Duplicates: repeated links and recurring topics

What this screen is

The sidebar calls it Duplicates. The screen itself is headed Duplicates & Recurring topics, with the subtitle "Detection of redundancies and repeated links/topics per brand". It exists to separate two very different kinds of repetition: a brand that reuses its footer, and a brand that has been pointing at the same landing page for six weeks.

Two things feed it. One is the destination each call to action points at, which appears in the rows as a full URL or as a path. The other is the topic categories Newsletrix puts on the email, free text with labels like barefoot, kids or sustainability. Anything that appears in more than one newsletter from the same brand inside the selected period is grouped and shown here.

Repeated links are split into three severity buckets so a stale hero link never sits next to a normal unsubscribe link. Repeated topics get a cadence label and a predicted next appearance date, which turns "this brand repeats itself" into something you can put on a content calendar.

What it does not do, despite what the help panel implies: it never compares email bodies and it never produces a text similarity percentage. There is no "this send was 94% identical to that one" anywhere in Newsletrix. Everything on this screen is derived from repeated destination links and repeated topic labels, nothing else.

Nothing here is an engagement measurement either. Newsletrix receives these newsletters as an ordinary subscriber, so every count is a count of newsletters, never of opens or clicks.

Open it when

• You want to know whether a competitor is recycling the same landing page across sends instead of shipping something new.
• You are planning your own editorial calendar and want to see which topics a brand brings back weekly, monthly or quarterly, and which one is overdue.
• You suspect a brand's "new" campaign is a re-send of the same destination with a fresh subject line.
• You are auditing tracking hygiene and want to see one base URL turning up with five different UTM variants.
• You need a single figure for how fresh a brand's content mix looks right now, to put in a competitive report.

Screen anatomy

One column, top to bottom. The header sits above the filter card, then come three tiles, the trend chart, the severity buckets with their expanded row list, and the cadence table at the foot of the page. Numbers on the mock match the walk-through underneath it.

1. Filter card Two dropdowns: Brand, which starts on the placeholder "All brands", and Time period, which starts on "Last 90 days". There is no Apply button. Changing either one reloads all three panels at once.
2. Total duplicate links How many repeated destination groups exist under the current filter. It adds the exact-duplicate and template-re-use rows together and ignores near duplicates, which have their own list.
3. Freshness score A 0 to 100 blend of topic variety and content recency, shown to one decimal, colour coded green at 70 and above, yellow from 40 to 69.9, and a red badge below 40. It is one brand's score, not an average.
4. Topic repetition trend A weekly line showing average mentions per distinct topic. Newest week on the left, so it reads right to left in time. With nothing to plot the card shows "No trend data available".
5. Severity breakdown buckets Three clickable tiles named Exact duplicate, Near duplicate and Template re-use. Each carries a row count, a one-line description and an Action line. Clicking one opens its list and closes the others; clicking the open one collapses everything.
6. Expanded bucket list A grey strip with the bucket name and count, then the rows. Exact and template rows show the brand chip, the URL or path, the destination type, how many newsletters contained it, its coverage percentage and the day span between first and last appearance.
7. Recurring topic cadence The distribution strip first: one small box per cadence label that has at least one topic, with its count. Buckets with zero topics are hidden entirely, so a missing "Quarterly" box means no quarterly topics, not a rendering problem.
8. