Do you really need that Looker subscription?
The final part of my Lean Data Stack series looks at data visualisation tools for tight budgets.
I mostly write (or plan to) about Data orgs, technology, metrics, and some miscellaneous work-related stuff, with the aim of publishing fortnightly-ish. If you want to read more of my work, please subscribe below. You can also find me on LinkedIn and Twitter. Thanks for reading!
Our hero the restaurateur has built the world’s most modern kitchen:
A magic fridge that grows and shrinks with the volume of food in stock.
A supplier of everything, who collates ingredients from across the world, and stores them in the magic fridge.
A smartknife, which automatically prepares all the raw ingredients, while performing quality tests and creating documentation along the way.
There’s just one thing left.
The restaurateur imagines a new world of cuisine, in which their guests are not bound by the confines of a menu. Instead, every guest has a touchscreen, allowing them to visualise variations on their meal before ordering it, and enabling them to answer such vital questions as:
What if this tonkotsu ramen were made with spaghetti instead of noodles?
What would this blueberry pie look like if it were presented as several rectangular blocks instead of one big circle?
In case I order too much, can I see a forecast of what the fried eggs will look like tomorrow if I take them home with me?
The guest then sends their order to Cooker, a tool that gives customers the power to choose how their meal is prepared and presented. Because the smartknife has already prepared the raw ingredients, Cooker can now cook any meal that the customer can think of. Haggis cheesecake? No problem. Borscht ice cream? You got it. Eel and marmite wellington? What? No. Get out, you freak.
In my increasingly tenuous restaurant analogy, Cooker is (roughly) to the modern kitchen what Looker is to the modern data stack. It acts as the interface between the data consumer and the underlying data.
Looker costs a fortune, but it’s not your only option. The data visualisation market is crowded, so by necessity I won’t cover every product in detail, and some I won’t cover at all. Because the purpose of this post is to highlight tools suitable for teams on a tight budget, I’ll focus on value for money and keep things brief elsewhere.
The point of this post is not to argue that Looker (or any other premium-priced software) is somehow bad. The market validates that it isn’t. My aim is rather to provoke reflection about whether a cheaper tool might serve your needs.
Can it be done for free?
💸 Tier 0: Google Data Studio and Google Sheets
Google Data Studio: Free, and much better than you’d think
Over at Impala Towers, we’ve got a lot of mileage from Google Data Studio. Since we use BigQuery as our data warehouse, Data Studio sits comfortably on top. It comes at no extra cost, beyond the query costs to populate your dashboards.
For early-stage companies, Data Studio is an underrated and underused tool. Although it has limited features, it does more than you might expect from a free tool, and has a shallow learning curve.
Limited functionality is not always a bad thing, as a lean feature set nudges you away from complexity. Most BI reporting use cases are (and should be) thematically simple, and Data Studio will usually have what you need for vanilla BI.
I would not recommend Data Studio for business-critical reporting in a mature business. It can be slow and buggy, and doesn’t support engineering best practices. Regardless, it’s a good entry-level tool for a small team (ideally already on the Google stack) operating on a small budget.
(say it quietly…) Google Sheets
It may seem odd to include Google Sheets here, since using spreadsheets for BI reporting is mostly anathema to proponents of modern data tools. Even so, Google Sheets is an underrated tool for basic reporting.
People usually see spreadsheets as data silos that float about in isolation from the real world. You dump some data in Excel, do something with it, and either update the data sometimes or forget about it. As time goes on, an unloved spreadsheet starts to look like that mouldy cabbage in the fridge.
It doesn’t have to be this way. Using Connected Sheets, you can connect a G-Sheet to a BigQuery table to refresh the data on a schedule. This allows you to build automated BI reporting in the simplest (and in some ways most flexible) data tool of all: a spreadsheet.
Of course, spreadsheets are less attractive than purpose-built BI tools, and tend to buckle under large volumes of data, but if you’re not fussed about aesthetics and your data set is small, G-Sheets may be a suitable tool.
Like any BI tool, G-Sheets has slicer functionality to filter your visualisations. However, selections made using slicers persist between different users, which isn’t ideal. If you have two people looking at a G-Sheet dashboard at the same time, this might get annoying.
I wouldn’t recommend attempting to build scalable reporting infrastructure using G-Sheets, but it’s often wrongly overlooked for simple use cases, where something needs to be built fast, and doesn’t require polish.
💲 Tier 1: Power BI, Metabase & QuickSight
Power BI: My first choice based on value for money. Not for Mac users
Microsoft Power BI costs $9.99 or £7.50 per license per month, with no distinction between developer and viewer licenses. This makes it one of the cheapest options on the market. Gartner’s Magic Quadrant also ranks Power BI as the market leader across both “ability to execute” and “completeness of vision”, whatever that means.
All things considered, Power BI is my favourite cheap dashboarding tool. It may not be the slickest platform on the market; you can always recognise a Power BI by its musty look and feel, a bit like that cabbage I mentioned earlier. That said, I find it intuitive to use, and effective at data transformation and modelling.
