Critique my (15 page cheat sheet) on my systematic approach to mobile campaigns (16)
Critique my (15 page cheat sheet) on my systematic approach to mobile campaigns
Hi guys my name is Paul.
To keep things really brief, I am brand new to PPC and coming over from Blackhat SEO. I am also an avid chess player and my best friend is a National Chess master--he has taught me the importance of developing a system, building on it and using it to act as objectively as possible.
For the last month I have scoured these forums trying to collect as much information as possible to then use and create a flow chart of all the elements that are important to be objective in PPC.
When I joined this forum I remember reading the follow-along of newbs vs. veterans and the big difference I noticed was that the veterans were far more systematic. When problems arose they developed hypothesis to why and then tested them--I was particularly interested in the questions that arose in veterans heads because I know they are based on years of experience and are what leads to their success.
I've done my best to create a "system" for PPC (based on the Main-Course) and what I'd love is if you guys could read it and let me know where I can add more and improve. If you can help me further develop the troubleshooting section because I think that is where the real, "pot of gold" is at.
I thought I'd share it so you guys could benefit from it as well.
Enjoy
Pay Per Click Flow Chart
By Jordanfan20
Psychological rules:
1. Stay systematic and statistical. Trust the process. Don’t let emotions come into play
a. Avoid overspending because of initial success.
b. Avoid quitting early because of a lack of success early on. Remember 10x payout rule for testing.
2. Never stop optimizing a good campaign further.
3. Keep with it. Let yourself fail one thing before moving onto another. And learn from failure.
4. Don’t optimize too soon. Focus on big picture first and become profitable before a/b testing minor things.
5. Never ask yourself if you think something will work. Just TEST!
Picking an Offer:
Use these criteria for picking an offer. Make sure you have at least 2 yes’s. Read newsletters from traffic source to find new offers.
Multiple networks Yes | No
Within testing Budget Yes | No
(Starting out) High volume on Network Yes | No
Is the offer scalable in multiple countries Yes | No
Verify Offer with AM (Get additional info)
Now that you’ve picked an offer ask your AM the following
1. What are some working angles/ideas?
2. What are geographical factors related to the offer?
3. What are some unique selling points of the offer?
4. What offers are hot?
5. Check what offers are being advertised on your traffic source
** Do not pick an offer just because you think, "oh, this looks good." Don't waste time randomly going through offers.
Angle Creation
Create a list of 15-30 potential angles for the product
Start with copying
1. Copy related offers (angles, banners and landers) on Whatrunswhere.com
2. Write down AM’s ideas from above
Basic Angle list creation
1. What benefits does the offer have?
2. What pain or problem does the offer solve?
3. What attributes linked to this to culture or country?
4. What subconscious needs could the offer affect?
5. What events could this be tied to? (Example: Christmas, World Cup..Etc.)
6. How will the offer affect the user? (Example: Faster Phone)
7 deadly Sins Technique:
Write each deadly sin down and come up with a spin of it for your offer.
Wrath, Greed, Sloth, Pride, Envy, Gluttony, Lust
Emotion Brainstorming method:
Fear & Emotion based banners are best..
+ Anger+ Disgust+ Fear+ Happiness+ Sadness+ Surprise
And then go even further:
Acceptance,Affection,Aggression,Ambivalence,Apathy ,Anxiety,Boredom,Compassion,Confusion,Contempt,Dep ression,Doubt,Ecstasy,Empathy,Envy,Embarrassment,E uphoria,Forgiveness,Frustration,Gratitude,Grief,Gu ilt,Hatred,Hope,Horror,Hostility,Homesickness,Hung er,Hysteria,Interest,Loneliness,Love,Paranoia,Pity Pleasure,Pride,Rage,Regret,Remorse,Shame,Suffering ,Sympathy
And for each, brainstorm what kind of images/headlines you can attach to them that relate to your product.
TIP: You can type in each into images.googlle.com and see what comes up, to build your list after you have that list it’s just a matter of firing up Photoshop and doing some work for each, then test testing them.
Angle is created: Make sure it is a good value-proposition
1. Does it answer the question, “If I am your ideal prospect why should I buy from you instead of someone else?” (Does it have an only factor?)
2. Is it specific and could it be quantifiable? (or at least more concise)
Use heuristic to quickly look and see if it has any final problems.
RC (Value force – Cost force) = Net force
RC=Clarity | Credibility Value force= Appeal/Exclusivity Cost force= Mental effort+ Material
Appeal= Relevence+Importance+Urgency
Problems?
