How to use popups without annoying the people you need

Use a popup when you have one job on one set of URLs: capture, recover, announce, or ask. Delay it, cap it, and give it a close control. Omnisend's 2025 data put delayed email forms at 2.4% vs 1.9% immediate. Google still treats first-load full-screen promotional overlays as intrusive interstitials. Cookie consent is a different dialog.

Which popup type should you start with?

Match the object to the job.

• Email: one field, a reason, content and collection URLs.
• Cart: shipping math or save-cart, on cart URLs only.
• Announcement: a bar, a real window, sitewide except checkout.
• Form: click-to-open on pricing.
• Survey: after the task.
• Promotion: scheduled sale, then it dies.
• Gamified: only if you can live with the list quality.

Entry popups (fire on load) are the type Google's intrusive interstitial note is written against, when they cover the page. Exit-intent is desktop. Scroll and click are the two triggers that still behave on a phone.

Confirm and error dialogs (delete this, login failed) are browser UX. They are not marketing popups. Do not build them in a popup SaaS.

How should you time and design the first campaign?

A first campaign that is hard to get wrong:

1. Goal: email, one offer, one URL group (e.g. /blog/ and /guides/).
2. Trigger: 6 to 10 seconds or 40% scroll, not both racing.
3. Device: sheet on mobile, modal or slide-in on desktop.
4. Fields: email. Stop. Omnisend saw conversion drop at four fields (2.1% to 1.5%) across 1.24 billion displays in 2025.
5. Close: visible X, Escape, tap-outside. No guilt copy.
6. Cap: once per session, 7 days after close, never for existing subscribers.
7. Off: checkout, account, cart unless this is a cart campaign.

Copy: what they get, when it ends, the button label. "Get the 10% code" beats "Submit."

Sleeknote's 2026 set of 516 million views put average conversion at 4.23%, mobile at 5.60%. Late, small, on-brand sheets can work on phones. First-load walls do not.

How do you keep this from wrecking UX and search?

Read do popups affect user experience for the longer version. The short list:

• Do not cover the article on a mobile search landing page.
• Do not inject a bar into document flow after load (CLS ≤ 0.1).
• Do not add a second popup vendor if INP is already near 200ms.
• Do not stack a sale modal on a cookie banner. ICO PECR is about cookies. The form is about email.

How do you A/B test without fooling yourself?

One change per test: headline, or trigger, or offer. Not all three.

Sample: Omnisend's average is 2.1%. If you have 10,000 unique overlay views a week, you can see a 2.1% vs 2.6% move. If you have 400, you cannot. Run through a weekend. Watch bounce and revenue, not only signup rate.

OptinMonster is best if testing is why you bought a tool. Poptin includes tests on a cheap tier. Popupsmart is enough to ship the default above if you will not test this month.

Legal: GDPR Article 7 for consent that is specific. US SMS: 47 CFR 64.1200 if you collect a number. None of that is optional copy.

If you only needed a bar, use Hello Bar and skip a modal. If you are on Shopify, start with the Shopify how-to. If you are on WordPress, start with the WordPress roundup.