There are various reasons that can prevent a web push notification from being delivered. One of them is the presence of subscriptions that are no longer in use: if a device / browser is no longer used, then its subscription to web push notifications is still valid, but the notification won't be delivered. That legacy subscription will eventually expire and will be removed automatically (by Pushpad), but this can take several months, depending on the browser push service (e.g. Google FCM, Mozilla autopush, etc.).
For the above reason, if you send a web push notification only to the browsers that have recently visited your website, then you will probably get higher delivery rates, because you avoid to send the notification to old, unused subscriptions. Some experiments described on other blogs confirm that.
Note that this strategy won't increase the total number of deliveries! It will just increase the delivery rate (i.e. the percentage). Basically it allows you to target only the subscription that are more likely to receive the notification successfully.
It is simple to use this strategy with Pushpad. When a user visits your website you can add a tag to that browser using the Javascript SDK:
var currentMonth = new Date().toJSON().slice(0,7); // e.g. '2018-09'
pushpad('subscribe', function () {}, {tags: [currentMonth]}); // add tag to browser subscription
Then when you send a web push notification you can target only people who have visited your website in the last months. For example, go to the Pushpad dashboard, create a notification, select Custom audience and enter a tag expression like this (please adapt it to your current period):
2018-09 || 2018-08 || 2018-07
This will target only people who have visited your website in the last three months.
If you prefer you can also remove the inactive subscriptions to reduce cost (if you pay per subscriber). The strategy of adding the tag is similar and then you can scan your subscriptions with the REST API and delete subscribers that have not visited your website recently.