# Did my change work? *The same days before and after, side by side* > The page https://skaldanalytics.com/features/did-my-change-work/ as Markdown. Every page and guide: https://skaldanalytics.com/llms.txt . Skald is source-available (Elastic License 2.0). Features · did my change work You rewrote the pricing page two weeks ago, and the only answer is a feeling. To measure website changes in Skald, mark the day on the page: it sets the same number of whole days on each side, a verdict per figure, and “too early to tell” until it can say. *A tracked change* is a mark on one page: the day it changed, what changed (content, a new title, a new backlink (a link to the page from another site), a redirect, or other) and a short note. Skald compares up to 28 whole days on each side; the day of the change counts in neither, because it changed during that day. [Start your 30-day free trial](https://skaldanalytics.com/beta/?choice=trial#signup) [Open the live demo](https://spcaeo.github.io/skald/) 30-day free trial · no card · we set it up within one working day · free to self-host - Mark the day you changed a page, and… - Up, Down, or just an ordinary week? - A missing day never counts as a bad… - Did it bring more visits, leads,… - The changes you forgot to mark, found… - A before-and-after you can defend in… Traffic › Pages › /pricing/ · sample Compared Changed 16 Dec: New title Whole days since the change: **28** 28 whole days after the change (17 Dec to 13 Jan) against the 28 before it (18 Nov to 15 Dec). Visits that started here: 3.3 a day before, 5.9 after, +80%. **Up** drag the days Sample visits a day on one page. The windows stay anchored at the change: after 28 days on each side they stop growing. [01] mark it ## Mark the day you changed a page, and stop guessing Open the page on Traffic › Pages and select **+ Mark a change**. Pick the day, what changed and, if you like, a note of up to 200 characters. Everyone on the site's team then sees it, until someone stops tracking it. - A change marked weeks late is fine: pick the old day, and once 15 days or more have passed the verdict is there at once. - The change shows as a marker on every day chart, so a jump in Sources or Pages has its reason beside it. - Only owners and admins mark and stop changes, on a Skald with accounts. A change marked today shows its figures from the next data update. [All Skald features](https://skaldanalytics.com/features/) [website traffic comparison](https://skaldanalytics.com/features/what-changed/) pick what changed /pricing/ northwind.example · Traffic › Pages Changed 16 Dec: New title “New title and a shorter plan table” Compared · 28 whole days each side In the address `pcm=new` `pcd=2026-12-16` `pck=title` The note kept in Skald with the mark, never in the address Sample mark. The dialog's address keys reopen it with its day and kind; the note you typed stays out of every link. [02] the verdict ## Up, Down, or just an ordinary week? Every page wobbles from day to day, so a higher average proves little alone. Skald calls a figure Up or Down only when the difference per day is more than 2 standard errors of the two sides' daily figures, the normal swing. Under that it says No clear change, and with too little to go on, Too few to tell. pick a page Visits that started here, 14 days each side Difference per day +2.29; the swing band is ±0.84 (2 standard errors). Up 3.3 visits a day before, 5.6 after: +70%, far outside the day-to-day swing. Up. Sample pages. The verdict is worked out here with the rule Skald's code uses: a side needs the figure on 7 days or more, the two sides at least 10 together (left at once: 10 visits on each side), and a difference beyond 2 standard errors. [Track your own page changes, 30 days free](https://skaldanalytics.com/beta/?choice=trial#signup) [Open the live demo](https://spcaeo.github.io/skald/) [03] fair sides ## A missing day never counts as a bad day A day can be withheld because its two counts disagreed, or was never recorded. Counting it as 0 would make your change look worse than it was. So that day goes in neither side, and when the sides end up with different numbers of days, Skald compares them per day and says so. pick a case [analytics data accuracy](https://skaldanalytics.com/features/counted-twice/) Before · 2 to 15 Dec 1. 2 2. 4 3. 3 4. 3 5. 5 6. 2 7. 4 8. 3 9. 3 10. 4 11. 2 12. 3 13. 5 14. 3 16 Dec · the day it changed · in neither After · 17 to 30 Dec 1. 5 2. 4 3. 6 4. 5 5. 7*withheld* 6. 4 7. 6 8. 5 9. 8 10. 6 11. 5 12. 7 13. 6 14. 4 14 whole days on each side: the totals compare as they are. Sample days. The withheld-day and leads notes are the page's own words, shown under the figures. [04] six figures ## Did it bring more visits, leads, search and AI readers? A page can win on visits and lose on leads. So each change is measured six ways, every figure with its own verdict: one change rarely moves everything the same way, and the table shows where it did. flip it MeasureBeforeAfterChangeWhat it says Visits that started here**3.392 in 28 days5.9166 in 28 days**+80%Up Leads and conversions that landed here**0.12 in 28 days0.25 in 28 days**+150%Too few to tell Visits from search engines**1.438 in 28 days2.261 in 28 days**+61%Up Visits from AI assistants**0.12 in 28 days0.13 in 28 days**+50%Too few to tell AI reads (live fetches)**0.925 in 28 days1.131 in 28 days**+24%No clear change Left at once**85.9%79 of 92 visits54.8%91 of 166 visits**−31.1 points Down Sample: the /pricing/ title change of 16 Dec, 28 whole days on each side, worked out with Skald's own page_changes code. AI reads are assistants such as ChatGPT-User fetching the page live for someone. For “left at once” a lower share is better, so its Down is good news. [lead attribution](https://skaldanalytics.com/lead-attribution/) [ChatGPT traffic](https://skaldanalytics.com/ai-traffic-analytics/chatgpt-traffic/) [05] nothing forgotten ## The changes you forgot to mark, found for you Edits often go unmarked. When your website's page list or your sitemap's “last changed” date (``) says a page changed on a day you have not marked, the page asks **Track that change?