BI Trends

Why Microsoft Doesn’t Understand Embedded Analytics

By Josh Martin
Share on LinkedIn Tweet about this on Twitter Share on Facebook

I’ll admit that I’m a Microsoft fanboy. Last year when I was still a mobile industry analyst, I referred to Microsoft as the most interesting company in mobile. In fact, I’ve given up on a number of apps because Microsoft offered something better:

  • Fantastical2 for Outlook’s Calendar – The Agenda view on Outlook is excellent
  • Individual E-Mail Clients (and Apple stock mail) for Outlook – It provided the best smart inbox to limit alerts on my Apple Watch (when I was using it)
  • Evernote for OneNote – I found the notebook structure of OneNote much more suited to my needs
  • Apple’s Camera app for the excellent and just released Pix application

I also convinced everyone in my office to use Office Lens (you should try it too!).

So why have I burned nearly 150 words praising a competitor? To prove unequivocally that I’m not a Microsoft hater.

But the launch of PowerBI Embedded shows why software vendors should strongly consider partnering with an OEM analytics expert when choosing to embed analytics. Because sometimes big companies – such as Microsoft – fail to grasp the basic fundamental question of pricing.

As a software vendor, the cost of analytics is a huge consideration when making a buying decision. Often, analytics are a competitive differentiator or a way to keep existing customers happy – not necessarily a new revenue stream – so cost considerations are paramount when deciding what analytics solution to embed within your software. I should know, this is all my team does at Logi (Don’t worry I won’t get preachy – this is an analysis not a sales pitch).

The Original PowerBI Embedded pricing penalized heavy use.

If you’re anything like me, you probably want your customers to use your software a lot. And if you’re really lucky, it becomes essential to making real-time business decisions throughout the day. So when Microsoft launched a Pay-Per-Render model, I was a bit taken aback, because it basically punished users for relying on the software.

Here’s how it worked. Let’s say, as an app provider, your customer had a dashboard with four visuals –each time they refreshed the page, you would get charged for four renders. Sure, this was a great pricing model if you thought your customers weren’t going to frequently use the data/charts to make better decisions. But if they’re not going to use them, why are you spending the time (and money) embedding them? On the other hand, if your software scaled (yay!) and usage was high (double yay!), your costs could run away from you very quickly.

Here’s the formula I used when determining costs (based on information available on Microsoft’s website)

(# of Visual Elements * Renders per Hour * Hours per Day of Use) * Cost per Render ($0.0025) = Cost per User per Day

The table below is a breakdown of several potential use case scenarios and their cost to you – the software vendor.

Visual Elements Renders per Hour Hours per day Total Renders Cost Per User Per Day
1 5 8 40  $             0.10
2 5 8 80  $             0.20
3 5 8 120  $             0.30
4 5 8 160  $             0.40

These numbers may seem small – but they are per user per day. If you take a fairly moderate use case of just 2 visual elements updated 5x per hour throughout the work day, you are paying $0.20 per user per day. Multiply that by 260 workdays and your per-user costs is $41.60 each year. Still not bad – but what if you have 10,000 users? You’re now paying over 400K per year. Of course, it’s likely you’ll have a mix of use cases – but isn’t the goal is to make your product so useful customers are constantly relying on it?

For infrequent usage this pricing works great – but why bother integrating a solution you don’t want your customers to rely on?

New PowerBI Embedded Pricing Limits Risk but Punishes Occasional Users

So, Microsoft saw the numbers, heard the customer gripes and made a change. It has now moved to a pay-per-session model.

My understanding of this model is that for one hour – the length of a session – users can render as many charts as they want and they are only charged for a single session. However, after the hour ends (and assuming you must re-render to begin a new session) the clock starts again. Unsurprisingly, the cost per session is much higher than the cost per render to account for the unlimited use customers are afforded.

