Category: app marketing

Personally one of the ongoing struggles of the app business is maximizing revenues from the various ad networks.  First off I’ll just say that there is no one perfect solution and a big part of this is opacity on the part of the ad networks.

The question is, how do we decide which network is working best?  How do we know if we’re getting what we’re worth from the ad networks that we’re using?

eCPM is the generally accepted model for measuring the effectiveness of an ad campaign (whether as a publisher or as an advertiser).  The problem is that sometimes this fails, which is why I’ve invented a new parameter to measure, eCPI.

eCPI – The New eCPM

eCPI means Effective Cost per Install.  You can work this out by dividing total revenue by total installs. I believe this is an important new parameter to measure to make sure you are using the best performing ad network and optimizing your ad revenue.

The best thing to make my point is to give you an example.

Here are some stats from an app of ours on Revmob.  You can see in the second column from the right that the eCPMs are pretty good, right?  $4-5 eCPMs are pretty high at the moment across various ad networks and $16-17 revenue per day is not bad for just one ad network, so if we were just looking at these stats we’d be jumping for joy (or would we?).

The thing is though, how much are we actually getting per install?  Without a doubt most advertisers are paying using a CPI (Cost Per Install) model.  Chartboost also helpfully publishes stats on average industry CPIs around that world so you can easily get industry benchmarks.

Just so you know, here are the current stats for November 2013:

Chartboost network CPI iPhone:   $1.65

Chartboost network CPI iPad:  $2.06

iOS US CPI:  $2.12  (with the range from $0.88 to $2.82)

So, if we crunch the numbers for what we’re getting on this app, we find out the eCPI is a shockingly low $0.43 (to get the eCPI divide the revenue by number of installs, in this case $16.60 / 39).   If we had been measuring eCPI alarm bells would have been ringing instantly that we are getting paid far below any acceptable industry benchmark.

How much should we have been getting?  If we take the network CPI for iPhones of $1.65 and multiply this by the installs we generated (39), we SHOULD have been generating revenues of $64.35 for that same game, which would work out to an eCPM of $17.79.

Impossible you say?

For comparison, on that same day on another ad network (Applovin) in the SAME APP we had an eCPM of $14.20 and generated revenue of $143.91.  I’d love to give you our eCPI figure here, but unfortunately Applovin doesn’t reveal the amount of installs generated.

(Note – this is not to say one ad network is better than the other.  Each network has its advantages and disadvantages).

So, the bottom line is that wherever possible you should be measuring your eCPI (effective cost per install) and comparing this to industry benchmarks to see if your ad network is doing you a good service or not.

CPM – Watch Out!

Another reason for taking CPM with a VERY BIG grain of salt is that this doesn’t take into account FILL RATE.  An ad network might show fantastic CPMs, but it might only be achieving 50% fill rate, meaning you are losing out on A LOT of impressions.

A better statistic is to look at rCPM (request CPM).  To understand this you need a bit of background about how ads are called.

Let’s say you have an ad placement on app open.  When the app opens, a REQUEST will be sent to the ad server telling the server “the app just got opened, please show me an ad!”  The server will then check if it has advertisers available, and if the answer is yes it will send an ad to be shown.  If not, then obviously no ad will be shown.

Why might an ad not be shown? Typically advertisers pay a lot of money to get users from the US, Australia, UK and other Western countries, since these users tend to spend more money in their apps and so are worth a lot more money to the advertisers.  For most ad networks, fill rates in these countries are pretty good, while in other countries fill rate can be non-existent or at least much lower since advertisers aren’t looking to get users from non-Western countries.

So by looking at request CPM (rCPM) this takes into account fill rate.  If the fill rate is 100%, then rCPM will equal CPM.  But if the fill rate is 50%, then rCPM will be 50% lower than CPM. This will give a much truer picture of an ad network’s performance, since fill rate is a critical revenue driving factor.

A Call For More Transparency From Ad Networks

Unfortunately us publishers are left way too much in the dark when it comes to statistics from our ad networks.  As you can see from what I’ve written above, ad networks can be pretty selective with what info they share with developers.  Chartboost leads the way by a long long way in terms of transparency, with publishers able to view even which ads are showing and detailed stats for each ad.

This should be the minimum:

• Fill Rate
• Impressions
• Clicks
• Click through rate (CTR)
• eCPM
• eCPI
• Revenue

Of course, it would be great if we could see even more detailed information, such as stats for various advertisers in our apps, stats for various countries, etc., but the above is really the bare bones of what I would want to see.

[Note:  while writing this article we corresponded with Applovin and they already implemented display of Fill Rate to their dashboard, so kudos to them!]

What can you do about all this?

First off, you really should be monitoring your stats.  Unfortunately as I pointed out above, most ad networks are missing out on vital information for us to make informed decisions, but you should be monitoring whatever you can.

Where possible, I’d advise you to keep track of your eCPIs (this is what we are doing).

