Power BI Template App Tips - Parameters, Models and Reports That Survive Installation
If you sell software in Australia, you've probably had a customer ask "can we get this in Power BI?" Usually they mean they want dashboards over the data your product holds, and they don't want to build them. A Power BI template app is one of the answers. You build the reports once, package them, and each customer installs their own copy connected to their own data.
The catch is that a template app has to work in a tenant you've never seen, connected to data you can't look at, for a user who may not know what a parameter is. Things that are fine in an internal report become support tickets the moment a stranger installs it.
Microsoft has a page of tips for creating template apps that's worth reading start to finish. Below is how those tips look in practice, plus the things we've learnt building these for ISVs and software teams.
Start with the install experience
Before you design a single visual, picture the moment a customer installs your app. They click install, Power BI creates a workspace in their tenant, and then asks them for parameter values and credentials. That screen is the whole first impression. If it asks for six parameters with names like pSrvURL_v2, you've already lost half your users.
So design backwards from that screen. What's the absolute minimum someone needs to type to get data flowing? For most SaaS products the answer is a tenant or account identifier and an API key or OAuth sign-in. Everything else should have a sensible default or be derived.
Parameters
Parameters are how you make one set of queries work for every customer. A few rules we stick to:
Name them for humans. The parameter name is what the installer sees. "Company subdomain" is better than "BaseUrl". Add a description too, because it shows up as help text.
Use parameters for anything customer-specific. Server URLs, account IDs, date ranges for the initial load. If a value is hardcoded in a query and differs between customers, it's a bug waiting for your second customer.
Keep the count small. Two or three parameters is plenty. Each extra one is another chance for someone to type the wrong thing.
Don't put secrets in parameters. Credentials belong in the data source authentication step, not in a text parameter that ends up stored in plain view.
Give the app something to show before connecting. Template apps can install with sample data so users see populated reports before entering their own details. Do this. A report full of blank visuals and errors looks broken even when it isn't. Sample data that shows a realistic picture sells the app.
Data model
The model is where internal habits cause the most trouble.
Name things properly. Tables, columns and measures should all have friendly names. Customers will see them in field lists if they explore or build on top of the dataset. "Total Revenue" beats SUM_rev_amt.
Hide what users don't need. Key columns, sort-by columns, technical fields. Hide them. A field list with 200 entries is unusable.
Format every measure. Currency, percentages, decimal places. Defaults look sloppy and nobody wants to see a revenue figure with seven decimal places. For Australian customers, think about date formats and currency too. Your US-built template showing month-first dates will get noticed.
Write descriptions on measures. Hover text explaining how a measure is calculated saves a lot of "why doesn't this match our finance system" emails.
Keep it lean. Don't import columns you aren't using. The model refreshes in each customer's environment, so a bloated model costs every customer time, and some of them will be on Pro capacity with its limits.
Reports
Build a landing page. The first report page should answer "what is this and where do I start". A short summary, a handful of headline numbers, and clear navigation to the detail pages.
Design for the data you can't see. Your test data has 50 customers. A real install might have 5 or 50,000. Check that visuals still make sense at both extremes. Slicers with thousands of values, bar charts with three bars, tables that scroll forever. We test every template app against a deliberately tiny dataset and a deliberately huge one.
Handle empty states. If a customer doesn't use a particular feature of your product, the related visuals will be blank. Add a text box or a measure that shows a clear message rather than leaving a confusing empty chart.
Use a consistent theme. Apply a report theme so colours and fonts match your brand across every page. It makes the app look like a product, not a side project.
Dashboards and app settings
The app itself has settings people rush through. Fill in the description properly, add a decent logo, set a theme colour and include a support link that goes somewhere real. If your support link is a generic homepage, you'll get angry emails instead of tickets.
Write a short "getting started" note, either on the landing page or in the app description: where to find the parameter values, how long the first refresh takes, what to do if it fails.
Limitations to plan around
Template apps don't support everything a normal Power BI workspace does, and you need to know this before you build, not after. The specifics change over time, so check the current list in the docs, but historically the things that caught people out included:
- Some content types aren't supported in template apps. Check before you assume dataflows, paginated reports or certain dataset types will package up.
- Row-level security has limitations in a template app context. If you were planning to rely on RLS for multi-tenant separation, rethink it. The better pattern is that each customer's install only connects to their own data in the first place.
- Customers can modify the installed content, and updates you ship can conflict with their changes. Tell customers what's safe to change.
- Credentials are per install. If your API tokens expire, every customer hits the same refresh failure at the same time, so plan your token lifetime and messaging accordingly.
The release process
Template apps have test, pre-production and production stages. Use them properly. Install the test version in a separate tenant, ideally one that isn't yours, with a user who hasn't seen the app before. Watch them install it. You'll find at least one problem every time.
Version your releases and keep release notes. When a customer reports a problem, the first question is which version they installed, and you want to be able to answer it.
Is a template app the right choice?
Honestly, not always. Template apps are good when you have many customers, a fairly standard set of reports, and a data source that Power BI can reach with simple credentials, usually a REST API or OData feed. They're less good when every customer wants heavy customisation, or when your data sits in a database you'd rather not expose.
Some of our clients have moved to Power BI Embedded instead, where reports live inside their own application and customers never touch Power BI directly. It's more work up front, but you control the whole experience. Others are now adding AI on top, letting customers ask questions of their data in plain English rather than clicking through report pages.
If you're weighing up template apps against embedding, or building one for the first time, our Power BI consultants can help you pick the right path. For software companies that want reporting as part of the product itself, our software development team has built both approaches. And if you've already shipped a template app and want to know what customers are actually asking about, that's the kind of question our business intelligence work tends to start with.
Build it for the stranger who installs it on a Friday afternoon with no help. If it works for them, it'll work for everyone.