Here’s the formula I used when determining costs (based on information available on Microsoft’s website)

Hours per Day of Use * Cost per Session ($0.05) = Cost per User per Day

Visual Elements Renders per Hour Hours per day Total Renders Cost Per User Per Day
1 5 8 40  $                     0.40
2 5 8 80  $                     0.40
3 5 8 120  $                     0.40
4 5 8 160  $                     0.40

 So, now Microsoft is limiting your upside exposure – which is great – but is now penalizing the mid-level and infrequent use cases.

Under the previous model if your customer could render 4 charts twice every hour for the cost of $.16 per day. Now there usage will cost you $.40. And yes, while the cost for heavy users in this scenario declines dramatically – just 25% of what it was – your costs will still likely increase based on your customer mix. In fact, according to the chart below – you don’t even break even with the pay-per-session model – until your customers render more than 160 visuals per day.

Chart - power bi

So now Microsoft has fully swung the other direction with its pricing model. Now heavy users are no longer the burden, but light users are! It seems pretty clear that Microsoft will likely need to revise their pricing again. But what this shows is that Microsoft simply doesn’t understand the embedded analytics market.

Here’s what I think Microsoft doesn’t understand.

1. Software Vendors will not be swayed by no/low upfront costs. No upfront costs sound amazing and Microsoft hopes to wow customers with this value proposition of PowerBI (except requiring you use Azure and the fees associated with that solution). But smart product managers know the ongoing cost of PowerBI embedded will be high, and hoping to negotiate with Redmond after PowerBI is embedded in your product and rolled out to market will be a fool’s errand. At that point, Microsoft’s got you. Companies that are serious about embedding analytics are willing to invest upfront in a robust solution.

2. Embedded analytics are inextricably linked to your product. Once you pick a solution, get your team trained on how to embed it, introduce it to the market and get customer using it you are committed. If the solution doesn’t scale for technology or business model reasons, rewinding the tape and launching something new will require significant internal resources and investment distracting from your core product improvements. Your customers will probably be frustrated as well.

3. Vendors crave predictability. While PowerBI pricing fixes the upfront cost it creates ongoing variable costs for vendors that can completely railroad operations if usage unexpectedly spikes. If companies budget for analytics and don’t reach the threshold then that extra money could potentially be used for something else – or saved to offset future overages. But it can’t be consistently applied to other ongoing operations because companies will need a slush fund to cover the usage costs of PowerBI. And what happens if usage spikes suddenly? Companies will be forced to find budget to pay for the unexpected overruns likely comprising other initiatives. This lack of pricing clarity is the antithesis of the concept of analytics – that have been designed to improve business predictability and forecasting.

4. You want analytics as a competitive differentiator. The software market is hyper-competitive today. The ability for customers to trial competitive software is easy. Contracts tend to be short. And customers want analytics. So, analytics needs to be more than just checking a box. Unless you offer analytics that your customers will want to use every day, there’s no point in integrating them. So, widespread adoption means widespread and diverse usage. Microsoft can’t figure out how to charge for this fairly.

Since joining Logi, I’ve been surprised by our pricing model – developed over many years of experience working with software vendors. We have created best fit licensing that enables us to work with customers based on their pricing model, customer base and growth plans to come up with a fair price that reflects the value of analytics within their application. Prospects are usually pretty surprised to hear this – I know I was when I joined. In the software world there are no one size fits all vendors and there shouldn’t be one size fits all pricing.

It’s important to work with a vendor that understands your unique needs, business model and go-to-market approach. I’m sure in several years Microsoft will have the experience required to hone its pricing strategy – but few companies have the luxury of time to release their next genius app with killer embedded analytics.


Originally published August 11, 2016; updated on August 9th, 2017

About the Author

Josh Martin is the Director of Product Marketing at Logi Analytics. Prior to joining Logi he was an industry analyst covering bleeding edge distribution channels and their impact on the consumer market. In this role he was a thought leader and advised clients on how to successfully benefit from market shifts while positioning products and services for long-term success. Josh holds a Bachelor degree in Business from Babson College.