You want a record of how a page changes: a supplier's terms, a competitor's price, your own privacy policy. Capturing it once is easy. Deciding how often to capture it again is the harder question, and the answer is never "as often as possible", because every capture costs storage and review time. It is also never "once a month" if the version you need lasted three days.
This article gives you a way to choose an interval, starting points for the common cases, and a way to correct your choice after the first month.
What a series of captures can and cannot show
A capture is a sample. It shows the page at the moment the browser loaded it, and nothing about the time between one capture and the next.
Two consecutive captures that differ prove that the page changed at some point between them. They do not say when. If the page changed and was changed back between two captures, neither one shows it. So the interval you choose is the longest gap in your record, and the longest gap is the one an opponent will point to.
Pages change without notice. In Pew Research Center's 2024 study, 38% of pages that existed in 2013 were no longer available, and 8% of pages from 2023 were already gone. Content that stays online drifts too: in a study of more than a million references in scholarly articles, over 75% had drifted away from what they were when referenced. Those figures measure years, not days, so they say nothing about how often any one page changes. They do show that "it will still say the same thing" is a bad default. We go through the numbers in Link rot by the numbers.
Three questions that set the interval
1. What does a missed version cost you? If a missing version means an awkward conversation, weekly is enough. If it means you cannot answer a regulator, a court or a client, capture more often than you think you need to.
2. How long does a version usually last? A terms page may stay the same for a year. A promotional banner may last a weekend. A price on a busy retail site may change several times a day. Your interval should be shorter than the shortest-lived version you care about. If you do not know yet, guess short and adjust (see below).
3. Does a rule set a window? Sometimes the law defines what you must be able to show. Under Article 6a of the EU Price Indication Directive, the prior price for a discount is the lowest price applied during a period not shorter than 30 days before the reduction. The Commission's guidance adds that where short promotions repeat, the reduced price from each earlier promotion counts. If you need to check a seller's claimed discount, your evidence is the price history of those 30 days, and a gap in your captures is a gap in that history. This is EU law; it does not apply in the United Kingdom.
Starting points
These are our own judgement, not rules. They suit a page you are watching from outside, without access to its owner's change log.
| What you are watching | Start with | Why |
|---|---|---|
| A page that is already part of a dispute | Once now, again before and after you act | The point is the exact state at each step, not a trend. Capture before you send the letter. |
| Your own terms, privacy policy and public claims | Weekly, plus a capture by hand on release day | You know when you change them. The schedule catches changes you did not make or did not hear about. |
| A supplier's or partner's terms | Monthly; weekly while a contract is being negotiated or renewed | These pages change rarely, and the change that matters is usually announced late. |
| A competitor's advertising claims | Weekly; daily during a campaign | Claims are rewritten when a campaign starts and withdrawn when it is challenged. |
| Competitor prices and promotions | Daily; at the hours the site changes prices, if you know them | Short promotions are exactly what a weekly capture misses. |
| A page before a known event (a sale, a deadline, a product launch) | Daily for the weeks before, then at the event | You are collecting the "before" that will not exist afterwards. |
| Government and policy pages | Weekly | Wording changes quietly, and rarely twice in a week. |
| A source you are citing | Once, when you cite it | You need the version you read, not a history. |
Hourly is for a narrow set of pages: ones that change within the day and where each version matters, such as a stock status or a limited-time offer. For most pages it produces a long list of identical captures.
Correct the interval after a month
Your first guess is a hypothesis. A month of captures tests it.
- If nearly every capture differs from the one before, first check whether the change is real. Dates, counters and rotating banners change a page without changing what you care about. Compare the text of two captures: if the differences are noise, keep the interval. If they are real, shorten it.
- If many captures in a row are identical, lengthen the interval. Eight identical weekly captures mean you can move to fortnightly or monthly and lose little.
- If a change turns up that you only half caught, shorten the interval for that page, and capture it by hand around the next event that might trigger another change.
A sitemap schedule needs the same review, because each page in it changes at its own pace. Group the pages by how often they change, rather than capturing a whole section daily because one page in it does.
Count the storage
More captures means more storage. The cost is easy to work out before you commit. A page of ordinary size, such as a terms page at 1.8 MB, captured daily adds about 54 MB a month. Captured weekly, it adds about 8 MB. Sizes vary widely from page to page, so check the size of a first capture and multiply.
If you are watching many pages, a daily capture of all of them is rarely what you need. Keep the daily schedule for the few where a missed day matters.
Make the gaps explainable
Whatever you choose, write it down. A short note on each page saying why it is captured, and at what interval, answers the question "why do you have a capture from the 3rd but not the 4th?" before anyone asks it. When something happens that matters, capture by hand at once rather than waiting for the next scheduled run. Our checklist in How to preserve a web page as evidence covers the rest of what each capture should record.
A public archive is not a substitute for your own schedule. You do not control when it captures a page, as we explain in Wayback Machine captures as evidence.
How Archivepanel handles it
Archivepanel captures public pages, one page at a time, without logging in. The free tier gives you 100 MB of storage and manual captures. Schedules are part of the paid plan.
- A schedule can run hourly, daily, weekly or monthly, at the hours you list and in a time zone you choose. A schedule set for the 31st still runs in shorter months, on their last day.
- A run that could not happen is not made up later. The schedule continues at its next time, so the gap shows in the record rather than being hidden.
- A sitemap schedule captures the pages a sitemap lists: 20 per run by default, up to 100.
- Compare with previous compares the text of two snapshots of the same address, with removed and added lines marked and changed words highlighted. It compares text only, so a change to images or layout does not appear.
- Each snapshot gets a trusted timestamp from DigiCert when it is captured, and a note and tags on the record can say why the page is being watched. The timestamp is a standard RFC 3161 token, not an eIDAS qualified timestamp.
Sources
- Pew Research Center, When Online Content Disappears, May 2024.
- Shawn M. Jones et al., Scholarly Context Adrift: Three out of Four URI References Lead to Changed Content, PLOS ONE, 2016.
- European Commission, Guidance on the interpretation and application of Article 6a of Directive 98/6/EC, Commission Notice 2021/C 526/02, Official Journal C 526, 29 December 2021.