The platform, release by release.

Re-processing an application is not new. Changing the policy template it runs against is.

Switch to any policy template and re-run, and that is the policy that runs, all the way through execution. The documents stay on the application, already processed. Nothing is uploaded again and nothing is read again, so you never burn credits twice on the same file. A second run is about the policy, not about the paperwork.

What that lets you do

• Run an application in two steps. A light policy template first for a quick pre-approval: a cashflow calculation, a few basic gates, an answer in minutes. Then the full one for the ones that pass, with every document and the credit investigation behind it. The applicant submits once, and you spend full processing only on the files that earned it.
• Test a rule change against real applications. Move a threshold, re-run applications you have already processed against the new version, and see what the change does before it meets live volume.
• Put an application through a different loan product. If another policy template fits it better, run it against that one instead of starting the file again.

The applications list, redesigned

Submitted Applications has been rebuilt around what you actually scan for. Every run sits on the application it belongs to, showing the policy template it ran against, whether it completed, the decision it reached and the score behind it.

An application run twice shows both runs together, one above the other, with both decisions and both scores in view. You can see what a different policy template changed without opening anything. Run status filters across the whole list, and search is one keystroke away.

Bank statements as images

Bank statement processing now accepts images alongside the document formats it already handled. A JPG photographed on a phone or captured from a banking app goes through the same analysis as a clean PDF, including transactions, balances, risk signals and monthly metrics.

When an image is split into separate statements, each one keeps its link back to the file it came from, so every number still traces to the document a person actually handed you.

One bad page no longer costs you the case

A bank statement that processed in parts used to be all or nothing. Now a statement where most sections come through cleanly and a few recoverable failures remain completes with warnings instead of failing the run.

The list shows that state for what it is: completed with warnings, not failed. And decisioning still runs on it. A single unreadable page no longer sends the whole application to somebody's desk.

See exactly what ran

Every application now has a Status tab. It lays out the run as it actually executed: each step in the flow, whether it completed, and how many finished, alongside the run's decision state and when it last ran.

If an application has been run more than once, pick the run from the dropdown and the whole view switches to that execution. This is where a case that completed with warnings tells you which step warned, rather than leaving you to work it out.

Re-process sits in the same header, so the next run starts from the page where you just finished checking the last one.

Export what the analysis produced

Processor output can now be downloaded straight from the application results screen. Financials Aggregator is first: pull the aggregated financials out as a file and drop them into a credit memo, a committee pack or your own model, without retyping anything.

Monthly bank metrics, measured over the real period

Average Monthly Credits and Average Monthly Debits are now calculated using the statement's actual coverage period. A statement covering six weeks is averaged over six weeks rather than rounded into whole months, so the monthly figures match the period the statement really covers.

Both metrics also have clearer names, tooltips explaining what feeds them, and number formatting restored. If you have been reading these figures on existing applications, re-process to see them recalculated.

Improvements and fixes

• Cleaner run summaries when an application has several executions behind it.
• More consistent decisioning signals across re-runs, including custom signal catalogues.

This release is about how you build. Creating a processor has been redesigned to be more intuitive, and authoring now happens inside the workflows where processors actually get used.

A rebuilt creation flow

Processor creation has been redesigned around the way people actually work. Authoring and workflow configuration are now one thing rather than two separate places you move between, and you can create a processor directly from the workflow that needs it. Build the step where you use it, then carry on.

Control over how each step runs

Workflows now support configurable Data Processing, and configurable processors support processing modes. How a step behaves is something you set, not a fixed default you design around. The configuration you choose is the configuration that runs on every application that passes through it.

Extractors follow the same rules

Extractor authoring and display now line up with Data Processing, including the same display and processing mode controls. One mental model across both, so what you learn in one place holds in the other.

Improvements and fixes

• Better handling of oversized transaction chunks on large bank statements, with configurable chunk size.
• Faster, more reliable processing across the board.

Two of the most requested features are now live in every account, along with a redesigned Decision scorecard and tighter policy governance.

Loan repayments and amortisation, visible at last

Floowed now detects existing loans straight from bank statements. It picks up loan proceeds, identifies repayment streams, and maps out the cadence: typical repayment amounts, how regularly they hit, and any failed attempts along the way. Everything is laid out on a monthly timeline, and repayments are linked back to the proceeds they belong to.

The result: you can see how an applicant is servicing their existing debt before you decide to add to it. No more reconstructing amortisation behavior by hand from raw statement lines.

Bounced cheques and risk signals, in full detail

Bank risk analysis just got a lot sharper. NSF and negative-balance episodes are now tracked end to end: how long the account stayed negative, how deep it went, and whether (and when) it genuinely recovered. Bounced payments are properly separated from reversed incoming receipts, related cheque events are grouped together, and every event drills down to the underlying transactions.

Risk events are also ordered correctly across statements, even when dates overlap or statements arrive out of sequence. The story the timeline tells you is the story that actually happened.

A redesigned Decision scorecard

The scorecard got a full redesign: a compact decision summary, hard-rule results, and score metrics in one view, topped with a grade-band score gauge. Every row expands to show the detail behind the number, so you can go from headline decision to underlying evidence in one click.

Tighter policy governance

Applications processed on an older template version are now flagged with a clear warning, and you can re-evaluate them against the current version in one action. Template selection is also limited to versions with an active policy, so an application can't accidentally run against a configuration that shouldn't decision. The policy your team approved is the version that executes, and now it's easier to prove it.

Improvements and fixes

• Workflow canvas stability: fixes for zoom, layout, and resizing behavior.
• Workflow-node configuration: direct edit access restored, and custom integration settings now save at the individual node level.
• Processor Schema Builder: longer-running processor tests are supported, with clearer errors when a test does fail.
• Faster processing and more reliable analysis across the board.

One of our biggest statement-review updates yet. This release is all about helping you spot what needs attention faster, and trust the numbers even on large, messy, or non-standard statements.

Clearer review status at a glance

• New colour-coded status card on each statement thumbnail, so you can see extraction status and any manual-review items instantly.
• Issue counts now appear on each thumbnail, so you can tell which statements need attention before opening them.
• New "Jump to issue" button in side-by-side review takes you straight to the transactions that need a manual review.
• Thumbnail status and the side-by-side screen now stay in sync, for a consistent review status across the platform.

Fewer unnecessary manual reviews

Recovered rows and normal balance history no longer trigger a manual review when no action is actually needed.

Better handling of large and complex statements

• Upload large files without timeouts.
• Improved page coverage, reducing the risk of pages being silently missed, with clearer flagging when a page may have failed or been dropped.
• Better handling of long statements and chunk-boundary pages.
• More reliable handling of statements with ambiguous or missing page numbers.

More accurate figures

• Improved balance verification, with a clearer distinction between the original expected balance and later recovered balances.
• Better debit and credit detection on statements that use a single combined amount column.
• Debit and credit direction is now inferred from opening and running-balance movement, with automatic correction where the balance clearly supports it. Fewer cheque payments mis-classified as credits.
• Genuinely uncertain transactions are still preserved for manual review rather than force-corrected on unclear evidence.