The Power BI learning curve is shallow, but to become an expert you need to master DAX, an analytical language similar(ish) to Excel. It takes some getting used to but isn’t especially difficult. There is a large, active community of Power BI users, with forums that will answer most of your DAX questions.
One downside is that Power BI has no Mac desktop version, so it’s not an obvious choice unless your BI team has Windows machines. If you’re on a Mac and you’re desperate to use Power BI, you could do this via a virtual machine.
Challengers: Metabase & Quicksight
I haven’t used Metabase or Amazon’s QuickSight, but both sit in a similar pricing bracket.
Metabase's "Starter" tier costs $85 per month including licenses for five users, then $5 per additional user thereafter. This may not scale far, as it lacks features like SSO, row-level permissions, and fine-grained access control. For these, you will need to upgrade to “Pro”, starting at $500 per month, which includes ten users. Each additional user costs $10.
QuickSight “author” (developer) licenses cost $24 per month, and “reader” licenses have session-based pricing ($0.30 per session), up to $5 per month.
I don’t see any obvious differentiator for QuickSight besides low cost at scale. You also have the option of adding “Q”, a natural language querying tool, for a further $10 per author per month, and a $10 reader price cap. This functionality comes out-of-the-box with some other platforms, including Power BI.
💲💲 Tier 2: Tableau & Qlik
Tableau: The most versatile and advanced data visualisation tool
Tableau is another of Gartner’s three “leaders”, albeit scoring lower than Power BI on the completely objective, absolutely watertight measures of “ability to execute” and “completeness of vision”.
Tableau is the strongest data visualisation and exploration tool on the market. If you know what you’re doing, you can build pretty much any data visualisation you can imagine with Tableau.
This makes Tableau an excellent tool for editorial purposes (infographics, data journalism, etc.), but it’s debatable whether this flexibility justifies the higher cost ($70 per month per creator, $42 per explorer; $15 per viewer) if you plan to use it primarily for BI reporting.
For interactive use cases, where users explore the data rather than looking at a static dashboard, Tableau has the edge over Power BI. Tableau follows the Grammar of Graphics, defining visualisations by their essential components, rather than a “scatterplot” or “bar chart”. This gives users more freedom to explore data, and adapt their analysis to specific contexts.
Tableau has a steeper learning curve than Power BI, and there’s so much you can do with it that it can be overwhelming to new users. It’s been around for a long time though, and has a large user base. Makeover Monday is a good resource for practicing on real data sets, and Andy Kriebel’s live viz-building sessions are worth watching to learn from an expert.
Qlik: The quiet leader
Qlik Sense is Gartner’s third “leader”, crucially edging out Tableau on “completeness of vision”, but catastrophically lagging on “ability to execute”. It’s a surprise to find Qlik in this category, as it rarely comes up in conversation. I suspect that it’s more popular among big corps than start-ups, which may explain why I hear little about it within my bubble.
The standard pricing tier is a flat $30 per user, which - depending on your mix of user types - may work out similar to Tableau.
💲💲💲 Tier 3: Looker & Sisense
Looker (owned by Google) does not share public pricing, but their sales team has twice quoted me $3k per month (including ten users), then $50 per month per additional user. Looker is relatively user-friendly, and comes with a data modelling language called LookML, which allows you to define business logic in reusable chunks of code. Along with better version control and CI/CD support, this may give Looker the edge over Tableau and Power BI on developer experience at scale. At early stage though, while your team is still small and figuring out the basics, you probably don’t need this.
Sisense (formerly Periscope) is another expensive tool with opaque public pricing. For a fully managed instance, expect a quote of around $35k annually at entry-level, including one admin, one designer, and ten viewers (you may get more if you negotiate). Each additional designer costs $1k per year, and each additional viewer costs $450. It’s reportedly a difficult tool to implement, and Sisense will often provide a consultant to support with this. So, quite the investment.
Which one to choose?
Compared to other parts of the data stack, visualisation tools are the easiest to replace. While data warehouses and ELT tools benefit from heavy vendor lock-in, most BI resources in an early-stage business have a short shelf-life. Company focus, metrics, and data models are liable to change regularly, and dashboards are rebuilt or deprecated accordingly. If you decide to build the next version using a different tool, so be it. This is why it’s a mistake to treat scalability as a significant factor when selecting a BI tool, unless the team is already scaling. Choose whatever will give you the most value for money today, as this decision is reversible.
Below, I’ve summarised the possible monthly cost at varying scales for all the tools mentioned above. The main caveat is that the ordering of tools by cost will vary based on your mix of developer and viewer licenses (e.g. Tableau might work out more expensive than Qlik, in some cases). Rates may also be negotiable as the number of licenses grows, so some of the L or XL tiers may end up cheaper than below.
That’s it from me on how to get the most value for money across your data stack. If you’re triggered by the topic of kitchens or cabbages, I’m sorry for the distress that these posts will have caused you. Otherwise, I hope you’ve found some of it helpful. I’ll be back soon with something else. 👋