Changing a risk factor means asking someone else
A new risk appreciation, a new product, a new market, a new segment. The model cannot describe it until the vendor has time.
Product · Onboarding
Customer Risk Assessment & KYC
Collect CDD data with your own KYC forms, score it against a risk model you build, and re-score the customer as they change. Every rating shows the factors behind it.
Configurable
FACTORS AND WEIGHTS YOU DEFINE
Simulated
TEST A CHANGE BEFORE YOU PUBLISH
Explainable
EVERY FACTOR & Detail SHOWN
Refreshed
PERIODIC AND EVENT-DRIVEN, update Risk Score on any change
Why this matters
The Risk model in your system/Excel is not the model that you want to use.
Risk models usually are configured once, at implementation, by someone else. When the business changes, adding a factor means a ticket, a quote, development and a release cycle. So teams stop asking. They mark segments high risk by hand, with their own way, keep an adjustment sheet on the side, and the real model quietly moves out of the system. Complead puts the model where the policy owner sits: change a factor, see what it does to your portfolio before you publish, and keep every version attached to the ratings it produced.
Real model moves to a spreadsheet
Manual overrides and side adjustments become standard practice. What the system holds is no longer what the team does.
Nobody can prove or explain which model applied
A rating from 5 years ago does not carry its weights, its thresholds or its approver. The change history is in email.
How it works
Scroll to advance
Four steps, from your policy to a defensible rating
-
Step 01
Configure the model
Set your factors, weights and thresholds, or start from a sector template. Conditional rules handle the cases a flat weighting cannot: a product that only matters in certain markets, a channel that changes the tier above a threshold.
-
Step 02
Collect what your risk model needs
Build your own KYC forms that ask only for what your model actually scores. Responses land on the customer record alongside screening results, geography and product data already held, so nothing is keyed twice.
-
Step 03
Rate, with the reasoning attached - AI also helps
A score and a tier come back with every contributing factor, its weight and its source. The model version that produced the rating is stamped on it, so the file stays defensible after the model has moved on.
-
Step 04
Change it without gambling
Two things change: the customer and the model. Behaviour, exposure and screening results trigger a re-score automatically, on event or on schedule. And when the model itself needs to move, simulate the change against your live portfolio first, see how many customers shift tier, then publish it as a new version.
complead / verify
- documentverified
- livenesspassed
- identifiersmatched
- geography30
- product25
- channel15
- PEP status20
- delivery10
complead / score
- source of funds → funding risklow
- ownership country → geographymed
- intended volume → productlow
- annual turnover volume → 5M$high*
complead / rate
- factorslisted
- EDD nottriggered
- geography +24
- product +15
- PEP none 0
- TOTAL *62*MEDIUM
- model v4.2
complead / refresh
- PEP tier 2 to1
- rating MEDIUM toHIGH
- reviewtriggered
CONFIGURE
COLLECT
Rate
ADAPT
Capabilities
What the Customer Risk Rating & KYC Product actually does
Start from a model, not a blank page
Sector templates come pre-weighted from consortium data, so a working model runs on day one. From there, change any factor, weight or threshold, add conditional rules, and run separate models per segment, entity type or jurisdiction.
KYC forms that feed the score
Build the forms that collect what your model needs, with questions that branch on earlier answers. Responses land on the customer record as scored factors, not as attachments someone reads later.
Simulate before you publish
Change a weight and see the effect on your live portfolio before it applies: how many customers move tier, how much EDD volume you just created. Publish as a new version when the risks are acceptable.
Benchmark your weights against peers
Compare your factor weightings and tier distribution against anonymised data from comparable institutions. The first outside reference for a model that was, until now, only ever checked against itself.
AI that proposes, you approve
The model watches its own output. When a factor stops separating risk, when a segment drifts, or when your distribution moves away from peers, it proposes a specific change with the evidence behind it. Every proposal goes through simulation and your approval before it becomes a version.
Re-scored on event and on schedule
A screening hit, a change in payment behaviour, a new market, an updated form response. Any of them recalculates the rating. Periodic review still runs by tier, but it is no longer the only thing that moves a score.
Product tour
The screens your vendor usually keeps
Model configuration, portfolio simulation and rating detail are normally on the vendor's side of the account. Here they are on yours.
The model is yours to change
Factors, weights and thresholds in one screen. Change one and the tier boundaries move with it.
- Factor list with weights, one being edited
- Conditional rule shown inline
- Tier thresholds with live customer counts beside them
- Version label and last edited by
complead / onboarding
- Retail modelv4.2
- GEOGRAPHY30
- PRODUCT25
- PEP STATUS20
- CHANNEL15
- SOURCE OF FUNDS10
- LOW 0-398,204
- MEDIUM 40-693,240
- HIGH 70+412
See the change before it applies
Nothing publishes until you know how many customers move tier and how much enhanced diligence you just created.
- The proposed change, stated in one line
- Customers moving tier, in both directions
- Impact on EDD volume
- Publish as a new version, or discard
complead / rating-detail
- GEOGRAPHY30 → 35
- HIGH 412 → 561+149
- MEDIUM 3,240 → 3,127-113
- LOW 8,204 → 8,168-36
- EDD QUEUE+149
Every factor, its weight, its source
The reasoning behind a rating, with the model version that produced it. Six months later, the file still answers the question.
