The second Microsoft Ignite event of 2021 included several important updates on a topic that I cover regularly on this blog: Microsoft Business Applications licensing. The biggest announcement of course was the pay-as-you-go licensing model (PAYG) for Power Apps:
Here’s a few thoughts and observations on what was announced & updated on the licensing front.
PAYG and Azure
Despite of its name, the biggest impact from the Power Apps pay-as-you-go plans isn’t necessarily the pricing model. Rather it’s the currency in which you pay for it: Azure money. This is something that most larger organizations will already have in their virtual cash reserves, which makes it considerably easier to spend on services.
Licensing agreements are both an enabler and a blocker for Microsoft cloud services, especially on enterprise level. Getting something new like Power Apps premium licenses included into the IT portfolio available to the business users can take a lot of time and effort. Once it’s included, though, there’s plenty of practical benefits for using these as opposed to some fancy new cloud service that might come along.
Having an official “exchange rate” that converts these new low-code tools managed on the Microsoft 365 side into tools and terminology compatible with the pro-dev work already done in Azure is key step in selling the Fusion Teams story to organizations. Even if the PAYG model wouldn’t be cheaper in practice, having Power Apps license costs show up in existing reports on Azure Cost Management side will make it less of an unfamiliar beast to IT and developers. Also the discounts customers may have negotiated based on their Azure spend will presumably apply to Power Apps, too.
PAYG and apps
PAYG is specified on environment level, by defining a billing policy and selecting which environments are included in it. However, you can exclude specific apps from the model and require their users to have a Power Apps Per User license instead. Mixing PAYG with Per App subscription plan isn’t allowed, though.
I believe that underneath the covers the PAYG Per App plan and the regular Per App plan are mostly identical. The major difference is that at the end of the month, the PAYG pass is removed from the user until he/she opens the app again. With the regular Per App subscription plan, the pass allocated upon the initial app sharing to the user and isn’t revoked until the app is explicitly unshared.
So, which one should you go for then with your Power Apps that use premium features? Subscription model or PAYG? First we must keep in mind that the recent 50% price drop in Power Apps subscription plans took the price of a Per App license down from $10 to $5. At the same time the Per App entitlements were updated to cover a single app, rather than up to 3 different apps (existing customers have been compensated with 3x passes). Now with the new options, you can either pay $5 per user per app per month in the subscription model, or pay $10 for the exact same features but only for those months when a user opens the app.
If there was a quick & easy way to assign, unassign and report on the Per App subscription licenses, it would clearly be tempting to save 50% and pay with the old model rather than PAYG. In practice, Per App licensing management remains a nightmare that has surely cost customers (and partners) lots of money in time spent figuring out who’s got access to apps, why someone doesn’t, and how much licenses are actually being used. It’s been over 2 years since the Per App plan was launched and we still don’t have a report in Power Platform Admin Center that would give customers the information needed for managing this process.
If the PAYG mechanism can both automate the assignment and removal of licenses, as well as provide detailed reporting in Azure Cost Management, then it could well be worth the price premium for process efficiency and cost visibility. For any smaller app experiments that you’re not yet entirely sure on how widely they will be used, starting with the Azure subscription based billing makes a lot of sense. You’ll be able to cut down on the initial admin hassle and get real data on the app usage to base your future licensing decisions on, whether to use PAYG or committed subscription passes/seats in the long term.
PAYG and capacity
It’s important to understand that the PAYG model doesn’t completely transform Power Platform into a pure consumption based model like Azure. After all, a single app launch during a calendar month will cost you the same as 1000 launches by the same user, and using 2 different apps will double the cost regardless of how much load you generate to the cloud services. Think of it more like buying a monthly pass for public transportation in your local area vs. taking an Uber and paying in proportion to the actual duration and length of your rides. If you travel to another city not covered by your current pass, you’ll need to purchase another pass. In the case of Power Apps, there aren’t any per hour or per day tickets sold, so the monthly pass is the minimum charge for the Power Apps per app meter billing rules.
Having said that, there is also the concept of pay-as-you-go capacity when it comes to Power Platform licensing. API capacity is something that Microsoft says a normal user shouldn’t have to worry about, since there is bundled capacity included with the app/user licenses. At Ignite an announcement was made that increased the request limits for both licensed users and non-licensed users (check out Gustaf’s blog for details). If you do run out of requests, though, there will soon be a PAYG Power Platform request meter that can be used to pay for the overage.
Something that does already exist today and is a much easier concept to approach than API consumption is the storage capacity. Unlike the traditional subscription licenses that are paid upfront, the PAYG model doesn’t have any storage capacity accrued per each user or app pass. This makes sense, since those numbers are only known as-you-go by the end of the month, whereas data storage is something a lot more persistent.
Whenever you activate PAYG for an environment, it means it no longer draws from the tenant’s common capacity pool. The environment does get 1 GB of “free” Dataverse database capacity + 1 GB file storage capacity, but after that you’ll pay for everything based on what the Dataverse capacity meter shows. These rats are slightly higher than subscription based capacity prices, at $48 and $2.4 per GB respectively (note that log capacity is always paid for at $12/GB). If you’re used to leveraging the tenant level capacity accrued from user subscription licenses for your storage needs (coming from Dynamics 365 users, for example), then this can be a surprising additional cost when activating the Azure based billing policy for your environment.
There’s one gotcha in the current way how PAYG storage capacity can be used. You see, in order for you to assign a PAYG billing policy to an environment, the environment must first be provisioned. What this means is that the tenant will need to have a minimum of 1 GB database capacity available for the environment creation to succeed. So, you’ll need non-PAYG capacity from elsewhere (such as a paid Power Apps or Dynamics 365 plan) before you can use an Azure subscription to start paying for the capacity instead. Presumably Microsoft will fix this issue in the near future, since it’s in no one’s interest to block PAYG as the sole billing method.
While storage capacity has been reported and technically enforced for a while already, Power Platform requests are still something that exist on paper but cannot be reported on nor paid via actually assignable licenses. In the FAQ section of their documentation page, there is now a new answer to the following question: What are the timelines for Power Platform Request limits?
The concept of limits was first introduced in late 2019 and documented limits were substantially increased in late 2021. Generally available reporting for Power Platform Requests is expected in the first quarter of calendar 2022. Any potential high usage enforcement wouldn’t start until six months after reports have been made available. Assignment of add-on capacity packs should be aligned to high usage enforcement.
Let’s assume that the reports would be launched in March 2022, after which there will be a 6 month grace period before the technical enforcement kicks in. This would mean that the October 2019 licensing changes that started all this confusion around what the platform can & cannot be used for will have taken full 3 years to truly come into effect. That’s a long, long time. Yet I do think it’s encouraging that Microsoft haven’t rushed into enforcing a rule that might have caused way too much collateral damage if not thought out well enough.
Power Automate licensing updates
There isn’t a PAYG model available for Power Automate. Sure, you can use premium connector cloud flows within the context of a Per App licensed app (although I’ve found this to be difficult/impossible in practice). The nature of a background process automation is fairly different from the interactive usage of an app UI, so it kinda makes sense why new options haven’t been introduced. Besides, if the Power Automate licensing model doesn’t work for your scenarios, then moving to the consumption based Azure Logic Apps is already a good option to consider.
While there weren’t any major Power Automate specific licensing changes announced at Ignite, there was a wealth of new documentation released on the topic during the event. The flow licensing FAQ contains some interesting answers when it comes to what we would have traditionally considered to be multiplexing. I’m one of the few people who have ever dared to blog about multiplexing, yet it seems that there are things I haven’t fully grasped about the topic either. This infromation in the FAQ caught me by surprise:
A common question is, “If a flow is triggered when a SharePoint list item is updated, and many users interact with that list, will there be a cost for each user?” The answer is if the flow does not use a premium connector such as calling Dataverse in the full production environment (not the Microsoft Teams environment), having an Office 365 license is enough. If the flow uses premium connectors, since the trigger is an automated trigger, only the owner needs a premium license.
Having a web form that any user could submit to create a lead record in Dynamics 365 used to be the scenario where I’d say “sorry, you’re violating the licensing terms”. Now, Microsoft explicitly calls out that if a user enters data into a SharePoint list and that in turn triggers an automated flow that uses a premium connector to push the data into Dataverse, it’s enough for the flow owner to have the premium license.
This little change opens up many scenarios where recording entries into Dataverse / Dynamics 365 may have previously been considered too expensive. Using a SharePoint / Microsoft List as the intermediate data storage is now officially a supported method to reduce the need for Power Automate premium licenses when using automated flows. I personally find it hard to understand how this statement wouldn’t contradict Microsoft’s core definition of multiplexing, but what can I say? Other than: welcome to Business Apps licensing!😂
One more small but significant change: there will now be AI Builder capacity included with Power Apps Per App plan (250 credits) and Power Apps Per User plan (500 credits).
How much AI that will actually give you depends on the usage scenarios, which you can try and estimate with the AI Builder calculator.
Why is this a big deal? Well, for some mysterious reason, Microsoft had earlier decided that the starting price for buying AI Builder capacity was $500/month for 1 million credits. Need just a few runs for your little cloud flow every month? “Sorry, we can’t be bothered to support anything below 1M.“
This was a prime example of how not to price cloud services for citizen developers, as they should be given the chance to focus on building apps instead of building business cases for software licenses. The starter capacity will now make it possible for app makers to leverage AI models in small solutions and experiments beyond just a 30 day trial.