[Metrics] Create Expense flows

[rlinoz]

Tracking issue for the Create Expense Flow Sentry dashboard.

Background

As we migrate customers from OldDot to NewDot, performance becomes critical. If NewDot feels slow or sluggish for common workflows, migrated customers become frustrated, stop using the app, and potentially churn. This puts existing revenue at risk and kills our word-of-mouth lead engine.

The create expense flow is one of the most frequently used features, so its performance directly impacts user perception of the app.

Goals

• Keep all interactions in this flow under 400ms (P90), since they shouldn't require network round-trips
• Ensure the metrics accurately capture real user flows and distinguish between different expense types (manual, distance, scan, split, track, invoice)
• Identify and fix outliers affecting the top 10% and 5% of slow experiences
• Continuously lower the target threshold as improvements land

Dashboard Structure

This dashboard tracks multiple sub-metrics across the expense creation flow, not just one number. Each widget measures a different part of the user journey.

[mountiny]

Note, this one is a bit different than others, where it is a health dashboard that consists of bunch of substeps, so it is not just one number

[marcaaron]

Sounds like the fun one.

[marcaaron]

Spent some time with Sentry today and the "Submit Expense" widget.

I think this flow is actually quite complex as the current dashboard is set up to catch some pretty broad stuff. And the metrics are maybe polluted or not really what we want exactly since the individual samples/traces come back with stuff like:

• RequestMoney ✅
• CreateDistanceRequest - well, this is a different flow with the same previous page origin 🤔
• TrackExpense 🤷
• Navigates to SearchMoneyRequestReport after confirm screen 🤔
• People sometimes bail without calling any APIs e.g. just navigated from one page to another, but didn't actually create any expense. Should maybe be tracked separately and seems like less of a priority. 🤷‍♂️

So, I really think next we just need to work on improving the dashboard and making sure it actually returns useful measurable scenarios otherwise we won't know if we are moving the needle on any specific scenario.

For now, I have been looking at and targeting spans that take around ~13 seconds and trying to uncover patterns.

[marcaaron]

I think a lot of these are people uploading uncompressed receipts on bad wifi 😂

[OlimpiaZurek]

Hi! I'm Olimpia from Callstack and I'd like to work on this.

[marcaaron]

Let's goooo!