FAQ & troubleshooting

Fast fixes for the usual Cohortum snags — failed loads, rejected rows, missing events, off numbers, cohort build failures, and no-access links.

Almost every Cohortum snag traces back to one of three things: a mapping you picked in the upload wizard, an event you excluded, or a filter that's narrower than you remember. Here are the questions we hear most, and the quickest fix for each.

Loading data

A load failed or shows rejected rows — what went wrong?

Open the source on the Sources page and expand its Loads history. Each load lists Rows loaded and Rows rejected, so you can see how much data actually made it in. Rows get rejected for one of three reasons: a timestamp Cohortum couldn't read, a required column you didn't map, or a file that doesn't match the type you picked.

You need three columns: Event Name, User ID, and Timestamp. If one is missing or points at the wrong column, Cohortum drops the rows that don't have valid values instead of guessing. For the exact file types, required columns, and timestamp formats we accept, see Supported formats & limits.

Auto-mapping mapped a column wrong — how do I fix it?

Read a file into the wizard and Cohortum guesses your columns for you, then flashes a toast like "Auto-mapped N columns — review the mapping below." Take it as a head start, not the final word — always look over the Mapping step yourself.

In the Map your columns table, open the Map to dropdown on any row and set it straight. Your choices are Event Name, User ID, Timestamp, User Properties (JSON), Event Properties (JSON), and Ignore. Anything left on Ignore stays out of the import. Fix the mapping, then re-run the load. The full wizard walkthrough lives in Sources & loads.

Sessions look wrong

Sessions come from the wizard's Sessions step. Turn on Auto-generate sessions and a new session kicks off after N minutes of quiet, where N is the Session timeout (minutes) field. Set it too short and one visit shatters into a dozen sessions; set it too long and separate visits blur into one.

Already have a session ID in your data? Point the Session column at it instead of auto-generating. If your session counts look off, this is the first setting to check.

Events and numbers

My event isn't showing up

A missing event is usually just hidden. Open the Events catalog and look at the Excluded status pill — an excluded event disappears from every analysis until you turn it back on. Mind the scope, too: an event can be excluded for just you (Personal) or for the whole Project.

If the catalog says the event is active, the culprit is your analysis. A filter, a tight date range, or the Min users: N% threshold may be pruning it out. Loosen those and rebuild before you decide the data's gone.

The numbers seem off or lower than expected

Three settings quietly shrink what you see. The Min users: N% threshold on the toolbar hides low-volume steps, transitions, and paths — drop it to bring back the long tail. Your Filter and date range in the left panel scope every number as well, so a filter you forgot about is a common offender.

Large sources get computed on a sample so results come back fast. The header tells you how much you're seeing (say, "73% from filtered"), and a Query more users button shows up when a sharper pass is ready. Click it to recompute against more of your data. For how these scoping controls play together, see Filtering & date ranges.

Cohorts and sharing

A cohort shows "Build failed"

On the User Cohorts page, a cohort can land in Build failed — usually because no users matched its definition, or its source changed and the definition no longer matches anyone. You can't rebuild in place: delete the failed cohort and make it again. If the definition was too narrow or pointed at a source that's since changed, fix that first — widen it or point it at the right source — then recreate the cohort.

Cohorts built from a saved filter (the Saved filter origin) are virtual: they re-run on every query, so they rarely fail. The ones that fail are the materialized kind — drill-click segments, cluster and transition origins, and CSV uploads. Each of those freezes a member list when it's built, which is exactly why they're the ones you'll recreate after the data changes.

A shared link shows "no access"

A share link carries the analysis, not the keys to the data behind it. Copy a Share link and a teammate can open it only if they already have access to the source it's built on. If they don't, they get a stripped-down view titled No access to this source, telling them to "Ask the source owner or a workspace admin to grant you access."

The fix lives on the Members side, not in the link. Have an Owner or Admin grant the recipient project access, or make the source Public so anyone in the workspace can reach it. Send the same link again and it works — no need to regenerate it.

Access and sign-in

Why don't I see the New source or upload button?

Uploading is gated by role. Owners and Admins can always upload; Viewers never can; a Creator can upload only when someone's flipped on the Upload sources switch for them. With it off, the Sources nav item disappears and you get an "ask a workspace admin to add one" empty state instead of the wizard. Ask an Owner or Admin to turn on your Upload sources switch — Members & roles covers who can do that.

I was suddenly signed out

Getting bounced to the login screen mid-session usually isn't a crash. When Cohortum can't match your account to an active workspace member, or your session token is invalid, it signs you out and sends you to /login on purpose. Just sign back in. If it keeps happening, your membership may have been revoked or never provisioned — ask an Owner or Admin to check your account on the Members & roles screen. A token quietly expiring mid-session won't sign you out, though — Cohortum refreshes it for you.

Still stuck?

Don't see your problem here? Reach the Cohortum team from the in-app help menu or email support@cohortum.com. Include the source name and, when it's relevant, the analysis link, so we can see exactly what you're seeing.

Related