- Score broken down by contributing factor
- Where each value came from
- Model version stamped on the rating
- Overrides recorded with their reason
- M. AydinMEDIUM 54
- GEOGRAPHY, KYC FORM+24
- PRODUCT, ACCOUNT DATA+15
- CHANNEL, ONBOARDING+15
- PEP STATUS, SCREENING0
- MODELv4.2
- RATED12 MAR 2026
Every factor, its weight, its source
The reasoning behind a rating, with the model version that produced it. Six months later, the file still answers the question.
- Score broken down by contributing factor
- Where each value came from
- Model version stamped on the rating
- Overrides recorded with their reason
complead / model
- M. AydinMEDIUM 54
- GEOGRAPHY, KYC FORM+24
- PRODUCT, ACCOUNT DATA+15
- CHANNEL, ONBOARDING+15
- PEP STATUS, SCREENING0
- MODELv4.2
Runs on Fusion AI
Fusion is what makes the rating a control, not a label
Every module writes to the same customer record on Fusion AI, and reads back from it. Screening, behaviour, ownership and media arrive as scored factors in your model. The tier that comes out goes back to those same modules as the setting they run on. Rate a customer high and the platform tightens around them, with no rule to duplicate and nothing to keep in sync.
This product
Customer Screening & Ongoing Monitoring
Screeening & Ongoing Monitoring against to Sanctions and PEPs are feeding your Customer Scoring
Feeds into
Adverse Media
If a customer showing in media with bad news, you should affect the risk rating for this customer.
You get
Transaction & Fraud Monitoring
If a customer trying a suspicios payment or money movement, block the transaction as real-time and automatically add a risk score to customer with Fusion AI
Run your model on your own portfolio
Bring your current risk model and a sample of customers. In 30 minutes you will see the tier distribution it produces, how it compares to peers, and what a change would do before you make it.
500+
Ready Customer Scoring Template
Flexible & Dynamic
Build Customer Scoring with your own data modal
AI Drives
AI provides insights that you haven't noticed
Integration
One API call to score, a webhook for everything after
Send the customer, get a rating and its factors back. Subscribe to the events that change it: a screening hit, a behaviour shift, a form update, a new model version.
// score at onboarding
POST /v1/rating "customer": {...}, "model": "retail_v4"
200 OK · rating MEDIUM · score 54 · model v4.2
200 OK · rating MEDIUM · score 54 · model v4.2
// subscribe to rating changes
PUT /v1/rating/webhooks "events": ["rating.changed","edd.triggered","model.published"], "url": "https://you.example/rating-events"
FAQ
Questions risk owners ask us
What is included in this product and what is not?
The CDD data is obtained from your own KYC forms, then evaluated using a risk model that you set up, and the customer is re-scored whenever there are changes. Document and biometric identity verification operates separately and is handled by your IDV provider; the verified identity is provided to the model as an input, not something that we generate.
Can we use the model that we already have running?
Yes, and that is the typical starting point: the factors, weights and thresholds are input exactly as they appear in your policy, together with any conditional rules that a fixed weighting cannot cover. If you prefer not to begin with your own sheet, sector templates are available.
After we go live, who will be altering the model?
Your policy owner, in the product. Adding a factor or moving a weight does not need a ticket, a quote or a release, which is the reason models drift into spreadsheets in the first place.
Can we see what a change does before publishing it?
Yes. Simulation runs the proposed model against your live portfolio and reports how many customers move tier in each direction and how much enhanced diligence the change creates. Publish as a new version, or discard.
Is it possible to use different models depending on the segment, entity type, or jurisdiction?
Yes, separate models with their own factors and tier boundaries can be used for the retail, corporate and high risk portfolios, meaning that one set of weights is not applied to groups which do not behave similarly.
What becomes of the ratings that a previous version of the model produced?
Every rating is stamped with the model version, weights and thresholds in force when it was produced. A file from three years ago still answers why that customer was rated the way they were, without reconstructing a model that has since moved on.
What causes a rating to change following onboarding?
A flag that is a screening hit, a change in payment behaviour, the introduction of a new market or product, a revised form response, or the release of a new model version. Periodic review by tier still takes place, but it is not the only factor that causes a score to change.
What triggers enhanced due diligence and how is it tracked?
As soon as the score reaches the level you have specified, EDD is triggered automatically and the case includes the factors that caused it to reach that level. Overrides can be granted and a reason for them is recorded, so that an approved exception is shown as a decision and not as a gap.
Will the AI alter our model by itself?
On the contrary, it monitors the model's output and suggests a specific change together with the evidence for it, such as in the case when a factor ceases to separate risk. Each proposal is subjected to simulation and requires your approval before it becomes a version.
Is our data put to use in the peer benchmark?
The benchmarking process involves comparing your weightings and the way your tiers are distributed with the combined data from similar institutions, and joining takes place on a contractual basis rather than being the default option. Information regarding residency, retention and the handling of customer data can be found in the Trust Center.
Testimonials
All case studies
What compliance teams say
We went from checking customers one by one to a single platform that screens more than 3,000 of them around the clock, so our analysts can finally focus on real risk.
We moved from screening customers one by one to a unified platform where our analysts can focus on what actually matters.
We focus on real risks, not false positives, meeting our AML obligations and our customers' expectations.