Ultiwrite connects portfolio acquisition, underwriting and merchant retention intelligence, so what is established about a merchant at one stage is available as context at the next. Acquisition intelligence is the front door; underwriting and retention are the adjacent decisions the same reasoning engine serves.
Read merchants individually and together, so what a book contains is visible rather than averaged.
The merchant among many othersRead the evidence a merchant arrives with and whether it describes one coherent business.
The merchant at the point of decisionRead what arrives afterwards against the picture already established, and separate ordinary variation from consequential change.
The merchant across its own historyA residual book arrives as a stack of monthly schedules and a trailing average. Ultiwrite reads the whole series as one system, treats every supplied file as evidence rather than truth, and hands back a decision, not a dashboard: what the recurring cash flow is actually worth, whether the seller’s history is durable, which cohorts belong in the forward book, and what should move the price.
The latest month, less residual from merchants without a track record, less one month of observed dollar attrition on the share of income that moves with card volume. Reported beside the seller’s trailing three, six and twelve months.
Merchant IDs collapsed to probable relationships from the names they trade under, then verticals. Top-three share and single-vertical share are stated with the HHI beside them, because the HHI alone read a 63.5% book as unconcentrated.
Income classified as contractual, volume-contingent, event-driven or one-time. A 10% fall in card volume is translated into dollars a month, and unclassified columns are reported as uncertain rather than dropped.
Correlated boarding and termination clusters, graded by how many independent reasons support them, with the dispute tail that follows them into a buyer’s first year.
Survival of the opening cohort across the series, and each seasoned merchant’s recent volume against its own baseline, with a guard that refuses a percentage on a near-zero base.
Two files claiming the same month. A merchant that processed in the last month it was listed and vanished from every later file. Columns that fail to map. Each reported with the question to put to the seller.
A transient batch inside the trailing months, a structural break that contaminates every figure computed across it, a cohort that behaves as its own book. Each is separated from the forward run rate, and the metrics it contaminates are marked for recomputation.
Residual that depends on a pricing tier, a split, or an assignment the buyer does not control, read against peers so a merchant that is merely moving with its category is not priced as a decliner.
The agreement, the month or the counterparty that would confirm or overturn each finding, ranked by expected effect on the price rather than by custom. One missing agreement can decide whether a large share of the income deserves a full multiple.
The acquisition workspace, the v1 API, or a written engagement memo. Each brief is frozen at run time beside the files it read.
Request accessThe coherence engine reads whether a merchant’s identity, entity, presence, documents and behavior tell the same story. For accounts found in diligence, for new merchants after close, and for any single merchant a buyer wants read on its own.
Timing, card mix, category behavior, revenue flow and structure fold into one deviation score with a confidence interval. One anomaly is noise; a cluster that corroborates is a pattern.
Statements and business documents read for internal consistency, the geometry of a numeric column, the timeline of a whole statement set, and the structural fingerprints of files that were altered or recycled.
The category a merchant declared against the one its descriptors and ticket sizes imply, whether the mismatch moved economics in the merchant’s favor, and the drift across a whole portfolio against a stated baseline rate.
Legal name, designator, formation date, identifiers and registries checked against each other before they are checked against anything else, with entity resolution across names, domains, addresses and principals.
Regulatory entitlement, footprint maturity, capability plausibility and dispute velocity, from public evidence the caller supplies. Absence is never treated as evidence.
Merchant Risk X-Ray as a standalone report, the underwriting workspace, and the v1 API.
Request accessAfter the book is bought or the merchant is boarded, the same engine watches it. Retention intelligence reads each relationship as it behaves, not as it was declared; portfolio intelligence compares every import to the baseline you underwrote and surfaces the drift, the concentration and the shared risk between accounts, while intervention still changes the outcome.
Information acquisition, competitive shopping, service dissatisfaction, processing divergence, business closure, and incumbent-induced friction. A single weak signal is noise; a sequence that corroborates across analyzers is a transition.
Every merchant is compared to itself, with a seasonal-consistency check so a December that looks like last December is not an alarm.
