356: The Right Way To Setup Your Website Analytics With Chris Mercer

Podcast: Download (Duration: 58:12 — 66.9MB)

Setting up Google Analytics for ecommerce starts on paper rather than in the platform. Chris Mercer’s QIA framework has you write down the questions you want answered, the information needed to answer them, and the action you will take based on each answer, all before opening analytics at all.

In this episode I sat down with Chris Mercer, founder of MeasurementMarketing.io, who specializes in getting businesses to actually use the analytics they already have installed.

Below is the whole system: why turning analytics on is not setting it up, the QIA process, how to measure campaigns without solving attribution, and how to diagnose a broken funnel from numbers alone.

Key takeaways

• Installing the code is not setup. That step only turns it on.
• Use the QIA framework: question, information, action, written down before you open analytics.
• Multi-touch attribution cannot be solved. Stop trying and measure campaigns individually.
• Give each campaign one job. Shorter time windows are easier to measure accurately.
• Ask results, then how. Every answer generates the next question down the funnel.
• Expect 1% to 3% overall conversion. Below that, something in the journey is broken.
• Every page has one job. A product page cannot sell, only produce add-to-carts.
• Forecast and be wrong. Guessing wrong generates the data that makes you right later.

Why turning on analytics is not setting it up

Installing the code or enabling a Shopify integration only lights the platform up, and most people mistake that for configuration.

Actual setup means creating views, configuring ecommerce settings, and defining goals, which is how you tell Google Analytics what you want reported back.

That step gets skipped most often, which is why so many store owners open analytics, look at visit counts, and close it again.

The QIA framework

Before opening any analytics platform, work through three columns on paper: question, information, action.

Q for question. Write down anything you want to know about your store. How many transactions am I making? What is my average order value? How many people see product detail pages? How many start checkout and abandon?

There are no wrong questions at this stage. This is a skill you build like a muscle, and early attempts being clumsy is expected.

I for information. Next to each question, write what you would need to measure to answer it. Average order value requires total revenue and transaction count, which you divide.

If you cannot identify the information for a question, set that question aside. It is not that the question is bad, it is that you need foundational answers first, like building a foundation before walls.

A for action. This is the part almost everyone omits. Write down what you will do based on the answer, always as a number tied to a specific action.

The format: if it is fewer than ten sales a day, I will check my traffic sources for one that underperformed. If it is more than ten, I will do something else.

Where the number comes from is your own baseline. If you typically get ten orders a day, ten is your threshold until you deliberately change something.

Why QIA matters more than the tool

Entering analytics without a question is what Mercer calls going in unarmed, and it produces overwhelm rather than insight.

The common pattern is opening the platform to see what you can see, getting frustrated, and deciding to come back in a few years.

Arriving with a question, the information you need, and a predetermined action makes every platform easier to use, because you now have context for what you are looking at.

The framework then compounds. Run it on transactions, then average order value, then whichever metric you touched last. After a week or two, you can answer large diagnostic questions like why people are not buying, because you already know which specific component failed.

Why multi-touch attribution cannot be solved

Mercer’s position is that multi-touch attribution has no answer and people should stop trying to find one.

His evidence is longevity. Marketers have asked this question for well over a century, which strongly implies it is not solvable rather than merely unsolved.

The reason is time. If you click a Facebook ad Monday, become a lead Wednesday after a Google search, and buy Friday from an email, the email closed the deal. Facebook contributed and did not close it.

His sales team analogy makes it concrete. If you employ three salespeople and email closes every deal, no sales manager would pay Facebook a commission for having spoken to the prospect first.

How to measure campaigns without solving attribution

Give each campaign one specific job with a short time window, then judge it only against that job.

The reasoning is measurement accuracy. Less elapsed time between the interaction and the intended result means a cleaner signal.

The sales pipeline analogy: some salespeople set appointments, others close. You do not judge an appointment-setter on closes. If a Facebook campaign’s job is getting people to read a blog post that educates them, that is what you measure.

If that campaign happens to produce a sale, that is welcome and it is not the point.

The important nuance is that this operates per campaign rather than per platform. One set of Facebook ads builds awareness, another drives product page views, and a third retargets people who abandoned checkout with a discount. Every traffic source can work at every stage.

What you learn from measuring this way

Once campaigns are individually measurable, you can genuinely compare traffic sources at each stage.

The questions become answerable: what is best at creating awareness, what is best at generating leads, and what is best at closing.

The picture assembles itself from there. Google organic might drive awareness, Google Ads and Facebook might convert awareness into leads at roughly 30%, and email might close 10% of those.

That produces a forecast. Next week a hundred people enter the store, thirty add to cart, ten complete checkout at a hundred dollars each, which is your expected revenue. Then you measure whether the store behaved as predicted.

How to generate the right questions

Ask what results you are getting, then ask how you are getting them, and repeat.

The sequence looks like this:

• How many sales am I making? People completing checkout.
• How are they reaching checkout? By clicking add to cart.
• How are they reaching add to cart? By viewing product detail pages.
• How are they finding product detail pages? By arriving from a traffic source.

Each answer produces the next question, and the chain reveals your customer journey without anyone handing you a list.

Mercer is dismissive of the seven-questions-every-store-should-ask genre, because those lists were designed to answer someone else’s questions about their own business.

How to diagnose a funnel from numbers alone

When a step deviates from expectation, look at the step immediately before it.

Mercer’s own example: his offer page normally sends 10% of visitors to the cart, and 35% of those complete purchase. One week the completion rate collapsed to around 10%.

Rather than examining the pages, he checked the preceding number. Traffic from the offer page to the cart had jumped from 10% to roughly 30%.

That combination told the story. Far too many people were reaching the cart, which meant they were arriving before they were ready to buy, which is why they were not converting.