So, if you see that an ad network is offering non-competitive CPIs, simply switch off their SDK and try out another ad network.  Compare the results and keep the better one.  This simple exercise can yield a huge difference in revenue, as I showed you above.

Apart from monitoring the payment per install, you should also keep in mind the massive issue of fill rates – this can be a huge source of missing revenue.

Ideally I would love to use an ad mediator for interstitial ads, using the highest paying networks first and then if there is no fill, to default back to a network which might be lower paying but which will have fill.  Unfortunately, to the best of my knowledge no such ad mediator exists.

So, in practice what we do at the moment is we use two ad networks at placements such as app open and on app movement to foreground.  Chartboost in our experience has the highest fill rate, so we will always use this in addition to Applovin or Revmob or other ad networks.

If you do these two things, you should see a significant increase in revenue.

What do you think?  Should ad networks be more transparent with their reporting? Write your thoughts below!

 

In this article I’d like to address a few misconceptions and share from my experience dealing with App store Optimization.

The basic formula everyone already knows – target high volume, low competition keywords.  So this sentence means different things to different people.  What this does NOT mean is to randomly chuck in single word keywords hoping some of them will hit the mark.

This is one of my biggest peeves of training people in ASO and dealing with so-called ASO experts. So we’ll dive into some of this slightly more advanced theory.

What does a high volume keyword mean?  How do you determine this?

Also, we’ll look at competition.  How many competitors do you need to compete against?  Or do we need to look at other factors in order to judge competition? (Hint: The answer is YES).

A Closer Look At Keyword Volume

Unlike with Google based SEO, there is no one perfect tool such as the GKWT (Google Keyword Tool, now called the Keyword Planner) for ASO.  Therefore as a keyword researcher you need to take in all the objective data at your disposal to determine what type of volume a particular keyword has.

Here is a summary of the data points that you should consider when trying to estimate volume for a particular keyword:

1.       Ratings pattern. 

Number of ratings is probably one of the most reliable proxies for measuring number of downloads.

When looking at the ratings pattern, always look at the Current Reviews (not all time) and keep in mind the time since the last update.  The reason I recommend looking at Current Reviews with respect to time since last update is that some apps may have tens of thousands of all time reviews, but have very few downloads currently.  This would be the case in apps that were once popular in the early days of the apps store, but which since then have sunk into obscurity.

How many ratings are a lot of ratings?  This is individual for each country.  You can get a sense for how much is a lot by typing in something “Big” like the kw “games” into a tool like Senor Tower (my favorite ASO tool).

You can see above that for the US app store, “a lot” of ratings is in the high hundreds to thousands of current reviews.

You can also see in the above diagram that ALL the top apps have “a lot” of ratings, which means that “games” is probably a very high volume term!

On the other hand, for the random kw phrase “boring games” look at the pattern below:

You can see a very different picture here.  Here the top 2 apps have only 10’s of current reviews and then all the results below that have only 1 or no reviews.  So “boring games” is probably NOT a high volume KW!

When looking at the ratings pattern you need to look at the whole PATTERN of the search results and use a bit of pattern recognition.  Even if the first result has thousands of ratings, but all the results below have very few ratings I would still class that as LOW VOLUME, since only the first result is getting any traffic.

2.       Category Rankings.

Look at the two examples above at the Category Rankings column.  This column displays if an app is ranked in the App Store Top Charts.

In the first example for “games”, you can see that every single one of the results also has a category ranking (and also that most of these rank in the Top 50).

On the other hand, for “boring games”, you can see that only 2 apps have a category rankings, so this also strengthens the fact that this is a low volume keyword.

Again, its important to look at the pattern of the top results.

3.        Sensor Tower Traffic Score.

This is pretty good for a quick estimation of traffic and to give a “score”.  The traffic store definitely gives a quick “snap shot” indication and is great to save time.

However, if you know how to look at the other parameters, it will allow you to “reality check” this score.  Also, for a lot of keywords a Traffic Score isn’t available, so knowing how to estimate traffic yourself is still important.

4.        Search Suggest.

If a term appears early in search suggest, this is an excellent indicator of high keyword volume.

Example: If I type in the letter “a” – all the results that appear here are VERY HIGH volume keywords.  Then when I type “ap”, the new terms that appear here are second tier, slightly lower volume keywords.  Typing “app” brings up third tier keywords, etc.  Also, the higher the term appears in the search suggest results, the greater the search volume.

I get my keyword researchers to categorize keywords they find using search suggest A to E.  “A” being if the term appears on typing one letter, “B” if the term appears on typing two letter, etc.

5.       Other.

There are other sources, such as GKWT that can provide an estimation of keyword volume and it is definitely a great proxy for app store search volume.

However, keep in mind that sometimes keywords that have high volume on Google might not have high volume on the App Store, since people search with different intent on these different platforms.

So, when considering if a keyword is this a high, low or medium volume keyword, you should take all of these elements into account.

