A handful of GA4 settings decide whether the data you collect is usable, and several of them cannot be applied retroactively. Turn them on late and the months before are simply gone. This is not a list of every checkbox. It is the short set where the default is wrong and the cost of leaving it wrong compounds daily.
The ordering principle is regret. Some of these can be changed any time with no penalty. Others silently discard data from the moment the property is created, and no amount of later configuration recovers it. Do the second group today.
The newsletter
Join our KISS newsletter
One short read a week on what actually moves revenue, in a free email. Read by 10,000+ operators and founders.
No spam. Unsubscribe in one click.
I.The settings that are irreversible, and why they come first
Three settings destroy data prospectively. Everything else can wait until Thursday.
A.Data retention, which is the one people find out about too late
GA4 sets event data retention to two months by default. Standard reports are unaffected, which is why nobody notices. Exploration reports, the ones you use for funnels, paths and segment comparisons, are limited to whatever the retention window allows.
Two months is not enough to see a year-over-year comparison, a seasonal pattern, or any B2B sales cycle. The discovery moment is usually someone trying to build a cohort analysis in January and finding that November does not exist.
Change it to fourteen months. It is free, it takes one click, and every day you delay is a day that will be missing from a report you have not thought of yet.
B.Internal and developer traffic
Out of the box, your own team is in the data. Every QA session, every content preview, every support agent reproducing a customer issue is a user in your reports.
For a large consumer site this rounds to nothing. For a B2B site with 4,000 monthly visitors and a team of thirty, it is a material share of sessions, concentrated on exactly the pages you most want to measure, behaving nothing like a customer. Define internal traffic by IP and activate the filter, and remember it is a filter rather than a retroactive exclusion.
C.Cross-domain, if you have more than one domain
If your marketing site and your application sit on different domains, the default behaviour splits one person into two users and attributes the second one to a referral from your own site.
The damage is specific and severe: your best-converting traffic source becomes yourself, and the real acquisition channel disappears. Our cross-domain tracking guide covers the configuration and its limits.
II.The defaults that quietly discard data
These do not lose data permanently, but they change what every report means, and none of them announces itself.
A.Google Signals, and the thresholding trade
Enabling Google Signals gives you cross-device reporting and demographics for signed-in Google users. It also switches on data thresholding, which suppresses rows where the group is small enough that an individual might be identifiable.
That is a genuine trade rather than a mistake. If you mostly look at aggregate trends, Signals is worth it. If you routinely examine small segments, and most B2B teams do, the suppression will remove exactly the rows you came for, and it does so without a clear indication that anything is missing. See GA4 data thresholds for the workarounds.
B.Session timeout and the campaign reset
Sessions end after 30 minutes of inactivity. The window is adjustable up to 7 hours 55 minutes and almost nobody changes it, even when their product is something people leave open for an afternoon.
The second rule is the one that causes reporting arguments: a change of campaign source mid-visit starts a new session. Someone who arrives from your email, searches your brand in another tab, and returns to buy has generated a new session at the moment of the campaign change, and the purchase lands in a session that never saw the earlier steps. This is the root of most funnel disagreements, covered in why your GA4 funnel and revenue report disagree.
C.Unwanted referrals, and the payment provider problem
Any offsite redirect that returns the visitor can overwrite the acquisition source. Payment providers are the common case: a customer arrives from a paid campaign, goes off to pay, comes back, and the purchase is attributed to the payment provider as a referral.
The referral exclusion list fixes it, and it is empty by default. Add every payment provider, every authentication provider and every domain in your own estate.
Settings by regret, which is the order to do them in
Configuration view| Setting | Default | If left wrong | Recoverable? |
|---|---|---|---|
| Data retention | 2 months | Explorations cannot look back | No |
| Internal traffic filter | Off | Your team is in your funnel | No |
| Cross-domain | Not linked | You are your own top referrer | No |
| Referral exclusions | Empty | Payment provider takes credit | No |
| Google Signals | Off | No cross-device, no thresholding | Partly |
| Session timeout | 30 minutes | Sessions split mid-journey | No |
| Enhanced measurement | On | Scroll and outbound noise as events | Yes |
| Conversion marking | None | Ad bidding optimises wrongly | Yes |
4.Enhanced measurement, which is on and probably should not be entirely
Enhanced measurement automatically collects scrolls, outbound clicks, site search, video engagement and file downloads. It is on by default and it is genuinely useful for two of those.
The problem is scroll. A 90% scroll event fires constantly, on short pages fires almost always, and tempts teams into marking it as a conversion, which corrupts ad bidding. Our piece on which events deserve to be conversions covers why. Leave enhanced measurement on if you want, but never let anything it produces become a conversion.
III.What no setting can fix
Four limits are architectural. Configuring around them is the most common way a good team wastes a quarter.
A.The four
Sampling. Above certain thresholds, complex explorations run on a sample and report an estimate. Setting changes do not remove it.
Thresholding. Reduced by turning Signals off, at the cost of the cross-device data you enabled it for.
Consent and blockers. Some share of events never arrives, and the loss concentrates on the pages nearest the money. No configuration recovers an event that was never sent.
The session as the unit. The deepest one. GA4 is built on sessions, and a session is an artefact of measurement rather than something a customer did. Every cross-device and multi-week question runs into this eventually.
B.Where the ceiling actually bites
It bites the moment somebody asks a question about a person rather than a visit. Which customers who read the pricing page in March eventually upgraded. What the people who churned did differently in their first week. Which acquisition channel produced customers still paying a year later.
Those are not configuration problems. They are the difference between a tool that counts visits and one that resolves behaviour to a person over months.
Kissmetrics captures the events itself and resolves them all to a person across devices and sessions, which is why there is no retention setting to get wrong and no session boundary to reason about. It also scans your site and configures the first metrics itself, so the twenty minutes of settings above become a review of a draft rather than a checklist somebody has to remember.
None of which is an argument for switching off GA4. It is very good at what it is for, it is free, and it is the tool your ad platforms expect. The argument is that the person-level questions are a different job, and no amount of configuration turns one into the other.
Verdict
Four settings destroy data prospectively: retention, internal traffic, cross-domain and referral exclusions. They take about twenty minutes and every day of delay is permanently missing from a report you have not thought of yet. Do those first, today.
Then stop configuring and check whether your remaining problem is architectural. Sampling, thresholding, consent loss and the session as a unit are the ceiling, and teams routinely spend a quarter tuning settings against a limit that no setting touches.
Continue Reading
How to Audit GA4 for Data Accuracy (And What to Do When the Numbers Don't Add Up)
If you have ever compared GA4 numbers to your backend and found a 20-40% gap, you are not alone. This guide provides a systematic audit process to identify where your data is leaking and what to do about it.
Read articleGA4 Data Thresholds Are Hiding Your Reports: Here's How to Work Around Them
When GA4 hides your data behind thresholds, you lose visibility into exactly the segments that matter most. This guide explains why it happens and gives you 5 practical workarounds.
Read articleWhy Your GA4 Funnel and Your Revenue Report Disagree
Nothing errors. The report renders, the percentages look plausible, and a team spends a quarter optimising a step that was never the problem. Here is how the gap opens and how to find which cause is yours.
Read article