Method
How we know when mail lands
1. Observations, not scans
Panelists photograph mail pieces when they find them in the mailbox. Each photo carries a timestamp and a region (the first three digits of a ZIP code). We never collect street addresses.
2. Matching by artwork
When you register a campaign, AI fingerprints the design — headline, brand, colors, format. Incoming photos are matched against that fingerprint. Strong matches are accepted; uncertain ones are reviewed; mismatches are rejected with a reason the panelist can see.
3. The curve and its band
For each day after the drop, we show the share of observed pieces that had landed. The shaded band is a 95% interval around that share — wide when few observations exist, narrow as they accumulate. The observation count is printed next to every curve.
Curves are recomputed as new snapshots whenever observations arrive. Old snapshots are never overwritten, so you can see how the answer changed.
4. When we can't answer
Below a minimum number of accepted observations, or where too few panelists live in your target regions, we show no curve at all — just a clear "we can't answer here." A thin guess is worse than no answer.
What this is not
Not delivery tracking. We measure arrival in aggregate and never identify a specific household.