The Accuracy Playbook For Data Suppression

Benchmarks, Precision/Recall, And What ‘Good’ Looks Like In The UK

Accurate suppression protects your brand, respects people’s wishes and keeps you on the right side of UK GDPR and PECR. The ICO’s accuracy principle requires you to take reasonable steps to ensure personal data is not incorrect or misleading and to erase or rectify inaccuracies without delay.

For live calls, you must not phone numbers on the TPS or CTPS unless you have specific permission from the subscriber, so you need to screen against these registers before making calls. ICO guidance states you should screen call lists and keep your own suppression list too.

In the UK, the Mailing Preference Service (MPS) is the postal ‘do not mail’ file. Use of the MPS is a condition of the DMA Code and ASA rules, and mailers should screen non-customer lists against it.

Sagacity provides market‑proven suppression and mover data via Connect - our on‑demand DaaS API for single or batch lookups, with a live demo to test suppression against multiple datasets.

What ‘Good’ Suppression Looks Like

Use these acceptance bands to judge results in a trial. They’re conservative, buyer‑friendly targets you can test pre‑purchase.

• Deceased suppression: aim for precision ≥ 98% and recall ≥ 95%. Minimising false positives comes first to avoid suppressing living customers

• Goneaway suppression: aim for precision ≥ 95% and recall ≥ 90%, recognising underlying data is less definitive than death data

• Marketing preference suppression: zero tolerance for contacting TPS/CTPS/MPS matches; evaluate cadence and coverage rather than precision/recall alone

• Deduplication: precision ≥ 99% at the individual level, with human review on borderline matches

These are recommended targets for pilot evaluation, not legal or industry standards.

Precision, Recall And The Confusion Matrix - A Quick Primer

To evaluate suppression accuracy, frame outcomes with a simple confusion matrix:

• True positive (TP): record correctly suppressed

• False positive (FP): record wrongly suppressed

• False negative (FN): record missed and should have been suppressed

• True negative (TN): record correctly kept

Key metrics:

• Precision = TP / (TP + FP): how often a suppression is right

• Recall = TP / (TP + FN): how many true suppressions we found

• F1 = 2PR / (P + R):  single score balancing precision and recall

These definitions are standard in information retrieval and classification.

UK Datasets That Drive Suppression Accuracy

• The Bereavement Register: launched in 2000 and now part of Sagacity. TBR states it is used to screen over 70% of UK direct mail, providing authoritative deceased suppression

• Royal Mail NCOA and data hygiene services: core sources for movers and gone aways that help maintain contact accuracy

• Mailing Preference Service (MPS): the UK postal ‘do not mail’ file that mailers should screen

• Telephone Preference Service (TPS/CTPS): statutory registers you must screen before live calls, with 28 days for registrations to take effect

• Experian documentation also recognises key third‑party sources including The Bereavement Register, Mortascreen and the Goneaway Suppression (GAS) file from Sagacity

A Transparent, Repeatable Trial You Can Run Before You Buy

Here’s a buyer‑side evaluation you can run with us so you can see results in your data, not a slide deck.

1. Define scope

• Channels in scope: mail, phone, email

• Suppression classes: deceased, goneaway, preference (MPS/TPS/CTPS), dedupe

• Risk priorities: agree maximum FP rates by class

2. Build a challenge set

• Stratified sample: by age, recency, region, channel, data source

• Gold labels: create ground truth using confirmed TBR registrations, MPS/TPS checks, returned mail codes and verified customer service outcomes

• Holdout: keep 20% hidden until final scoring

3. Privacy‑preserving matching

• Exchange only what’s necessary for matching - e.g. salted, keyed hashes or tokens to pseudonymise personal identifiers

• Follow ICO guidance on pseudonymisation techniques, including use of modern hashing with salt and appropriate safeguards. Guidance updated 28 March 2025

4. Run the test

• Vendor‑blind: we process the sample with our standard production services and return suppression decisions with reasons and data‑source flags.

• Validation: validate a subset via contact attempts, TBR registrations, MPS/TPS checks and returns

• Freeze versions: record dataset versions and dates for audit

5. Score and diagnose

• Compute precision, recall and F1 by class and segment

• Inspect FP/FN root causes - data quality, matching thresholds, source disagreements

• Tune thresholds to your risk appetite

6. Decide

• Compare to the acceptance bands above

• If we meet or beat targets, move to phased rollout with continuous monitoring

Reporting Templates You Can Lift And Use

Use or adapt these in your RFPs and trials.

1) Confusion Matrix Template

• Precision = TP/(TP+FP)

• Recall = TP/(TP+FN)

• F1 = 2PR/(P+R)

2) Class Benchmarks Template 

3) Source Coverage Template 

MPS and TPS/CTPS roles are documented by their owners and regulators.

How We Reduce False Positives Without Missing Real Issues

• Multi‑source corroboration: we require agreement across independent sources for higher‑risk classes before suppression, with different thresholds per class

• Conservative matching: deterministic keys first, then probabilistic scoring with strict cut‑offs, and manual review on high‑impact cases

• Proven datasets: e.g. The Bereavement Register for deceased, NCOA for movers, MPS for mail and TPS/CTPS for calls

Frequency - How Often Should You Screen?

• Calls: screen against TPS and CTPS before each campaign and at a minimum within the last 28 days because registrations can take up to 28 days to take effect

• Mail: screen prospect lists against MPS before each mailing, as required by the DMA Code

• Deceased and movers: schedule regular hygiene runs and always screen before large campaigns; Royal Mail maintains the NCOA and offers audits if you need a snapshot

Why Sagacity For Suppression Accuracy

• Proven deceased suppression lineage: The Bereavement Register originated with our business and is widely used across the UK

• Live, testable services: Try suppression lookups and forwarding‑address checks in our Connect DaaS environment today

• Respect for people’s choices: Our consumer preference centre makes it simple for individuals to request suppression

• Practical compliance guidance: We design trials and rollouts to align to ICO guidance on accuracy, direct marketing and pseudonymisation

Buyer Toolkit - Copy And Paste Into Your RFP

• Share a 100k stratified sample with pseudonymised identifiers

• Require vendor to return - suppression flag, source(s), match reason, confidence score, timestamp

• Score precision, recall and F1 by class

• Demand an FP/FN audit with proposed rule tweaks and source adds

• Ask for TPS/CTPS screening cadence proof, plus MPS screening for mail

Ready To Test Us?

• Run a suppression accuracy pilot with our Connect API and get a shareable report you can take to governance

• Want to discuss scope or data‑sharing options? Speak to our team using the button below.

We’ll help you design a fair trial, set realistic acceptance bands and implement continuous monitoring so suppression accuracy stays high as your data changes.