A verification waterfall is a sequence of verification checks run in order until identity or risk is resolved, with a cheaper, lighter check first and harder, costlier ones only when needed. It optimizes cost and coverage so you avoid running every vendor on every user.
What is a verification waterfall, in plain English?
A verification waterfall runs your checks in sequence, not all at once. A cheap, low-friction check goes first, and most users pass right there. Only the ones it cannot resolve fall through to the next check, which is usually more thorough, more expensive, or more intrusive. The user cascades down the steps until their identity or risk is settled.
The point is efficiency. Running every vendor and every check on every user is slow and costly, and it piles friction on people who never needed it. A waterfall spends your budget where it is warranted: the easy cases clear quickly and cheaply, and the harder or riskier cases earn the tougher checks. It also gives you failover, so when one provider misses or goes down, the next step still resolves the user.
What you actually tune is not just the individual checks but the sequence itself: which check runs first, what counts as a pass at each step, and how much latency the chain adds. A well-ordered waterfall is cheaper and smoother; a poorly ordered one raises both cost and friction for no gain.
How the waterfall runs
- Cheap first — Run the lightest check. A fast, low-cost, low-friction check clears the majority of users who are easy to verify.
- Fall through — Escalate the unresolved. Users the first check cannot confirm drop to a more thorough or more expensive step.
- Resolved — Pass threshold met. Stop here; no need to run costlier checks on a clear pass.
- Unresolved — Still uncertain. Continue down the waterfall to the next verification step.
- Escalate — Add heavier verification. Harder cases reach document checks, biometrics, or a step-up that costs more but resolves more.
- Resolve — Decide or fail over. The user is approved, sent to review, or declined; if a provider is down, the next step covers it.
Who is in the waterfall?
Who | Their role |
Risk or onboarding owner | Designs the ordering, the pass thresholds, and the escalation logic. |
Verification vendors | Each provides a check that sits at a step, with its own cost, coverage, and speed. |
Data and platform engineering | Wires the sequence, handles fallbacks, and keeps latency under control. |
Finance and analytics | Watch cost per verified user and push for the cheapest ordering that still resolves risk. |
What it looks like in practice
In practice
A fintech onboarding flow starts with a cheap database check that matches name, date of birth, and address. About four in five applicants clear it instantly, at a few cents each, and go straight into the product. Only the ones who fail or come back uncertain fall through.
Those users hit a document and selfie check, which costs more and adds friction but resolves the harder cases. When the primary database vendor has an outage one afternoon, the waterfall fails over to a backup data source at the same step, so onboarding keeps running instead of stalling. The team later notices the second step is being reached too often and reorders the front of the waterfall to catch more good users cheaply, cutting cost per verified customer without hurting fraud capture.
Why it matters to operators
A verification waterfall is how you make identity checking affordable at scale. Verification vendors charge per call and each adds friction, so running them all on everyone burns money and drives away good users. Ordering checks from cheap and light to expensive and thorough lets the easy majority clear fast while you reserve the costly steps for the cases that need them.
Tune it on three things: the ordering, the pass thresholds, and latency. A poorly ordered waterfall raises both cost and friction, for instance by running an expensive check on users a cheap one would have cleared. The trade is thoroughness against speed and spend, so the sequence itself, not just the individual checks, is what you optimize, and failover between steps is what keeps the flow resilient when a provider misses or goes down.
What to watch when you tune it
- Ordering cost. If an expensive step is reached too often, a better first check or a tuned threshold can clear more users cheaply upstream.
- Pass thresholds. Set them too loose and risk leaks through early; too tight and good users fall needlessly to costly steps.
- Latency stacking. Each chained check adds time; watch total onboarding latency, not just the speed of any single vendor.
- Failover gaps. If a step has no backup, a vendor outage stalls the whole flow; make sure each critical step can fail over.
- Coverage overlap. Two steps checking the same thing waste money; each step should add genuinely new resolving power.
Quick questions
Why not run every check on every user?
Because it is slow and expensive and piles friction on people who never needed it. A waterfall clears the easy majority with a cheap first check and reserves costly steps for the cases that actually require them.
What order should the checks go in?
Generally cheapest and lightest first, escalating to more thorough and expensive steps only for unresolved users. The exact order should be tuned so the front of the waterfall clears as many good users as cheaply as possible.
How does a waterfall help with vendor outages?
It gives you failover. If a provider at one step misses or goes down, the next step or a backup source can still resolve the user, so onboarding keeps running instead of stalling on a single point of failure.
What is a pass threshold in a waterfall?
It is the bar a user must clear at a step to be considered resolved and stop cascading. Set it too loose and risk slips through early; too tight and good users are pushed unnecessarily to slower, costlier checks.
How is a waterfall different from orchestration?
A waterfall is a specific sequential pattern for verification checks. Orchestration is the broader capability of configuring and sequencing any risk or onboarding steps, of which a waterfall is one common design.
How do I know my waterfall is well tuned?
Look at cost per verified user, the share of users reaching each step, total latency, and fraud capture. A good waterfall resolves most users cheaply and quickly at the top, with expensive steps used sparingly.
Go deeper
- FFIEC BSA/AML Examination Manual ↗ — The manual US examiners use to assess BSA and AML programs.
- FTC Consumer Advice: Scams ↗ — US consumer guidance on current scams and fraud, and how to report them.