No (continue) Yes (rework or grab a new one)
Banner Creation:
1. Go to traffic source and determine what sizes your banners should be.
2. Copy relevant banners from Whatrunswhere.com and modify those using techniques below.
3. Write 1 or 2 sets of ad copy and keep it the same throughout the first set of banners.
Create 10 banners for each Angle with as much variety as possible.
-Out of the ten have 4 proven banners (from what-runs-where) that are modified, 3 that are proven high CTR banners (from STM), 3 completely original banners. Make the majority static at first then when profitable try more animated ones.
Use chart to create as much variety of banners as possible:
http://media02.hongkiat.com/cheatshe...-ad-design.jpg
Banner Principles:
Copy: Make a promise in exchange for a click
1. Skunk and White: power words positioned earlier. (top left)
2. Concise and clear. Use slang only if targeting a specific location. Precise modifiers.
3. Use quantitative measures where possible.
4. Avoid overused marketing drivel
5. Write to where they are in their thought sequence
6. Power words like “free,” and “get.”
7. Try some ads with words that excite, "Wow, Crazy, Insane, Earth Shattering, Mind Blowing."
Buttons
1. Use CSS generator to make fast buttons
2. Consider button on ad and page, not just in a vacuum.
3. Use action words, “download.” Don’t’ make the next task anxiety provoking.
4. Check the top 3 performing submit buttons on Peerfly.
Font:
1. Use spy tools to find cool fonts.
2. Use Google chrome extension to add text to a page and then screenshot it to get access to the font.
Image
-To find high CTR images Google, “keyword” + crazy, wtf, fighting, weird, weird hat, bleeding, gun, kissing, club.”
Translations if necessary:
1. Convert words to basic synonyms.
2. Consider using slang.
3. One-hour translation or FIVER.
4. Consider using local landmarks.
Banners are done. Produce 1 or 2 landing pages per campaign.
Landing Page Optimization and Building Principles
4(motivation) +3(strength of Value proposition) +2(incentives-friction)-2(anxiety)= Conversion Rate
Friction= length/difficulty, complexity of eye-path
Anxiety= Psychological concern
**Take everything with grain of salt until tested. Remember there are a lot of converting tactics; you will have to limit them on mobile. Use formula as a way to hypothesize improvements and originals.**
Steps
1. Keep one goal throughout. Use methods to enhance that goal. Get rid of whatever doesn't.
2. Ignore user motivation. It is outside of control via landing page.
3. Communicate Value Proposition in headline and carry it out throughout each element of the page. (View angle creation to find value proposition.)
Headline Creation:
1. All marketing messages must be centered primarily on the interests of the customer. Therefore, when it comes to crafting headlines, emphasize what the visitor gets rather than what they must do.
2. The goal of a headline is similar to the goal of the opening scene of a movie — to arrest the visitor’s attention and get them into the first paragraph. Therefore, utilize a “point-first” structure (i.e. place the value at the front of the headline).
3. Don’t sell too hard with the header. Use the header to direct the reader onward.
I. The headline draws your reader into the next line (subhead).
II. The subhead connects them to the first paragraph (copy).
III. The first paragraph engages them with your offer.
4. Connect dots. People read passively don’t assume they will connect the dots. (E.G, “A 400 HP engine.” Say, “A 400hp engine, that means it’s really fast.”
Body Copy:
Write for your audience:
Slang sites:
· http://www.slanguage.com/
· http://www.coolslang.com/
· http://localspanish.com/
· http://www.peevish.co.uk/slang/
Ex-Pat Sites:
· http://www.transitionsabroad.com/lis...websites.shtml
Principles:
1. Be as clear as possible (precise modifiers)
2. Be at the right place in the customers thought sequence.
Long vs. Short
· High commitment and high anxiety =long copy
· Low commitment and low anxiety = short copy
Image:
Does the image help promote the value proposition?
· Is the picture relevant?
· When you look. What do you learn?
· Does the image help visualize the offer?
Some image resources:
· http://www.deviantart.com/
· http://www.worth1000.com/contests/
Incentives:
Always assume there is a more optimized incentive. Once you are profiting always further optimize incentive.
Forms of incentives
· discounts
· content: free books, articles, emails
· order/delivery incentives
· accompanying product incentives
· accessories
· complementary products
What makes a good incentive?
· exclusivity
· perceived value (consider what people in that society are conditioned to value)
Friction:
1. Draw eye path and make sure it is simple and obvious. Consider the linear geology.
2. What is the tone? Is the tone repulsive or draw people in?
Anxiety:
Main anxiety concerns and how to deal with them:
· Quality: satisfaction guarantee
· Reliability: Testimonials
· Security: Third-party seal
· Price: low price guarantee
Use different types of testimonials to reduce different anxieties:
· ease of use
· product quality
· customer satisfaction
· cost justification
· admit weakness
· only say verifiable truth
Place anxiety reduction factors around locations on the page that produce that anxiety.
