Dashboards
13 dashboards
Business
Customer Volume
Business view of billable volume from laminr-prod: processed (billable, status=2) vs submitted loan packages per day with support-ticket load overlaid as a cost signal; time-to-processed p50/p95 from the package state history (packages_historicalloanpackage); customer trends (top 14 customers + Other) with counts/share views; and a range-aware top-customers table with share, billable rate, growth vs the prior window, and new-customer badges. Internal/test tenants excluded (Laminr, demo and pentest accounts). Ranges: 7/30/90/180 days. Support tickets from the Open Support Tracker 'Laminr Tickets' tab. Static snapshot captured Jul 20 2026.
Prospect Engagement
A sales working view of new prospects: tenants created in the last 7 or 30 days, and how far each has got. Tracks the onboarding ladder (signed up -> invited a teammate -> invite accepted -> someone signed in -> first package -> first processed -> came back -> active now), with a per-prospect roster showing invites sent/accepted/pending, seats, last sign-in, days to first package, and 14-day momentum. Each prospect is classified No invites / Invite pending / Onboarded / Dormant / Fading / Steady / Growing so a rep can see who is stuck and at which step. Two benchmark panels (activation curve and post-signup ramp) sit at the bottom and deliberately cover ALL prospects, not the selected window, so a new cohort can be read against the historical norm. Internal/test tenants excluded. Static snapshot captured Jul 20 2026.
Customer Experience
Customer-satisfaction view of laminr-prod: unique active customers and users per day and per window (with prior-window growth); time to first terminal state (under review OR processed — how fast the customer gets an answer) vs time to final processed; review rate (share of packages needing support review); and support latency (review → processed — the support team's contribution to wait time). Fast processing = happy customer. Internal/test tenants excluded. Ranges: 7/30/90/180 days. State-history coverage begins Apr 13 2026. Static snapshot captured Jul 20 2026.
Customer Health Scorecard
One row per customer combining the other dashboards' snapshots — 30-day processed volume and growth (Customer Volume), override burden, packages touched, and zero-touch rate (Overrides), and daily active users (Adoption) — into a simple health grade: Healthy, Watch, or At risk. Built entirely from sibling data.js files with no database dependency; refreshes after the source dashboards each morning. Static snapshot captured Jul 20 2026.
Product
Adoption & Engagement
How actively lender teams use the product, from the API audit log: write-actions per day with daily active users, what those actions are (file uploads, overrides, income modeling, package management, settings — endpoint families, never raw paths), and per-customer engagement. Excludes internal admin actions and internal/test tenants; weekends excluded; ranges 7/30/90/180 days. Static snapshot captured Jul 20 2026.
Bank Coverage
Which financial institutions our customers' borrowers actually bank with, and how well we handle each: bank-statement file volume per institution (normalized brand names) over the trailing 180 days, weekly trend of the top banks vs the long tail, per-institution parse-failure and cashflow-check pass rates, and newly-appearing institutions — a roadmap prioritization tool (which bank to fix or integrate next = volume × failure). Institution detection began 2026-05-07. Internal/test tenants excluded. Static snapshot captured Jul 20 2026.
Document Types
What flows into the intake pipeline: documents split from borrower uploads per day (files_file) by classified type — bank statements vs everything else — with docs-per-upload, the unknown/unclassified share as a classifier-quality signal, duplicate detection volume, and a per-type league table (files, packages, avg pages, parse failures, first seen). Type classification began 2026-05-06 and duplicate detection 2026-05-21. Internal/test tenants and weekends excluded; ranges 7/30/90 days. Static snapshot captured Jul 20 2026.
Overrides
Product-health view of manual overrides from overrides_override (laminr-prod): every override is work our product failed to do for the customer — fewer overrides = more joy. Tracks override burden in aggregate (overrides per processed package, inclusion vs exclusion split) and per customer, the zero-touch rate (share of processed packages with no override and no under-review stop — the north star, since 2026-04-13) with a touch-mix breakdown, plus an automation-opportunities panel grouped by theme (keyword-bucketed: payment processors, payroll labeling, transfers, …) listing the customers who need each rule, with inclusion-vs-exclusion direction analysis that flags contradictory signals — consistent themes are global-rule candidates, mixed themes need tenant-scoped rules. Internal/test tenants excluded; weekends excluded; ranges 7/30/90/180 days. Static snapshot captured Jul 20 2026.
Package Journey
Where packages stall between submission and processed: the submitted → under review → processed funnel per submission-day cohort, the share that goes straight through vs. stops in review, and time-to-processed medians for each path plus review dwell time (submission → first review → processed, from package state history). Window starts 2026-04-13 (state history is only reliable from then). Internal/test tenants and weekends excluded; ranges 7/30/90 days. Static snapshot captured Jul 20 2026.
Support
Parsing
Cashflow Check Waterfall
Parser agreement dashboard over waterfall_cashflowcheckresult + waterfall_parser (laminr-prod). The truth parser is per-unit: the last parser that actually ran on that unit (file × account × statement period) — not necessarily the last parser configured in the waterfall. Each earlier parser's attempt is scored against it: MATCH when both pass or when both extract the same cashflow within tolerance (matching numbers on a bad document count as parsing success), MISMATCH when the extracted cashflow differs. Consistent matches per bank = skip-check / early-parser optimization opportunity; consistent mismatches = parser improvement opportunity. Truth-parser failures are user errors (bad documents) — excluded from metrics, shown for volume. Includes waterfall config with fail_on_cashflow flags, bank × parser agreement matrix, and a Today/Yesterday/7/30/90-day range selector. Static snapshot: data since Jul 18 2026, captured Jul 20 2026.
Splittr
How the document splitter performs and whether its splits survive: job volume and strategy mix (gemini / gemini_chunked / infrrd fallback), success rate, processing time and documents per upload from the splittr_telemetry warehouse — correlated against the final page ranges on files in laminr-prod, which reflect support remediation. Every upload is classified exact / merged / fragmented / reshaped, and rolled up to package level — one bad split anywhere in a package pulls in support, so package accuracy is the rate the team actually feels. Failed jobs are listed with bucketed causes. Also tracks throughput (seconds per page and total split time per package), the gemini-vs-infrrd provider comparison (call volume, error rate, latency) with strategy failover paths, and model-call latency. Telemetry starts 2026-07-10; the accuracy comparison starts 2026-07-14 when file ids began populating. Static snapshot captured Jul 20 2026.
Waterfall Operations
Operational health of the parsing waterfall from waterfall_waterfallrun, waterfall_waterfalltaskexecution, and waterfall_filepageparser (laminr-prod). Daily run outcomes (completed / failed / cancelled) with average run duration and stuck-retry counts; per-parser task execution performance (volume, failure rate, avg/p50/p95 duration, retries, skips); pages parsed per day by parser with count/share views (which parser is actually doing the work — cost view); and task errors by type over time with top parser × error hotspots. Rolling 30-day window with Today/Yesterday/7/30-day range selector. Static snapshot captured Jul 20 2026.