Google IAP v3 returning 'cancelled' events as 'failed' events on devices

Solar2D Monetization (in-app purchases, ads, etc.)
#1

Just wondered if anyone else had noticed this.

I was running some tests on an Android Google Play device (based on some user feedback for IAP Badger) and found this behaviour:

  • code initiates a call to store.purchase to Google IAP v3
  • user is presented with dialog asking to confirm purchase
  • user cancels purchase (eg. by pressing back button)
  • Corona returns what should be a  cancelled  transaction event as a  failed event with error code  -1005

Has anyone else come across this?

I know it’s the most devastating mis-classification in the world, but it’s an inconsistency that might throw some apps.

Also - does anyone know if this is a new change or something that’s escaped my notice in the past?  Maybe a change in behaviour back when Google moved to IAP v3?

Simon :slight_smile:

#2

I’ll ask engineering to look into it, but I’m pretty sure those messages are just passed through.  If you’re looking at “adb logcat” you’ll see it coming from the Finsky process (Google IAP).

Rob

#3

Hi guys , 

I am facing exactly behavior as Simon described ,

my problem is that i currently run an open alpha test, and got feedback from testers that received the “this item is not available” error ,

I am trying to measure this problem with flurry , but cannot distinguish between errors, and normal cancel.

When i try purchase from my local device billing going smoothly , with real charges,

Rob - did you manage to get a response from the engineers about this issue?

Maybe you heard about problems with IAP during google open alpha tests?

Thanks, 

Barak

#4

Let me check.

Rob

#5

This is a Corona bug…

What you need to do is check in  transaction.state == "failed"

if transaction.isError then

and then handle accordingly…

(easier than waiting for a core fix)

#6

Thanks for the replies guys , 

I’ve tested the isError idea now on my local device ,

however i am getting isError=true in both of my cases (it is true also when the user click cancel) 

#7

Rob - how safe would if be to look for error code -1005 and translate this manually into a failed event?

Are these error codes subject to change?

Simon

#8

I would tested against that response and handle it as failed, but Google may also have other error codes or may use new ones in the future that would need to be included after we discover them.

Rob

#9

The number of download per day have decrease per 600% since the first user have problem with in app. I think google don’t show or show less app with in app problem.

Please Rob Miracle find a quick solution because it’s not only our income who go down.

Stats is an image who show where the first problem appear. (for Stronghold) Since this date I have a lot of trouble. My income are also down. I was at 300€ per month and now I am at around 2€…

For my other apps, I have the exact same problem.

Google have updated : https://developer.android.com/google/play/billing/billing_reference.html during summer

#10

@remi I’m a bit confused.

How do IAP cancellations that come back as error code 1005’s impact downloads?  

How do you earn income?  IAP? Ads?

How does this “cancellation as an error code” impact successful purchases? If I understand the problem correctly this is an issue with people cancelling purchases. These would never be sales anyway.

If you’re code isn’t handling the transactionState failed and crashing, I could see where that might lead to lower ratings or Google down ranking you for crashing. 

I need more information before I can go to engineering.  As of now I see a problem with Google returning a wrong, but completely hand-able condition.

Rob

#11

@Rob Download come from only by the google play. Google decide to show or not my app. When an app make the user disappointing he won’t download more app. When he have trouble to do an inapp, the next time he will think twice before buy something.

For an inapp google take around 20% of the price, that’s mean they have all interest to do some ad / show the app if they win money.

That’s why their algorithm don’t show much my app.

My app make the user disapointing of inapp that’s mean the probability for the user to buy another inapp is reduce.

All inapp are refund, that’s mean google won’t win some money.

I am very sorry for my english, I am not use to write. If there is something not cristal clear I can explain it

Best regards,

Rémi

#12

IMHO there is zero correlation between IAP and downloads.  Well apart from you can’t have sales if users do not download.

Now, there is a correlation between buggy/slow apps and apps that ANR and crash a lot and your ranking in searches - which is a positive move to surface apps that provide a great user experience.  After all who wants to play a game that crashes all the time?

I pushed Corona hard to fix IAP and FB bugs and can report a massive drop in random ANRs and crashes (from many hundreds a day to low tens a day) - although there is still some work to be done here.

#13

@sphere Game Studios it’s very strange what you say because when I integrate in app in my game and a first user buy one, my app was show a lot of more

#14

If your app is crashing, it will show in your ANR and Crash reporting. You can draw a correlation between those numbers and downloads. Likewise you can look at your star rating and reviews. A low rated app will be down ranked too. Google still shows your app but instead of page 3 it might be page 10.  

If your app is crashing you should fix it. That’s my point with this IAP situation. Getting a " failed  event with error code  -1005"  isn’t a crash. It’s communications from the IAP server saying something went wrong. It’s your responsibility to handle that and let your users know that the transaction didn’t go trough. If it’s truly Google sending the wrong message for a “Cancel”, the user won’t be bothered if you present a message saying the purchase failed. The user will expect that since they intentionally “failed” the transaction.

For someone to get an IAP error means they have already downloaded your app. Of course if the IAP fails when it shouldn’t (i.e. they did not cancel), they of course can give you a poor rating. But the original poster here is talking about “Cancels” translating into -1005’s. Is there evidence that actual purchases are generating -1005s?

Finally Games have something called a “shelf-life” which means that purchases/downloads will tail off over time, rankings naturally fall off. It’s part of the natural life-cycle of a game unless you’re advertising your game to keep the game in front of new users.

Rob

#15

Rob is mostly correct… there is a definitive life-cycle for most games (there are clear exceptions for Candy Crush and Clash of Clans, etc).  But for most of us our games will grow, plateau and decline.

Now as my first game is starting its 3rd year I will share some stats… installs have shrunk to 2-3k per day but sales and profitability have soared. No longer am I getting pointless installs but better value installs so it is all good.

Remi, feel free to PM me on slack if you want to chat.

#16

I’ll ask engineering to look into it, but I’m pretty sure those messages are just passed through.  If you’re looking at “adb logcat” you’ll see it coming from the Finsky process (Google IAP).

Rob

#17

Hi guys , 

I am facing exactly behavior as Simon described ,

my problem is that i currently run an open alpha test, and got feedback from testers that received the “this item is not available” error ,

I am trying to measure this problem with flurry , but cannot distinguish between errors, and normal cancel.

When i try purchase from my local device billing going smoothly , with real charges,

Rob - did you manage to get a response from the engineers about this issue?

Maybe you heard about problems with IAP during google open alpha tests?

Thanks, 

Barak

#18

Let me check.

Rob

#19

This is a Corona bug…

What you need to do is check in  transaction.state == "failed"

if transaction.isError then

and then handle accordingly…

(easier than waiting for a core fix)

#20

Thanks for the replies guys , 

I’ve tested the isError idea now on my local device ,

however i am getting isError=true in both of my cases (it is true also when the user click cancel)