Buttons:
Index fingers are at least 45-57 pixels: Thumbs are 72 pixels wide
1. At least 70 pixels wide and 45 pixels high
2. More specificity tends to do better
3. Don't consider button in a vacuum
Speed Tips:
1. Eliminate Java Script and CSS from header. Keep CSS in same file.
2. Keep Jpegs to around 20kb
3. Before using images run them through smush-it
4. Test on Pingdom.com
Colors
Color theory: http://www.smashingmagazine.com/2010...ning-of-color/
1. Various shades of color are used to emphasize something
2. Color shouldn't draw people out of the form.
3. Use color to draw the readers eyes to what is important
4. The contrast between background and text needs to be high. Contrast is important to web design.
5. Colors have respective meanings associated with them.
http://wordpress.tis.edu.mo/shevonwi...y-colors-2.png
· Hue is color (blue, green, red, etc.).
· Chroma is the purity of a color (a high Chroma has no added black, white or gray).
· Saturation refers to how strong or weak a color is (high saturation being strong).
· Value refers to how light or dark a color is (light having a high value).
· Tones are created by adding gray to a color, making it duller than the original.
· Shades are created by adding black to a color, making it darker than the original.
· Tints are created by adding white to a color, making it lighter than the original.
Mobile Site Final Checklist:
5 key ingredients checklist (brief how-to’s.)
1. Does landing page auto resize, so it properly fits all phones?
2. Is there anything that distracts from the main idea?
3. Are you holding the users hand throughout?
4. Are you giving the user a sense of urgency?
5. Did you use words like “instant,” “get it right away, ”get it now,”
6. Did you use the word “free,” if you are selling a free concept? People love the word.
7. Is the landing page VERY small file size. Your page needs to load ninja fast 200 m/s goal. Bigger pages send to Wi-Fi and smaller send to 3g and below.
8. Limited text/images etc... Screen space is limited, don't overload the user
9. When creating landers on mobile, make sure you code links to open on the same page.
To do this use
<a href="http://yourlink.com" target="_self">Click here</a>
10. Does lander fill screen? Add to header if not:
<meta name="viewport" content="width=device-width, initial-scale=1.0, maximum-scale=1.0, user-scalable=no" />
6. Did you take advantage of Ad handset detection? Instructions if not:
1) Signup at www.handsetdetection.com & Login
2) Create a website profile with no redirection rules.
3) Grab the basic Express (JavaScript) snippet to paste into the <head> section of your web page. It'll look like this
<script type='text/javascript' src='https://localhost/sites/js/xxx.js'></script>
where xxx is your site id.
4) Access the information via the HandsetDeteciton object. Like this
<script>alert(HandsetDetection.vendor + " " + HandsetDetection.model);</script>
Use jQuery or JavaScript to put the vendor and model into your webpage.
Need to redirect a certain type of phone:
Java script for resending phone: Replace links with other landing pages
<script language="Javascript">var h=window.screen.availHeight; var w=window.screen.availWidthif(h=="1" && w=="1") document.location="http://google.com"; elseif(h=="2" && w=="2") document.location="http://yahoo.com"; else document.location="http://anythingelse.com";</script>
Ask AM to Quickly Review the Campaign.
Setup Campaign
1. Create 4 campaigns per angle based on phone setup. (E.G Offer-Angle-Wifi-Site)
2. Label everything clearly. (Fill in “offer” with offer name and “Angle” with Angle name)
3. Add in 10 banners for each Angle.
4. Make 7 offers direct to LP and 3 to direct offer.
5. $7-10 daily spends per angle.
6. Enter 1.5x payout of offer for bid.
7. Ask A.M to Review one last time.
8. Run personal test and make sure tracking is working.
9. Let Angles collect $14-20 spend
Spend #1
While collecting impressions the $14 spend watch for and cut
Cut any of the following if this occurs during the first $14 spend
• 5000 impressions on a banner and no conversions
o Unless the banner has a really good CTR
• Cut a placement if it is getting 50% or more of the total impressions.
o Unless it is converting really well
• Cut if placement has really high CPC both 3x average and 2x payout spent
Now that the $14 has been spent check each Campaign for ROI
Ø Greater than (-50%) ROI Place in Promising
o Jot down some reasons you think this campaign might be doing well and things that can be improved.
