T11 · Review before summarising
Missing is not zero.
Use one row per project, site, station, night and taxon/group. The explorer checks declared counts and effort. It does not derive passes from raw audio or certify a dataset.
CSV instructions, column definitions and worked example →01 / Open records
Start with a file or a worked example.
No observations opened.
Without a schedule, the tool cannot infer missing nights.
02 / Define and map
Make the assumptions explicit.
Column mappings appear when a file is opened.
Open an example or your CSV to begin. No data is uploaded.
03 / Keep the audit visible
Issues, omissions and decisions.
Audit findings will appear here.
Errors block rates. Correct the source file and reopen it. Warnings stay visible; the tool never silently removes duplicates or converts missing counts to zero.
04 / Compare like with like
Counts and rates answer different questions.
A summary requires a valid audit and your dataset declaration.
Pooled rate = total accepted passes ÷ total accepted recorder-hours. Equal-group mean gives each displayed group equal weight. Neither is a count of individual bats.
Practical questions
Can this compare different detector types?
Only within an analysis design that addresses recording and detection differences. Dividing by hours does not adjust for microphone sensitivity, placement, settings, weather, habitat or taxon-specific detectability.
What happens to unreviewed classifications?
Included rows marked unreviewed or uncertain are omitted from both the rate numerator and denominator. The audit reports them. A classifier score does not become a reviewed count automatically.
Does this replace R, Python or a statistical model?
No. It is a preparation and descriptive-review tool. The existing R/Python resource covers reproducible scripts and separate uncertainty examples. Choose a model that reflects the actual sampling design.