My M@bile Results (13)
Hi All,
I was hoping someone could take a look at this campaign stats I ran. There are 4 campaigns targeting...
WifiSite: spent $30
WifiApp: spent $30
MobileSite: spent $15
MobileApp: spent $15
I've included the top few placement stats that matter. The last 3 columns are from my tracker, which tracks conversion and actual clicks received. I tried to blacklist placements that have spent more than 3x payout. Some of them spent more than that as can be seen.
These are the stats for my banners. It also calculate how far from the average each banner performs.
I'm abit lost on how to move forward with this campaign. Here are some of my options:
- It's a failed campaign and move on.
- Add a lander to see if it will improve
- Focus on the best performing creative + placement combination and optimize in that direction.
- Give it more budget to see give it a more even spend on the other placements that haven't have as much of a chance to work.
- Try other size banners
Other questions on my mind:
- Some of these placement's are questionable. Should I even bother spending on them?
- The offer payout is $0.34. Does that mean I need quite a high conversion rate for the campaign to be profitable? All my campaigns got under 10 conversions which is less than $3 for campaign spend of $15-$30..yikes!
1. I would make a clone of the campaign and put all the placements in there with potential (just make one clone, IMO no need to split in Wifi/3G and Mob/APP)
2. This new 'Target' campaign needs to be optimized by creatives. Let it run, see if you can find a clear winning angle and optimize further bases on that angle
3. By this time you'll have some data to optimize on device type or OS version and on connection type.
4. You can let your RON campaign run to keep finding more placements for in the Target campaign. Make sure to blacklist the heavy losers and the ones which go in the Target campaign
Your current banner data is useless IMO because it doesn't have conversion data. Make sure to collect enough conversions first.
In the current data dump format, I don't know how you (or anyone else) can even begin analysing your activities.
- I don't see any P&L anywhere. How can you analyse how much money you are making?
- Where are important metrics like conversion rate, cost per conversion, revenue per conversion?
- And please, for the love of God, use the Excel sort function and sort your data.
There are some halfway decent campaign Excel templates floating around STM ... you should track these down and use them.
Originally Posted by keepthecar
I've done what you mentioned with putting potential placements in separate campaign in the past. I was just trying to use the method described by Caurman with splitting into wifi/mob/app/site.
What do you mean by "2. This new 'Target' campaign needs to be optimized by creatives."?
How would you decide which placement has potential? Either by:
- ROI? (eg. -50% and above)
- Spend? (eg. 3x payout)
- Conversions?
- Clicks?
Be prepared for the ROI to drop of the placements you put in the target campaign. Often your sample size isn't large enough to determine whether the placements are profitable
You can put negative ROI placements in here as well if you are convinced that they can be optimized heavily
Originally Posted by keepthecar
Yes in that case I would move on to something else
I'm currently seeing this in my Decisive reporting stat. "Undefined" in my Site placement data sucking up my budget for the campaign. The response from Decisive support was to black list it, but it's not available to blacklist.
Has anyone had any experience with this?
Originally Posted by quantum27
I'm currently seeing this in my Decisive reporting stat. "Undefined" in my Site placement data sucking up my budget for the campaign. The response from Decisive support was to black list it, but it's not available to blacklist.
Has anyone had any experience with this?
I'm currently split testing LPs for campaign of mine. From my understanding of this post
http://stmforum.com/forum/showthread...-Landing-Pages
I should start collecting data until I get 5 conversion for each lander to see which one to cut.
What about the placement, and how often the LP was tested on each placement. Maybe some LPs work better on certain placements, and some worse. Should there be enough data to collect for each placement testing between the different LPs.
Or maybe, we just take all the stats from all the placements to see the winning LP.
Unless you are doing serious volume, optimising at the placment>lander level is going to seriously slow you down.
Look at the data, sure, but make decisions at the campaign>lander level unless there is a massive outlier in the data that is worth your time.
I'm trying to run a modified version of the Mobile Cookbook on pop traffic.
Currently overall numbers are:
Revenue $6.16
Spend $14.04
PnL -$7.88
ROI -56.13%
Payout $0.56
My Split test result of 3 LP show's LP2 having less than 10% of being the best from split test calculator.
LP1
Views: 732
CTR: 23%
Conv: 5
CVR: 0.68%
LP2
Views: 732
CTR: 6%
Conv: 2
CVR: 0.27%
LP3
Views: 734
CTR: 9%
Conv: 4
CVR: 0.54%
From my initial $14-$15 spend here are my questions:
1. It's currently sitting at -56% ROI, but I'm not sure if I'm testing enough placements to say if it's a "promising" or "unpromising" campaign for me to either keep going or look for outperformers.
2. I'm currently cutting placements at 3x payout spend which is $1.68, that's only enough to test 9 placements fully ($15/spend each stage), not to mention the small amount of spend scattered amongst other placements. What I've done in the past, is not worry about overall spend each stage. Just spend enough on each placement, and pull out placements with -50% ROI to a promising campaign to optimise.
3. From what I can see, I should cut 1 LP and add a totally new one. How many LPs should I test at 1 time, because the more LPs I test the less accurate I think the data will be. Because the spend would be more spread across the different LPs, making the data less statistically significant.
I'd recommend testing 2-3 landers at once if you're going for relatively small spends. Bear in mind that once you have statistical significance, getting more data doesn't actually help very much - it's essentially wasted money.
Are you using the Bayesian split-test calculator? If not I'd recommend doing so.
LP1 looks to be a smidgen above -50% ROI from my rough calculation - given that, it's definitely worth keeping testing the campaign for a bit. Keep killing placements, testing new landers and offers, and see where it goes!
Home > Mobile > Follow-along Campaigns