Ø Less than (-50%) ROI Place in unpromising
o Jot down some reasons this campaign might not be doing well and some things that could be improved.
If both go in unpromising consider changing offers. Talk to AM or STM and decide whether campaign might have potential with new angles.
Promising campaigns before next spend:
1. Cut Bad Banners performing banners and replace with new ones.
2. Run in ROI sheet and make sure the ROI reaches goals. (Shoot for break-even early on.)
3. Cut either direct link or Lander.
4. Create 3 more LPs. High variation.
5. Check tracker stats and review most used placements and make sure all are doing well. Google Chrome Developer Tools or BrowserStack. Resolution and normally functioning.
Unpromising Campaigns before next spend:
1. Look for good stats in placements
a. Good device/carrier
b. Good lander
c. Good image
2. If found cut down to just good element
3. Replace failed landers and landers
Spend #2:
While running spend #2 cut if
1. 5x payout without a conversion
Check bids midway:
2. If winning 90% of bids, drop bid down
Review after next $14 spent for Spend #2
1. Test landing pages against each other using Bayesian Inference A/B
a. http://www.peakconversion.com/2012/0...al-calculator/
b. See if anything is winning 90%. If so cut other ones. Replace with 2 variations and completely new landers. Do n
2. Cut banners that have -.5 ROI and clear bad offers with -.5 ROI. Replace
3. Cut carrier and handsets that are underperforming
4. If unpromising is still negative then for good.
5. If promising goes bad then give it another spend.
Optimize Banners
1. Create variations of top performing landers and banners. And split test after every spend.
a. Test new images
b. New headlines
c. New CTAs buttons
d. New Background colors/images
e. Anything else that is relevant.
Scale:
Requirements: 30% ROI for 3 days
1. Double Spend daily budget.
a. Continue until ROI drops significantly
2. Ask for Pay bump or faster payments
a. Tell them you need a pay bump to scale
3. Duplicate into other GEOs
a. Fiver or Onehourtranslation.com
b. Run campaigns to $25 spend and see if they maintain a positive ROI
4. Duplicate on more traffic sources
a. Run campaign for $25 spends and see if it maintains positive ROI.
Repackage:
o Add time in (E.G Popular in 2014)
o Hack up logo or brand
o Change up new angle
Troubleshooting:
What you should do in the following scenarios.
If good ROI campaign goes bad?
Wait and see if it stays bad for a week. Ask AM if there could be an explanation for the sudden bad ROI. If not, drop the offer.
Not getting enough traffic?
Is there a bad CTR?
1. Does banner ad make a specific quantitative claim
2. Do you provide a compelling reason to click?
3. Does the image elicit a response?
Is there a bad CVR?
Recreate the experience(the funnel) in your own mind and attempt to reinvent the questions that the customer will ask based on different levels of motivation (I.E viewing from an app, site, Wi-Fi). Use the following questions below as a logical tool to come up with your answers.
You need to get inside the mental conversation they are having. Answer the following:
1. What expectations would you expect if you clicked on the ad?
2. What motivation is the person likely to have?
Are you orientating the viewer within 7 seconds?
1. Where am I?
2. What can I do here?
3. Why should I do it?
Does the page follow these principles?
1. Congruence: Every element of the page expresses or supports the value proposition
2. Continuity- Ensuring that every step of the conversion process expresses or supports value proposition
3. Does every component of the LP carry the value proposition into the landing page?
4. Are you utilizing the headline of your page to immediately inform the prospect where they have arrived?
5. Is your landing page objective clear?
6. Are you starting with broad stuff and then getting more direct?
Does it satisfy the following?
1. Principle of alignment (how it aligns with what is on the page)
2. The principle of timing (is the call to action too soon. Where is the person in the buying phase?)
3. The principle of absorption (Is there too much information)
4. Principle of negation (Too much, simplify)
5. Principle of redundancy(Are you restating the value proposition?)
Hey there,
I like the way you are approaching this.
Just a quick note: PPC is a payment / bidding model and not a traffic source.
PPC can thus refer to search, display, social, mobile and many other digital channels and traffic sources.
Much of what you write here seem to be more applicable to PPC for mobile display inventory on DSPs and exchanges. Some of this would not apply at all (and indeed would be wrong) for other channels.
In addition, I know a lot of people on STM talk about the need to test with a budget of 10x payout or 3x payout or whatever. Honestly, I don't know where this comes from. It may just be a heuristic rule of thumb, but as a statistical methodology, it is not correct.
Originally Posted by cmdeal
Do you have any recommendations on a better test budget that is based on stats. Or is their a way to calculate an optimal test budget.