How attractive the account looks from outside: volume, tenure, economics and pricing staleness, read without any behavioral event so it stands as independent evidence.
How long since the relationship was touched, and whether the friction the merchant is feeling is being caused by the processor it is about to leave.
Every assessment can be replayed as of any past date, and promotion out of shadow mode requires measured lift on a partner’s own history rather than a threshold search.
Import the book by CSV, XLSX or a gateway connector and every merchant is compared to the baseline you underwrote. Drift candidates are ranked into a review queue, with the specific change behind each one.
Per-period dimensions fold into one trajectory per merchant, read for its slope and its acceleration, so an account bending the wrong way surfaces before the loss rather than in the post-mortem.
Each merchant tracked against the thresholds the card-brand monitoring programs watch, surfaced as a watch-list for your team. Ultiwrite never asserts a merchant is in breach.
Merchants that share a principal, an agent or a behavioral fingerprint, read on tokenized identifiers so raw owner data never enters the layer. Loss concentrates in the connections, not inside any one account.
Event ingestion over REST or batch, file imports and gateway connectors, the retention and portfolio workspaces, and the v1 API. Enterprise tenants add webhooks.
Retention intelligence is described here as what it reads, not as a proven outcome: nothing in it has been validated against a partner’s history yet. Any signal can be marked shadow-mode individually, which carries the caveat through to the reader.
Request accessConventional systems treat underwriting, retention and acquisition diligence as separate workflows. Ultiwrite treats them as different questions put to one evolving body of merchant evidence.
The evidence the merchant arrived with.
Everything observed since, read against that first picture rather than on its own.
And that trajectory set beside every other merchant in the book.
A fact established while the merchant was being evaluated is still the thing later evidence has to agree with, months after the decision was made.
A material change identified after boarding can alter how that merchant is interpreted when a portfolio containing it is examined.
A pattern that is only visible across many merchants can put a single merchant in a different light than its own file supports.
Not three independent analyses. Progressively richer context, applied to different decisions.
Does the available evidence describe a coherent operating business?
Has something material changed from the business that was originally evaluated?
How does this merchant’s trajectory affect the quality and outlook of the book around it?
Same merchant. Different decision. More context.
The rules every lens shares. The engine asks whether the data agrees with itself, whether the unit of analysis is the right one, whether several findings share a root cause, what rests on one unstable assumption, what would overturn a finding, and what changes economically if it is wrong. These four disciplines are what make the answer safe to paste into a model, forward to a lender, or put to a seller.
Observed, derived, inferred, contradicted, uncertain. Every finding carries one, so the caveat travels with the number.
High, medium or low, each with its reason. No decimal confidence until a calibration corpus exists to earn one.
Every finding names the request or the evidence that would overturn it. A conclusion that cannot be wrong is not a conclusion.
No fraud, no intent, no ownership findings. The engine reports what the files show and what would resolve it; a person decides what it means.
| The question | Checklist | Spreadsheet | Ultiwrite |
|---|---|---|---|
| What does the trailing average contain? | The average | Seasoned and unseasoned residual, decayed on the volume-contingent share | |
| How many relationships sit under the merchant count? | The count | Probable relationships, top-three share, single-vertical share | |
| Which merchants were boarded and dropped together? | Waves, graded by independent reasons, with the dispute tail | ||
| Do the files agree with each other? | Contradictions, each with the question to ask | ||
| How far can each figure be trusted? | Pass or fail | A standing on every finding | |
| What should it do to the price? | A verdict, a run rate, and the structure to protect it |
Run a brief, keep every past one, record what you did with each finding. Tenant-scoped from the first byte.
The same engine over the v1 API: analyze a residual book, list briefs, fetch one. Webhooks on enterprise tenants.
For a single book on a deadline, the team delivers a written brief a buyer can forward to a lender and a lawyer.
Every file is retained before it is read, under your own tenant, with a receipt on every upload.
Bring a merchant file, a portfolio or a diligence question, and read the resulting intelligence in context.