have a CPA greater than 1.5times of payout
3.4 After 10,000 impressions for a banner ensure that maximum possible conversion rate is greater than minimum viable conversion rate – kill if max possible conversion rate < min viable conversion rate
3.5 Keep any banners that are profitable running regardless of the CTR after ensuring that maximum possible conversion rate is greater than minimum viable conversion rate
3.6 Replace any killed banners with new ones to always have 5 banners in rotation for each campaign (alternate between basing new banners on exiting banners that won’t be cut and completely new ones)
4. LP Optimization (after at least 5 conversions between all 30 LPs): Q6: Is this “5 conversion rule between all LPs” (which is mentioned in part 11 of the tutorial guide) dependent upon the total number of LPs at all or not? And how “long” do you wait until you have 5 conversions? In other words, after how many clicks and no conversions should you realize your LP is just not good enough?
4.1 Use AB Split Test Calculator to determine approximate probability of being best – select LP with 90% probability; remove LPs with less than 10% probability of being best
5. Choosing the best offer between the 2 initial offers
5. 1 If an offer has had clicks worth 4 times the payout of the offer but no conversions, stop running that offer
5.2 Compute the minimum and maximum EPCLs of both offers, if any of them have a maximum EPCL below the other offer’s minimum EPCL, stop that offer
6. Track revenue and spend every day – stop any campaigns that are not close to making a profit after a couple of rounds of optimization and launch new campaigns - Q7: What is a good rule of thumb for the amount of rounds of optimization to perform before stopping a campaign? Is a couple really meant to mean 2 or more than 2?
C. Scale & Don’t Slack .....(to be continued)
Sorry it took so long to reply, I imagine the post got covered by other bumped follow-alongs (if it goes 3 days without a reply feel free to bump it bro!)
Q1. this is fine for mobile and what you're trying to do.
Basic criteria
So have you picked two verticals? If not ask your AM - say you're interested in all 4 and ask if they can recommend two based on what's doing well at the network.
From there, focus on a single region e.g. SEA that those offers/verticals are doing well in.
When it comes to offers by the way, don't limit yourself to finding the same offer on each network. Quite often if your angle/banners/landers are broad enough you can plug in multiple different offers on the back end - indeed this is a good way to facilitate offer testing.
Approach
A. Set-up
1. Again, don't limit yourself to offers that are on each network if possible. A better situation is an iPhone sweeps campaign where there's actually 5 different offers on each network that could be plugged in and you just rotate 3-4 depending on what's doing better.
2. You should reassess your campaign setup. Ideally, make a new campaign for each angle that you are taking within a vertical (ignoring any trivial splitting by WiFi/app/etc. here). Don't make a new campaign for each offer if they are similar - rotate these with the tracking system. Same with landers - don't make one per offer, make one per angle/vertical, or several, and rotate them. Your goal here is not to make a completely different and customised funnel for every offer or angle, it's to take multiple designs/variations and split-test them all, even if it requires a little tweaking of the ad copy for angle consistency.
Q2. An app or site is a placement at Decisive, as observable in the app or site tabs. A place where your ad has been delivered. If one of these has used a lot of $$ and garnered no conversions, kill it.
Q3. Different banners for different angles and verticals. Again, try to find similar offers that you can reasonably rotate on the back-end without compromising your angle.
Q4. You may be confusing where the campaign splitting needs to occur. Split campaigns at Decisive by WiFi/Carrier and App/Site to segregate their performance. These can all link to the same tracking202 campaign as long as you pass variables that tell the tracker where the clicks came from, e.g. using the {{{media}}} and {{{carrier}}} tokens. This way you can drill down in tracking202 and know which clicks came from what campaign. I'm not sure if Decisive passes "WiFi" as the {{{carrier}}} value when targeting WiFi. If it doesn't you could just add a custom token (remove one that you don't need of the 10) and manually but WiFi or Carrier in the tracking URL for each different campaign.
Again, don't make new campaigns for different offers unless the offers, or the angles you are using, are so different that the same banners/LPs simply won't work.
E.g. if you were advertising DU Speed Booster, Clean Master and some other speed-up-your-phone utility, you could use the same campaign and rotate these on the backend. But you'd need separate campaigns for Zero Launcher and some messaging app.
Q5: LPs get set up in tracking202 only, they are separate from anything Decisive. Don't make separate LPs for each banner. Don't make 3 LPs per campaign. Make 3 LPs for each vertical/offer type and rotate them all. Modify these landers to suit different angles. Don't overthink things here, you don't need to spend time re-doing the same thing over and over to make custom paths for each campaign. This goes against the intuitiveness of split-testing.
B. Optimization
1. I'd cut off at -50% ROI.
2. Sure but focus on placements first as otherwise killing off devices and carriers may be premature.
3. For 3.3., make sure something has 2 conversions before doing this if it's CPA is right in that 1.5x zone and it's had a conversion already. E.g. if at X spend and 1 conversion with CPA of 1.6x payout, a single conversion would then make that placement profitable, so give it a bit more time.
Q6. Look at the earnings per click of the lander and use that as a guide for optimisation. Use that in the Bayesian calculator tool for comparisons. If the landers are at 55/7, 49/5 and 45/1, there is no point waiting for 5 conversions on the last lander when statistically it simply is very, very unlikely to beat the first two.
5. Again, use a calculator to compare offer performance. An offer that has had 4x the payout in clicks to it might be a dud but if the others are performing similarly badly... it could be your angle or the placements sending clicks. These thing should all be relative to the other variations when optimising.
Q7. Depends. If it's at -80% ROI on day 1 with enough data to do a big optimsiation, then -60% on day 2, then -55% on day 3, then -47% on day 4, it's probably not going into the green - at least if you have made substantially optimisations. There's no real rule of thumb as different campaigns will have different variables that have different levels of impact on the campaign performance.
Better to roll with it and present some data so that we can provide advice for the specific campaign.
All the best!
Hi Zeno,
Thanks for the reply and advice! I'll ne implementing this next week and will keep you posted with progress and data.
Thanks a again. Appreciate your help.
Certainchance
Home > Mobile > Follow-along Campaigns