Work  /  Push Notification AB Test

CRM & Push  ·  Spain · UK · US  ·  2026

+50% more applications. Zero extra spend. Just better copy.

Building a push notification copy testing system from the ground up — and the infrastructure to keep improving it.

AB testing CRM & push Copywriting Data analysis Product marketing

+50%

Applications per delivered notification · Spain

+37%

Click rate lift (users)

10+

Notification types audited

15 days

50/50 test · 3 markets

Job Today's job alert notifications — sent every time a new role matches a job seeker's saved search — were written from the employer's perspective: "Company X just posted a new job." The copy announced a posting. It didn't speak to the job seeker.

No copy testing had ever been run on these notifications. There was no process, no tooling, and no baseline hypothesis. With millions of notifications delivered weekly in Spain alone, even a small improvement in conversion represented significant additional applications — at zero extra cost.

Control

⚡ [Company] just posted a new job! ⚡

[Job role] – [Job description excerpt]

Test

[Company] is looking for a [Role]

Your profile gets you halfway there. Apply with a tap!

The hypothesis

The current copy is company-oriented — it announces a posting. The test reframes it around the job seeker: their profile is already a match, they just need to tap. Job-seeker language should drive more clicks and more applications.

Marketing Manager — initiative owner end to end. I worked with engineering (who built a reusable testing tool) and a BI analyst (data extraction), but the strategy, creative, metrics definition, and analysis were entirely mine.

Initiative owner Copy & hypothesis Metrics definition Data analysis Cross-functional lead

The test ran for 15 days across Spain, UK, and US on a 50/50 split. Spain — the most mature market with the highest volume — gave the clearest signal.

Click rate % — Spain (users)3.6%5.0%6.4%7.8%20/0522/0524/0526/0528/0530/0501/06
Apply rate per delivered notification % — Spain0.5%0.7%0.9%1.1%20/0522/0524/0526/0528/0530/0501/06
Test Control

Spain

SHIP IT ✓

+50% apply/delivered

UK

SHIPPED ✓

+10% apply/delivered

US

SHIPPED ✓

Low volume · NYC only

The apply rate per click was slightly lower in test — a broader audience was now engaging. But the metric that matters, applications per delivered notification, was up 50% in Spain. Same send volume, zero extra spend, significantly more applications.

Beyond the immediate result, this project established a repeatable testing capability that didn't exist before. It applied inbound marketing principles — meeting users with the right content at the right moment in their journey — to a CRM channel that had never been optimised for it. Engineering built a self-serve tool that allows new tests to launch in under a day. With 10+ notification types still untested, the pipeline for lifecycle conversion optimisation is now open.