The test budget should be determined by the the sample size you require in order to get to statistical significance for your target conversion rate.
Basing your test budget on a crude multiplier for the payout may or may not be useful as a rule of thumb, but methodologically, it is wrong. The probability density function for the conversion rate of any campaign and channel is independent of any payout amount. The probability of winning at craps in Vegas does not change if you play at a $10 minimum bet table instead of a $1000 minimum bet table. The dice don't care about your payout.
@cmdeal - the payout multiplier approach takes an implicit assumption of ROIs approximately similar to those of other affiliate campaigns. Within those assumptions (ROI in the range +30% - +250% or so), it seems solid to me. Obviously it breaks down if you're only looking to test for very large ROIs or if you also wish to include small ROIs.
However, I know a few people have asked about this in the past, so I'm going to write up a thorough investigation - a "paper" of sorts - into the appropriate sample size for testing affiliate campaigns. I'll be interested in your thoughts and comments on the statistical methodology - particularly if it's wrong, as if so I want to know ASAP 
EDIT - It also breaks down, or at least has to be heavily modified, if a source has more or less confounding variables. I'd generally recommend a much higher multiple of payout to test an offer on mobile because of all the additional variables muddying the data.
Originally Posted by caurmen
However, I know a few people have asked about this in the past, so I'm going to write up a thorough investigation - a "paper" of sorts - into the appropriate sample size for testing affiliate campaigns. I'll be interested in your thoughts and comments on the statistical methodology - particularly if it's wrong, as if so I want to know ASAP
You can think about this intuitively.
- If for one day, an advertiser decides to double the payout, that doesn't mean you now have to double the amount of spend testing.
- Or, if one day, someone (a beginner newbie) grossly overpays for his/her media buys and gets to 10x or 3x payout of an offer with just 2 clicks and getting nowhere near statistical significance, this does not mean that his/her job is done.
- Similarly, it is not hard to imagine that a more advanced "superaffilate" uncovers a ridiculously underpriced media channel and gets to statistical significance while just spending even a fraction of the payout amount.
At the campaign level, the target conversion rate and the results you get from your samples determine the the size of the tests you need in order to get to statistical significance. But this can amount to 100x payout, 10x payout, 1x payout or even 0.1x payout.
The amount of payout does comes into consideration a priori as it determines a required conversion rate you need in order to pay for the cost of a specific media channel in order for it to be profitable for you. If, for example the payout implies that you need to get 120% conversion rate for a certain media channel, you know that this is something not worth doing. In the same vein, you probably want to think twice if the payout implies you need even a 25% conversion rate on a channel such as Facebook.
Many people (including myself) often rely upon simple rules of thumb as a heuristic for making quick decisions. If people are using some sort of payout multiple as a rule of thumb, that may be fine. But they need to understand this for what it is.
@jordanfan20, I would turn this into a large poster and print it out then slap it on your wall!
Originally Posted by cmdeal
That being said I was struggling at coming up with a system to manually calculate statistical significance. Primarily, because I don't know how to calculate a representative sample size because I have no idea what the population is for any given traffic source.
Do you guys have any tips to finding an N-value that is representative?
The one alternative method I came up with was to stick with the Bayesian Inference, but the big disadvantage to sticking with it is if I have enough sample data a normal null-hypothesis test would probably be more accurate.
Thanks guys
@cmdeal - I'll write up my thoughts in detail and we can discuss from there! Having said that, I do agree about some of the implicit assumptions - if you believe that your click costs are likely to change significantly after you do the test, that definitely invalidates it.
Likewise if you expect to be able to get massive bumps on your payout.
It's definitely a rule of thumb. Gathering data, running a binomial confidence interval calculation and continuing the test from there is more accurate, but largely because it makes the implicit assumptions explicit.
@Jordanfan - I'd usually just use a binomial confidence interval method to calculate expected CTR / CVR - is that what you're looking to do or do something more sophisticated?
@caurmen No I think I'm just a little confused about the binomial confidence interval. When computing it, do you need to have an estimate of the population size that is being sampled? Or does the binomial confidence interval calculate a independently of knowledge of the population size?
Originally Posted by cmdeal
Originally Posted by zeno
Great idea!
@jordanfan - you're calculating the interval of probable values given the results of a set of experiments, so you just need number of trials and number of successes. It's particularly well suited to AM for that reason - you don't need pre existing knowledge of the environment in which the experiment is running. Obviously the usual limitations ( np > 5 and n(1-p) > 5 as I recall) still apply.
Home > Mobile >