A reporting tool has failed when two analysts export “last week” and cannot explain the delta without opening a Google help article. GA4 gives you several ways to get a different number from the same property: sampled Explorations, thresholded cards when Google signals is on, consent-gated collection, and the ordinary confusion between users, sessions and events. Teams then blame “the data team” for a product that is working as designed.
SiteLine does not sample on-site pageviews and custom events. That is not the same as “SiteLine has infinite precision” or “SiteLine replaces GA4’s warehouse”. This page separates Google’s sampling and thresholding from Search Console’s row cap, and from SiteLine’s visibility caps — three limits that get mashed into one Slack thread.
Sampling in GA4 is mostly an Explorations problem
Standard GA4 reports are generally unsampled for ordinary site traffic. Explorations (free-form, funnel, path) are where sampling shows up once you exceed the event quota for the exploration — on the order of 10 million events at the standard tier, far higher on 360. The UI will say the report is sampled. People screenshot anyway.
The practical failure: a funnel you built in Explorations for a board deck used a sampled 12% of events, while the same conversion as a standard report did not. Nobody labelled the slide. Next month the exploration is unsampled because traffic dipped, and the conversion “improved”.
360 and BigQuery export are Google’s answers. They are also why “just use GA4” is a platform decision, not a tag decision. SiteLine does not offer BigQuery export. If unsampled raw hits in a warehouse are the requirement, you are not shopping for cookieless page analytics.
Thresholding is a privacy feature that wrecks small sites
Turn on Google signals (or otherwise trigger Google’s identity mixing) and GA4 may hide or card numbers so individuals cannot be reverse-engineered from a slice — country × device × landing page on a quiet Tuesday. The report looks empty or oddly rounded. You then disable signals to see the real counts and lose the cross-device story you enabled them for.
This is coherent as a privacy control. It is miserable as an analytics product for a SaaS marketing site with a few thousand visits a day. Cookieless tools with daily hashes avoid this particular Google-signals threshold; they also cannot offer that cross-device stitch. See cookieless vs GA4.
Search Console’s 25k cap is not sampling
The Search Console interface shows 1,000 rows. The API will return up to 25,000 rows per request. Request a date range with dimensions and you silently truncate everything after the first day’s budget — which is why any serious importer (SiteLine included) requests one day at a time for query and page breakdowns.
SiteLine imports up to 25,000 rows per day per report, backfills 16 months, re-imports the last 7 days every 4 hours, including fresh rows, because Google revises recent days. The newest day can still change. CTR is computed from clicks and impressions. Queries Google anonymises never appear in query reports, so query totals can sit below site totals. That is Google, not a SiteLine filter.
| Limit | What it does | Who imposes it | What SiteLine does |
|---|---|---|---|
| GA4 Exploration sampling | Uses a subset of events above a quota | Google Analytics | Not applicable — no Explorations product |
| GA4 data thresholds | Hides/cardes small slices when signals apply | Google Analytics | Not applicable — no Google signals |
| GSC UI 1,000 rows | Truncates what you see/export in the web UI | Search Console | Import via API instead |
| GSC API 25k rows/day/report | Truncates the long tail of queries/pages | Search Console API | Import up to that cap, one day at a time |
| Anonymised queries | Omits queries from query reports | Google Search | Cannot invent them; site totals still import |
| Starter/Growth visibility cap | Hides post-quota on-site events until upgrade | SiteLine billing | Soft ingest continues; Scale meters $2.99 / extra 1M |
The last row is SiteLine being honest about its own ceiling. Events keep ingesting after 1M (Starter) or 10M (Growth); charts hide the overflow until you upgrade. Trial visibility is about 50k events. That is a billing gate, labelled in the product — not silent sampling. Quota rules: /docs/billing.
Unsampled still needs an identity you can explain
An unsampled count of a bad metric is still a bad metric. GA4 “users” after a 50% consent rate is an unsampled description of the consented population. SiteLine unique visitors are an unsampled daily hash: truncated IP + UA + salt, reset UTC midnight, no cookie. Goals and funnels are derived from events you already sent (/docs/goals), so you are not sampling a second pipeline — you are slicing the first.
If the weekly fight is “why doesn’t this match last week’s Exploration?”, leaving GA4 for cookieless analytics plus Search Console is a reasonable move. If the fight is “we need hit-level export into the warehouse”, stay on GA4 360 / BigQuery or a product-analytics tool. SiteLine will not grow a warehouse connector to win that paragraph.
MCP access (Claude, ChatGPT, Cursor) includes coverage notes so a model does not treat Google’s two-day Search Console lag as a collapse, and so soft-caps are visible in the answer. That is more useful than an unlabelled sampled funnel.
Related: why teams leave GA4, SiteLine vs Google Analytics, GA4 for SaaS. Signup: /register.
FAQ
Are SiteLine on-site reports sampled at high volume?
No. Volume is a visibility or overage issue, not a sample rate. Starter and Growth hide data after the plan cap; Scale bills overage. Upgrade unlocks hidden history on the capped plans.
Why is my query total lower than site clicks in Search Console?
Google omits anonymised queries from query breakdowns. SiteLine stores what the API returns. Compare site-level totals (date dimension) to query reports and expect a gap; it is documented Google behaviour, not a lost import.
Can I get GA4-style Explorations on SiteLine?
No. You get pages, referrers, countries, devices, UTMs, goals and funnels on data you already collect. There is no sampled exploration canvas and no BigQuery dump. If you need arbitrary slices of raw events, you are in product-analytics territory.