What Businesses Should Do Before Even Thinking About Using BigQuery

If you use Google Analytics, chances are you've heard of BigQuery by now. After all, BigQuery is the "second product" that Google is selling with the free version of GA4 (after Google Ads and I guess GA4 360).

BigQuery is incredibly powerful and it enables you to pretty much answer any business question by using it - building custom attribution models, comparing over long dates ranges, answering complex funnel questions, etc...

...most businesses have so much work to do before they should even think about using BigQuery, however.

They are trying to run a marathon after a week of running - it's possible, but without a foundation you will have to go backwards in your training.

Before I get into what you should be doing instead, you should setup the link between GA4 and BigQuery so that you are storing your data.

Before we ignore BigQuery, set up the link!

This way, you will be storing your GA4 data (and owning it) for very little cost every month if/when there is a time you need to use it.

The most we've seen an account spend for storage costs every month is $6 (hundreds of thousands of total users every month to the site).

What should come before BigQuery then?

Now that we own our GA4 data and will be able to answer deeper questions when the time comes, we can focus on the fundamentals of data driven Marketing.

Planning

You should have a dedicated phase of planning in your analytics.

• Start by listing out your business questions you want to have answered (notice how I didn't say analytics questions - don't restrict your thinking, just list 'em!)
• Map your actions to your questions (play the hypothetical game here - "If I was to tell you the answer to the question was 'x' - what would you do with it?" (If you can't answer this, de-prioritize it from the building phase since it's by definition a vanity metric)
• Document the user journey so the "funnel" is a shared definition - no assumptions are made:

Building

• Take those questions you have, build the data collection in a thoughtful* way in GTM (don't waste your time with event creation in GA4)
• Use explore reports in GA4 (they are incredibly powerful) to answer the questions you originally set out to answer
• Use Looker Studio to tell a much better story and have a highly interactive view into your data (filters are a nightmare in explore reports, use 'cross filtering' in Looker Studio to save a lot of time and headaches).

Reframe how you think about Measurement

"Did this happen like it was supposed to" is one of the most powerful ways to shift your thinking about Measurement.

It amazes me how many Marketers will not do this, yet resort to questions like "Why aren't my Meta Ads matching GA4 with sales numbers last week?" (Spoiler, they will never match and that's because their definitions are different...switch your thinking to trends and patterns and "if Meta is up 10% in sales than GA4 should be up about 10% as well"...ironically, numbers rarely matter in Measurement).

What this reframe is doing is forcing you to use the Scientific Method in your thinking. Literally no better way to be a data-driven Marketer. He's a recap of what that is:

The Scientific Method

1. Ask a Question
2. Research
3. Form a Hypothesis
4. Experiment
5. Collect and Analyze Data
6. Draw a Conclusion
7. Communicate Results

Now of course using all these steps for every part of Marketing is complete overkill. But try applying these concepts and you'd be amazed and how your actions change.

Step 3 is by far the most missed part I see with Marketers. They aren't forming hypothesis.

I often say, "What are your expectations here?" to avoid the H word (it can feel intimidating, especially if you aren't running a controlled test/don't do that regularly).

...but if you aren't setting expectations, you won't know what you are looking for in the data. (Great way to waste time in GA4)

You will use the two "I words" in data that I hate:

• Insights
• Interesting

Why do I hate those? Usually no actions are ever taken after people use those words.

(Great podcast on how "insight" is terrible over-used: https://share.snipd.com/episode-takeaways/09ae9f2c-4bca-4154-a6c4-6be88a29d568) (And your new favorite podcast player if you haven't heard of Snipd before 🤓)

I wrote more here on how you should be setting up goals to match the user journey/full conversation.

Qualitative Research = No Longer Optional

This is the biggest miss for Marketers I see today. Before you ever write a line of SQL in BigQuery (And no, Generative AI writing it for you still counts) you should be doing these forms of research regularly (one per month, minimum):

• Behavioral data (Microsoft Clarity does a great job, we use that for our clients) - Heatmaps and Video Recordings can give you a quick idea of engagement on pages. For a more in-depth setup of copy measurement, check out my edition on Element Visibility.
• Website Surveys - Probably one of the most underrates forms of research on users. Exit intents and TY pages are my "go to" places to start. Bottom of funnel conversion pages (checkouts, form pages, sign up flows) are a no-brainer to be asking "What is the #1 thing holding you back from purchasing?" Replace "purchasing" with whatever action is relevant for that page. Analytics data will never tell you this. You can only make assumptions.
• Incentivized Surveys - Usually the first thing we start with for a client if they haven't done this (the right way) in the last 6 months. We recommend to incentivize these (when done right, the responses you get back are a goldmine of actions and copy in the words of your customers).

Start with buyers/sign ups/leads (whatever your business model) in the last 3 months, but if you only have so many, then you'll need to go back further to 12 months. You don't need "statistical significance" with any form of qualitative research, you need saturation. In this case, themes of responses.

Look at how few responses you really need given your sample size (list you are sending to):

Tool I used to calculate that, here: https://www.qualtrics.com/blog/calculating-sample-size/

I'll have a dedicated edition just talking about incentivized surveys (including a coding template) but just know these are some of the most actionable data you can get about your customers. Forming hypothesis (and copy) is SO much easier when you are basically in the minds of your customers.

P.S. I like to start with buyers here for a reason (and yes, of course it's a biased audience, but it's a healthy bias because we want to understand what caused them to act...you can also survey non-buyers - leads that didn't convert, and comparing those two data-sets is where the magic really starts happening).

• Talking to Customers - When did we stop doing this and why? Oh, we have 20 more KPIs to report on and a reduced budget, makes sense - there is no time, right? Wrong. Talk to 1 customer a week for a month and you will know more about your customers, their pains, and have better empathy than 99% of Marketers out there.

Alex Hormozi says "You can solve all of the problems in your business if you simply talk to enough customers".

(I don't disagree at all)

• User Testing - 5 second tests, copy tests, preference tests, tree testing, moderated (task-driven), unmoderated (task-driven), card sorting, dairy studies, etc...

Marketers could learn a lot from Product Marketers here.

• A/B Testing - People forget that this is a form of research, you don't always just say "Oh this won so therefore we have to turn it live". I defer to Ruben de Boer's Hierarchy of Evidence:

• Meta Analysis - Combining research together to truly use data informed rich decision-making actions. This is why you should have a repository of research, priorities, experiments run, etc so that seeing these trends and patterns becomes easy. It's impossible to make good decisions without this. (Why we recommend and use Airtable)

Diminishing Returns in BigQuery

Yet, we want to be building custom attributions models in BigQuery and we haven't talked to a customer ever...but we "sent out that NPS survey after they purchase" 🙄

BigQuery is awesome. But you have to grow into it. Do the fundamentals of Marketing first. Run a 5k (and 10k and probably half marathon) before you run 26.2.

I heard an analogy that compares BigQuery to a DSLR and GA4 to an iPhone that I liked...

...An iPhone takes great photos - often indistinguishable from a DSLR.

Most businesses think they need a DSLR (BigQuery), but they’d benefit more from mastering iPhone fundamentals (GA4): rule of thirds, framing, lighting, etc. - paralleling the key marketing fundamentals of Measurement and Qualitative research I outlined here.