Piano Analytics Experts

The absence of sampling is the most useful property day to day: a figure read in the interface is the figure, not an estimate. Eco-mode itself samples nothing, it estimates deduplicated counts with an average relative error stated below 2%. The exemption regime does exist and does work, but it is not a checkbox: the documentation requires masking the visitor identifier, anonymising IP addresses, purging at twenty-five months across the whole organisation, measuring on your own domains and apps only, and crossing no imported data. A contract annex seals it.

Raw data leaves through two distinct doors, and confusing them is expensive. Data Flow delivers incremental files to FTP, Amazon S3 or Google Cloud Storage. Data Sharing opens the underlying Snowflake database in SQL, where a table reprocessed nightly carries the computed properties the stream does not have, visit duration, entry and exit pages. There is no connector to another warehouse. And the Reporting API caps at two hundred thousand rows per request, with no pagination workaround.

Piano measures what reaches your own surfaces. It does not separate traffic from Google's generative answers from the rest of Google traffic, and the vendor says so plainly: the referrer is identical in both cases. Its dedicated assistant source only covers those serving their answers on their own domain, and it is not retroactive. More broadly, an audience measurement tool attributes on what it observes: impressions seen without a click and offline conversions escape it.