Push notifications for ecommerce: when to use them and how to write them
Web push is cheap, instant and the fastest way to be switched off permanently. It is an interruption rather than a message, and almost everything that works in email fails here.
Web push has an obvious appeal: no per-message cost, no deliverability to manage, no inbox to compete in. It also has an unusual property — when somebody has had enough, they do not unsubscribe from a campaign, they block notifications from your site at the browser level, permanently, and you will never reach them this way again.
That asymmetry should shape everything about how you use it.
How to ask for the permission
The browser prompt is a one-shot resource. A refusal is usually permanent for that site, and there is no second attempt.
- Never prompt on the first page view. A prompt before anybody knows what your site is gets denied by reflex.
- Use a soft ask first — your own in-page prompt explaining what they will get and how often. Only trigger the real browser prompt after they agree to that.
- Ask at a moment with obvious value: on a sold-out product, at an order confirmation, after somebody has viewed several pages.
- Say what and how often, and keep it. "Back-in-stock alerts, and nothing else" is a promise worth being able to keep.
What it is actually good for
- Back in stock, for a product this person asked about. The best push message there is.
- Price drop on something they viewed or saved.
- Order and delivery updates, where you have no better channel.
- A genuine deadline, used sparingly, and only when it is real.
What fails: weekly campaigns, category announcements, anything that could equally have been an email, and anything sent because the calendar said so.
Writing for an interruption
- The title is about 40 characters and the body about 90 before truncation, varying by platform. Write to the shorter end.
- Lead with the specific thing, not the brand. The sender is already shown.
- One action. A notification with an ambiguous outcome is a notification that gets dismissed.
- Deep link to the exact page, never the home page. Landing somewhere generic after tapping is the single most common reason people switch push off.
- Respect local time. A push at 02:00 is not a lost impression, it is a lost subscriber.
Frequency is the whole risk
Email has a graduated response: people ignore, then unsubscribe. Push has a cliff. The practical limits most shops land on after learning the hard way:
- At most two or three a week, and fewer is usually better.
- Never two in a day unless one is transactional.
- Cap across channels. Somebody receiving an email, an SMS and a push about the same sale has been contacted three times about one thing.
- Track the opt-out and the block rate, not just the click rate. A rising block rate is permanent attrition and it is the number that should govern the calendar.
Consent, and the part people forget
The browser permission is a technical grant, not by itself a complete legal basis for marketing. In the EU the same principles apply as to any other electronic marketing: the consent should be specific and informed, which is what the soft ask is genuinely for.
- Record that consent was given, when, and what was promised — as for every other channel.
- Offer a way to stop that does not require browser settings, in your own preference centre, and honour it.
- Do not treat a push subscriber as an email subscriber. It is a separate permission and a separate record.
Sources and further reading (4)
- MDN — Push API
- MDN — Notifications API best practices
- ePrivacy Directive 2002/58/EC — Article 13
- EDPB — Guidelines 05/2020 on consent
Checked on 23 September 2026. Provider prices, mailbox rules and legal guidance change — verify anything you plan to act on.
A frequency limit that counts all of them together
Auralata records consent per channel and can cap how many marketing messages one person receives in a week across email, SMS and messaging — so a single sale does not reach somebody three times.