Studio · First-party correlation
Your numbers, against culture signal
Upload a daily CSV of revenue, sign-ups or units. We align it to an entity's composite and scan every lag from 0 to 28 days where the culture signal leads your metric.
“Upload series”
LOCKEDCSV only, max 2 MB and 5,000 rows. Required columns after mapping: date and value. Optional: metric_name, spend, channel. Validation runs server-side and names the failing row.
UNLOCKS WITH STUDIO
“Lag scan”
LOCKEDPearson correlation at every lag 0–28 days, returned with the full lag curve, best lag, r, n and a p value.
UNLOCKS WITH STUDIO
“What we do with your data”
- Rows are written under your own account and restricted to you by row-level security.
- No other user and no anonymous visitor can read a byte of it.
- It is never used to train anything.
- It is deletable by you at any time from your account.
“Read this before acting”
This is a correlation between two time series, not evidence that one causes the other. Both may be driven by seasonality, a campaign already in market, or an unrelated third factor. Correlations found by scanning many lags overfit easily. Treat a result as directional only when the relationship holds across multiple entities, survives a holdout period, and matches something you can explain.
Additional rules
n under 30 renders muted and labelled LOW CONFIDENCE, SMALL SAMPLE. A best lag at the edge of the scan range, 0 or 28, is flagged as likely spurious.
The product reports the relationship. It will never render a recommended spend number.