A Closer Look At Keyword Competition

In iOS 6+, search results are displayed one at a time on the iPhone.  Therefore in order to hope to have any visibility, you must aim to be ranked in the Top 3 results for the keywords you are targeting.  Otherwise you are basically invisible.

That said, its much easier to optimize for the iPad, since there are 6 results displayed at any one time on the screen and scrolling is much easier, so even if you are ranked in the Top 6 results, you will have pretty good search visibility on the iPad.

(Interesting stat: Over 50% of our downloads across all our apps come from iPad users, even though iPads make up only about 20% of the total iOS devices – conclusion #1: ASO is FAR more effective on the iPad; conclusion #2 – if you’re not putting out universal apps, you are missing out on A LOT of traffic).

In general, to have any hope to rank in the Top 3 results, we need to take into account the following rules:

1) Keywords in Title have more strength than Keywords in Keyword Field.
2) The strength of keywords in the Title is determined by the following hierarchy:  Exact Title Match > Broad Title Match > Partial Title Match > No Title Match  (see point 4 under the Title section for more info).

As a quick aside, here is what I mean by title match…

Exact Match – all the KW in the KW phrase are present in the title in the exact same order.

Example:  For the KW phrase “Christmas Games”, exact match means that the words making up this keyword appear in exactly this order in the Title.   “Santa VS Elves – Christmas Games of Fun” and “Christmas Games of Santa VS Elves” are both examples of Exact Matches for the KW “Christmas Games”.

Broad Match – All KWs are present in the Title, but not in the exact same order.

For example:  “Santa VS Elves – Games of Christmas” or “Santa VS Elves Games – Great Christmas Fun” are both examples of broad match types.  Also, the higher the KW density, the higher the app will rate for that KW phrase (all other things being equal).

Partial Match – Only some of the words making up the KW are present in the title.

Example:  For the KW phrase “Christmas Games”, “Santa VS Elves – Christmas Fun” is a partial patch.

No Match – None of the KW appear in the Title (i.e. the keywords are only in the KW field).

Let’s look at the example of the term “Pinball HD Collection”, which is a popular pinball app (it is the 3^rd results after we type “pinb” into search suggest).

You can see above that there are 9 iphone apps total that appear for this phrase.

The top 2 results have an EXACT title match.  Have a look at Xmas Pinball, ranked #4.

The full title of the app is:

Xmas Pinball Retro Classic – Cool Christmas Arcade Game Collection For Kids HD FREE

It has a BROAD title match for the keyword phrase “Pinball Collection HD”.  If you wanted to prioritize this keyword phrase above others and increase the chances to rank in the top 3 for this term, we should include the keyword phrase “Pinball Collection HD” as an exact match in the title.

Example:

Xmas Pinball Retro Classic – Cool Christmas Arcade Pinball Collection HD FREE

Why is Xmas Pinball currently ranked #4, when the #3 result has NO TITLE MATCH?  The reason is that game #3 (Wild West Pinball) has more download volume.

You can see though that the reason Xmas Pinball is ranked #4 is because all the other results have either NO TITLE MATCH or only a PARTIAL TITLE MATCH.

So, in order to properly assess competition, you need to look at the Title Match of the top results.

All other things being equal, if for a particular keyword phrase:

The top 5 results have NO TITLE MATCH:  You need a PARTIAL TITLE MATCH in order to beat them and rank in the Top 3 results.

The top 5 results have PARTIAL TITLE MATCHES:  You need a BROAD TITLE MATCH in order to beat them and rank in the Top 3 results.

The top 5 results have BROAD TITLE MATCHES:  You need an EXACT TITLE MATCH in order to beat them and rank in the Top 3 results.

The top 5 results have EXACT TITLE MATCHES:  You will not be able to rank in the Top 3 results.

Understanding this is very important in order to know which keywords you can effectively compete for.

Note: some very high volume keywords, e.g. “racing games”, even though the top results don’t have an exact title match, due to the very high download volumes of these games, there is no chance to rank in the top 3 results.  For this sample term, you can see some exact match result further down the list (spots 12 onwards) and you can see how they don’t rank in the top 3.

Note:  Sensor Tower makes it easier to spot EXACT TITLE MATCHES by marking these results with a red triangle at the top right of entry.

Also, in order to see the full app title, you can press on the Keyword Spy symbol at the right of the entry.

UPDATE (3rd December 2013):

I got some important feedback from readers of this blog, which I’d like to share here and was not made clear enough in the article.

One of the MAJOR factors in determining where you rank in the search results is the download volume / velocity and ratings.  Major publishers BUY a lot of traffic and this is a major factor in their high search rankings.

This article is aimed at indy developers, who typically DON’T have a budget of tens of thousands of dollars to spend on paid advertising campaigns.  The section on competition is given that all other factors (i.e. download volume, etc) are held constant.

Paying attention to proper placement of keywords (title vs keyword field and exact title match / high density title match) cost nothing and can result in a very high ROI, so it is definitely worth doing!