Neither column is wrong. The left one is what happened; the right one is why. Only one of them is kept by default.
Shopify Analytics is accurate and it is not a diagnostic tool. Every figure in it is an aggregate of something that already finished: sessions, orders, conversion rate, revenue. The causal layer sits one level below that, in what individual visitors searched for, filtered by and abandoned, and none of it is stored by default. That has one brutal consequence: it is not retroactive. If nothing was recording last month, last month's cause no longer exists anywhere, and you can only diagnose forward from today.
Outcomes are not causes
There is nothing wrong with Shopify Analytics. It does what a reporting tool does: it counts finished things and slices them by date, channel and device. Every number in it is correct.
The problem is a category one. A conversion rate is an aggregate of things that already happened. It can tell you the rate fell from 2.1% to 1.3%. It cannot contain the reason, because the reason is not an outcome, it is a sequence of individual behaviours that the aggregate was built by discarding.
Here is the same distinction applied to a real week:
| The report says | The cause was | Where the cause lived |
|---|---|---|
| Conversion fell 0.6 points | A size filter now returns an empty page | A filter event |
| Sessions flat, orders down | A best seller's main variant went out of stock | A product page view |
| Mobile revenue down 30% | Add to Cart moved below the fold after a theme update | A viewport |
| Bounce rate up on one collection | A theme update reset the sort order | A collection sort |
| Search-driven revenue down | A misspelling of your top product returns nothing | A search query |
Every entry in the third column is an event. Events are not kept unless something keeps them.
A report is a record of outcomes. A cause is a record of behaviour. Only one of the two is being written down by default. The Dropshipping Playbook
The part that cannot be undone
This is the practical consequence, and it is why this chapter sits in the middle of Level 03 rather than at the end.
None of it is retroactive. You cannot install anything today and learn what people searched for in June. There is no archive to open, no export to request. If nothing was listening, the sound was not recorded.
Which means a store diagnosing a drop has exactly two options: work with what was already being captured, or start capturing now and accept that the diagnosis begins today. Most stores discover this on the worst possible day, which is the day revenue moved and somebody went looking for a reason.
When the numbers themselves are wrong
Separately from the blind spot, there is a smaller problem: sometimes the outcome layer is not accurate either. Three known ways.
Merchants have reported Shopify session counts sitting well above GA4 during bot spikes, while conversion rate appeared to collapse from around 2% to under 1%. Nothing about the store had changed. The denominator had.
The tell is a conversion rate that falls exactly as fast as sessions rise, leaving order count flat.
A theme with analytics built in plus an app that adds its own tag produces duplicate events in GA4. Pageviews and add-to-carts inflate, purchases usually do not, so every rate in the funnel drops at once without any single step being at fault.
One documented case saw an apparent 40% fall in paid conversions that turned out to be attribution failing rather than customers disappearing. The signature is specific: a sudden climb in "direct" traffic with no matching fall in conversions anywhere else. Direct is where a tracking system files what it cannot identify.
All three produce a chart that looks like a store problem. The five-minute reality check in chapter 3.1 exists to rule them out before you spend a month on the store.
What to record, starting today
You cannot recover the past, so the only useful move is to make sure this month is diagnosable later. In rough order of what pays back fastest:
- Every search query typed into your store, and which ones returned nothing
- Which filter combinations produced an empty result page
- Products viewed repeatedly by the same session and never added
- Carts opened and then closed without proceeding
- A written change log: every theme edit, app install, price change and shipping rule, with the date
That last one is free, takes ten seconds per change, and turns step six of the 30-minute diagnostic into a glance at a file. Chapter 5.2 makes the fuller case for it.
What this does not fix
Worth stating plainly, because the opposite claim is everywhere. Recording the discovery layer does not tell you why an individual person decided not to buy. Nothing does. What it does is convert an unanswerable question into a countable one: not "why did conversion fall" but "1,400 people searched for a term that returns nothing, and here is what that is worth."
That is a smaller claim than most analytics marketing makes. It is also the difference between the merchants in chapter 3.1 who spent months guessing and a store that fixes one thing on Tuesday.
Common questions
Why don't my Shopify sessions match GA4?
Is a spike in direct traffic good news?
Can I get last month's search data back?
Do I need GA4 as well as Shopify Analytics?
One number pays for the rest
What a single order is worth decides what every other fix can afford. Chapter 4.1 is where that number moves.
Go to chapter 4.1