Going to the offer page with that specific hypothesis revealed the cause immediately: someone had removed the price. People clicked through to the cart to find out the cost, then bounced back to the sales page.

It was a five minute fix, and the numbers returned to normal with more total purchases than before.

The expectation engine

Every step sets an expectation that the next step must meet, and most breakdowns are a mismatch between them.

The clearest example is ad-to-landing-page. An ad promising 10% off a summer sale landing on a page announcing a winter sale creates a disconnect that damages opt-in rates.

The fix is either direction. Change the ad to match the page, or change the page to match the ad.

For opt-in rates specifically, Mercer expects 25% to 35% from cold traffic. Below 25% he goes back to the ad rather than the page.

When a step underperforms, the diagnostic question is which expectation was set and which was not met.

How to know if your numbers are good

Roughly 1% to 3% overall conversion is the general benchmark, and everything below that has to be judged in context.

Benchmarking against other stores is useful and dangerous. Someone reporting 47% conversion on product detail pages might be converting only 2% of those to purchases, so a single number without the full journey is meaningless.

Mercer’s example of a misleading benchmark: someone reporting 80% checkout completion is almost certainly seeing returning visitors with coupons, which is not scalable and would make a store doing a healthy 40% feel like a failure.

Business model changes everything. A store selling 45 variations of one product has completely different browsing behavior than one selling three dog collars and two treats.

The practical starting point is measuring what you have, treating that as your baseline, and improving against yourself.

Why every page has exactly one job

A product detail page cannot sell anything, because you cannot hand it a credit card.

Its job is producing add-to-carts. Judging it on revenue misattributes the work entirely.

The same logic runs through the funnel. The cart’s job is getting people into checkout, and checkout’s job is facilitating payment.

This makes measurement dramatically easier, since each page has one number that tells you whether it is working.

How to diagnose a page that is not converting

Behavior data tells you which of two completely different problems you have.

If people spend five seconds on a product page, ignore the images, skip the reviews, and never scroll, the problem is upstream. Wrong audience, wrong product, or a headline that did not match what brought them there.

If people examine images, spend three minutes reading reviews, scroll through the details, and still do not click add to cart, the problem is the page design. Mercer has seen carts with four red buttons doing four different things, where finding the right one requires too much effort.

Same symptom, opposite fixes. Measurement is what tells you which one you are looking at.

The measurement journey

Nobody starts sophisticated, and trying to skip stages is what makes people quit.

Stage one: light things up. If analytics is off or you have not opened it in a year, just start looking. Get familiar.

Stage two: the valley of visibility. You begin interacting with reports and noticing things.

Stage three: level up implementation. Now you set up goals, measure add-to-cart clicks, and configure what was missing.

Each round of bigger questions requires another implementation upgrade. A question you cannot answer today usually means your implementation is not there yet, which is fine.

Why you should forecast and be wrong

Guess how your store should perform, expect to be wrong, and use the gap to get better.

The five pillars Mercer teaches: planning through QIA, building out the integrations, reading the reports, forecasting, and then acting.

The tendency people fight is needing to be right immediately. Being wrong in the beginning is close to guaranteed, and the wrongness is what generates real data.

What improves is the questions. Each cycle makes your questions and forecasts better because both are now grounded in what actually happened rather than what you assumed.

How often to look at your numbers

Frequency should match your role rather than following a universal rule.

Mercer reviews funnel-level numbers weekly across product lines, while his marketing team checks daily because moving those numbers is their job.

His framing: whoever is responsible for making a number move needs to be in that number often. An investor does not need daily funnel alerts, and a marketer does.

Collection is automated once configured, and reviewing is not. Google Data Studio is free and builds dashboards that color-code steps red, yellow, or green against your expected percentages.

His argument against alert-only monitoring is that regular exposure surfaces trends and insights you did not know to look for.

The curse of a good offer

Not knowing how your machinery works is dangerous specifically when things are going well.

The mindset is common: as long as money is coming in, the details do not matter. That holds until the day it stops.

The failure scenario is scaling to $50,000 a day on a platform, having the results vanish without warning, and discovering nobody in the company knows how anything worked.

Mercer’s uncomfortable observation is that businesses without a working offer are frequently in a better position, because they are forced to understand their numbers.

The underlying principle is that sales are a symptom of a system. Focus on the machinery, and the results follow.

Frequently asked questions

How do you set up Google Analytics for an ecommerce store?

Installing the code only turns it on. Actual setup means configuring views, ecommerce settings, and goals so the platform reports what you need. Before any of that, work through the QIA framework on paper to decide what you are trying to learn.

What is the QIA framework?

Question, information, action. Write down what you want to know, what information would answer it, and what specific action you will take based on the answer, with a number attached. All of this happens before you open any analytics platform.

Can you solve multi-touch attribution?

No. Marketers have asked this question for over a century, which suggests it has no answer. Instead, give each campaign one specific job with a short time window and measure it only against that job rather than against final revenue.

How do you know what questions to ask about your store?

Ask what results you are getting, then ask how you are getting them, and repeat. How many sales, how are they completing checkout, how are they reaching checkout, how are they finding product pages. Each answer generates the next question.

What is a good ecommerce conversion rate?

Roughly 1% to 3% overall as a general benchmark. Single metrics from other stores are misleading without the full journey, since a 47% product page conversion means little if only 2% of those buyers complete checkout.

How do you diagnose a funnel step that is underperforming?

Check the step immediately before it. When Mercer’s cart completion collapsed, the cause was traffic to the cart tripling, which meant people were arriving before they were ready to buy. The offer page had lost its price.

How often should you check your analytics?

Match frequency to your role. Whoever is responsible for moving a number needs to see it often, so marketers check daily while owners might review weekly. Collection is automated once configured, and reviewing is not.