**. Every tracked change is then listed on Overview › Insights, under “Done, and did it work?”. `` ` https://northwind.example/blog/fall-launch/` ` 2027-01-09` `` Your sitemap says this page changed on Sat 9 Jan 2027. press it Pages you changed: did it work? Pages you changed, sample, today 15 Jan | Page | Changed | What changed | Visits a day, before | After | What it says | |---|---|---|---|---|---| | /blog/fall-launch/ | 9 Jan | Content update | counting | counting | Waiting for the next update | | /features/ | 26 Dec | Content updateRewrote the top section around server logs | 4.1 | 4.4 | No clear change | | /pricing/ | 16 Dec | New titleNew title and a shorter plan table | 3.3 | 5.9 | Up | Sample list, as Overview › Insights draws it. A new mark waits for the next data update, then counts every whole day after 9 Jan; its verdict comes on 24 Jan. [06] what it can and cannot say ## A before-and-after you can defend in a meeting The tracker is built to be quoted. That means saying plainly where its answer ends. ### What it does not do - It does not prove cause. Up says the figure rose beyond the day-to-day swing after that day; a season, a campaign or a search update in the same weeks moves figures too. It is a before-and-after comparison, not a split test (one that shows two versions of a page to visitors in the same weeks). - It compares one page at a time, at most 28 whole days on each side, and needs 14 whole days after the change for a verdict. - Marking needs accounts (signing in); owners and admins mark and stop, the rest of the team sees. On the demo nothing can be marked. - The figures come from visits by landing page, kept from the first day collected after the 2.35.0 upgrade; a change on an older day says “not recorded” for the days before. A page group shows changes for the whole site only. - A day keeps its 200 busiest landing pages; a quieter page can show a floor, marked with a +. - Leads need the conversions data. Each change is counted twice, the second time straight from the stored days; a change the two counts disagree on is left out and named. [What changed this period](https://skaldanalytics.com/features/what-changed/) [See the plans](https://skaldanalytics.com/pricing/) keep reading ## More on measure website changes - [What changed](https://skaldanalytics.com/features/what-changed/): Now against before: key figures, sources, landing pages, and the pages that need work. - [Google Search Console](https://skaldanalytics.com/integrations/google-search-console/): Queries, clicks, impressions and position, beside the visits your log counted. - [Skald pricing](https://skaldanalytics.com/pricing/): Per site, a 30-day free trial with no card, or free to self-host. questions ## Measure website changes FAQ Straight answers, including what Skald does not do. [Read the docs](https://skaldanalytics.com/docs/) [Ask us a question](https://skaldanalytics.com/contact/) **How does Skald tell if a page change worked?** You mark the day you changed the page. Skald takes the whole days after that day, up to 28, and the same number of days just before it, and sets six figures side by side: visits that started on the page, leads and conversions that landed there, visits from search engines, visits from AI assistants, AI reads, and the share who left at once. **Why does it say “too early to tell”?** A verdict needs at least 14 whole days after the change. Until then the page says how many days it has and the day the verdict comes, so a lucky first week is never read as a win. **What does “No clear change” mean?** The difference per day is within the normal day-to-day swing: 2 standard errors of the two sides' daily figures. The figure may have moved a little, but not by more than an ordinary week moves it. **What if a day is missing or withheld?** A day that is not whole, not verified or not recorded goes in neither side, and is never counted as 0. When the two sides then have different numbers of days, Skald compares them per day and says so. **Who can mark a change?** Owners and admins, on a Skald with accounts (signing in). Everyone on the site's team sees the tracked changes. Without accounts the page says why nothing can be marked. **Does an Up mean my change caused it?** No. Up means the figure is higher after the change by more than the day-to-day swing. Other things happen in the same weeks, such as a season, a campaign or a search update, so read it with what else you know. It is a before-and-after comparison, not a split test, which shows two versions of a page to visitors in the same weeks.