Scoring: Ranking Every Record in CRED Against Your Own Definition of a Good Fit

Tom, CPO at CRED, walks through the launch of custom scoring: building a score from factors and weights, having a score suggested from a list of your existing customers, creating a score by talking to the assistant, applying it to a filtered set of companies, and reading the score and its justification.

Why use Scoring

Custom scoring is something CRED has long built for customers by hand — complex data science work, done with our engineers, then enriched into the platform. Scoring brings that same capability to every CRED user in a self-serve format: you define what predicts a great customer, and CRED scores its database against it.

Use it when you want to:

• Turn your ICP from a gut feel into an explicit, weighted model your whole team scores against

• Run a different score per sponsorship asset — a stadium sponsor target campaign scores differently from a shirt sponsor or a digital media sponsor

• Hand CRED a list of your previous customers for a given asset and have it reverse-engineer the score that predicts lookalike buyers

• See why a company or person scored high or low, and use those reasons as personalisation for outbound

• Rank your target market so your reps work the best-fit accounts first, not the newest ones

• Feed the score into a dynamic list or an automation, so high scorers get worked without anyone checking a dashboard

Where to find it

• In the left sidebar, open Discovery → Scoring. The page lists every custom scoring model in your workspace — ones you've built and ones colleagues have built for you: "Custom scoring models for this workspace. Each score ranks records against the factors and weights you define."

• Use Filters, Search, and Sort: Updated at to find a score, and + Create score (top right) to build a new one.

• The scores table shows Score name, Filters, Entity, Factors, Method, # of applications, Status, CRM sync, Last run and Next run for each model at a glance.

• Open any score to reach its three tabs: Application, Activity and Factors, plus Run History in the top right.

• You can also start from a list: open any list and use the green Score this list banner.

• Or start from AI Assistant in the top bar, which can build and edit scores for you at any stage.

How to set it up

Prerequisites:

• Decide whether you're scoring companies or people. A score's Entity is fixed to one or the other.

• Scoring reads data CRED already holds. Factors with no underlying data can't score anything — the Coverage column tells you where you stand before you commit.

There are two stages to custom scoring: build the score (the factors that predict a good fit), then create an application (the set of records you run it against).

1. Build the score — pick your route

You can get to a score three ways, and they all produce the same editable thing:

2. Set your factors and weights

Open the Factors tab and add one row per variable. Each row takes:

Weighting across all factors must add up to 100 — the Total row at the bottom shows where you are.

💡 Tip: Watch the Coverage column — it tells you how effective a factor will be before you spend anything. A factor at 99% coverage scores almost every record; one at 31% only scores a third of them, and CRED flags the score with a Low data badge. We can't score on factors without data, so either lower that factor's weight or enrich the field first.

3. Have a score suggested from a list

If you'd rather not guess at the factors, open a list of companies or people you already know are good customers and click Score this list → Suggest a score. CRED runs a commonality analysis across the list to work out what those records share, and builds a score from those characteristics.

This is often the fastest route, because working out what your best customers have in common by hand is genuinely hard.

4. Or describe the score to the AI Assistant

Open AI Assistant and describe what you're after — for example, "create me a score based on the similarity between soft drink companies that sponsor sports brands; I want to find lookalikes for my shirt sponsor." The assistant reads CRED's proprietary data on the topic, runs a commonality analysis, and outputs a score you can then edit.

• You can attach screenshots of your current sponsors and have the assistant reverse-engineer a score from those.

• Use it at any stage of the journey — creating a score, editing one, or applying it.

• Your conversation is saved alongside the score, so you can pick it back up later.

5. Preview before you commit

Open Preview to see how a handful of companies or people score against the model as it currently stands, then adjust the factors and weights until the ranking looks right. Click Save changes when you're happy.

💡 Tip: Preview is a beta feature, and its results are an estimate rather than a final score.

6. Create an application

An application is the set of records the score runs against. Open the Application tab, name it, build your filters, and apply them. The table below the filters previews exactly which records are in scope, with a result count at the bottom.

For example, filtering to HQ Continent Europe plus an executive sports interest in soccer narrows a very large universe down to around 30,000 companies.

💡 Tip: Trim your filters deliberately. Scoring is charged on the size of the enrichment, so the tighter your application, the less a run costs.

💡 Tip: A score with no application never runs. If you see the banner "scores have no active scope — they're not running. Open each to create a scope and start it," those scores need an application before anything happens.

7. Choose how it runs

Set the Method to Scheduled or Manual. Scheduled scores take a Frequency — for example Monthly — and CRED then queues every future run for you.

8. Confirm it worked

A run applies your full scoring logic to every record and writes a justification for each, so expect it to take up to two to three hours. Results are blank until it finishes, and you'll get an email when it's ready.

Open the Activity tab to watch progress. The parent row is your application, showing Frequency, Status (Active), Last run, Duration, Next run, # of records, # of change records and Credit spent. Expand it to see every individual run — completed ones carry a date, duration and record count; upcoming ones sit at Scheduled.

A healthy score reads Completed in the Status column back on the Scoring list, with a Last run and a Next run date. Not running means no active application; Last run failed means the run errored and needs re-running.

9. Read the results and put them to work

Scored records give you two paired columns in the grid:

• The score column, holding the score for each company or person

• The justification column — unlock it on a record to read why that record scored the way it did, and which factors moved the needle

Sort the score column descending for a prioritised target list, or ascending to see what's falling short and why. Then:

• Use the score as a condition in a dynamic list, so the list keeps itself populated with your best-fit records

• Use it as a trigger or filter in an automation, so high scorers get actioned automatically

• Use the justifications as personalisation angles for outbound

• Go back to the Factors tab any time to change fields, conditions, thresholds or weights — or just ask AI Assistant to edit the score, run a new application, or re-score against the updated model

Credits

Scoring is charged on the size of the enrichment, so cost scales with how many records your application covers — a tighter filter is a cheaper run. Credits are debited when a run executes, not when you create or edit the score, and every run reports exactly what it used in the Credit spent column on the Activity tab.

Full breakdown: credit consumption table · FAQs (Credits)

Permissions

• Each score has its own Visibility setting — set it to All Members to share the model with the whole workspace, or restrict it so only you see it.

• Scores built by colleagues appear on your Scoring page, so a model your team agrees on only has to be built once.

• Admins manage the workspace credit allocation that scoring runs draw from, so scheduled runs depend on the workspace having credits available.

• Scores only ever rank records you already have access to in CRED — scoring does not widen anyone's data access.

More on roles: CRED Role Overview: Admin & Member

Data privacy

• Scoring reads data CRED already holds on the records in scope and writes back a score and a justification. It does not pull in new personal data on its own.

• Scores, applications and run history stay inside your workspace — nothing is shared with other CRED workspaces.

• You only ever see scores on records you already have permission to view, so a shared score never exposes a record you couldn't otherwise open.

• Where a score uses personal fields such as seniority or job title, that data is processed in line with CRED's GDPR commitments.

Details: GDPR & Data Privacy in CRED

FAQs

Does running a score consume credits?

Building and editing a score is free — including having one suggested from a list or built by AI Assistant. Running it is charged on the size of the enrichment, so the number of records in your application drives the cost. The Credit spent column on the Activity tab shows what every run actually used.

How long does a run take?

Up to two to three hours. CRED applies your full scoring logic to every record in the application and writes a justification for each. Results stay blank until it finishes, and you'll get an email when it's ready.

Can I build a score from a list of my existing customers?

Yes, and it's usually the easiest way. Open the list and click Score this list → Suggest a score — CRED runs a commonality analysis across those records to work out what they share, and builds a score from it.

Why is my score showing Not running?

It has no active application. A score needs a set of filters defining which records to score before it can run — open the score, go to the Application tab, build your filters, and apply them.

Why can't I see a score another team member built?

Scores have a Visibility setting. If it isn't set to All Members, only its creator sees it. Ask them to change Visibility on that score.

What does the Low data warning mean?

At least one of your factors has low Coverage — most records have no value in that field, so the factor can't meaningfully score them. Either reduce that factor's Weighting or enrich the field and re-run.

Can I see why a record scored the way it did?

Yes. Every score column comes paired with a justification column — unlock it on a record to read which factors moved the needle. Those reasons make good personalisation angles for outbound.

Can I check a score before spending credits on a full run?

Use Preview to see how a few records score against the model as it stands, then adjust and Save changes. Preview is a beta feature, so treat its output as an estimate.

Can I change the factors after a score has already run?

Yes. Edit fields, conditions, thresholds or Weighting on the Factors tab at any time — or ask AI Assistant to do it — then re-run to re-score against the updated model. Past runs stay visible in Run History.

Can I score people as well as companies?

Yes. A score's Entity is set to either Company or Person when you create it.

Can I use a score to drive a list or an automation?

Yes — that's the point of it. Use the score as a condition in a dynamic list so the list keeps itself current, or as a trigger or filter in an automation so your best-fit records get worked automatically.

Related articles

• FAQs (Credits)

• CRED Role Overview: Admin & Member

• GDPR & Data Privacy in CRED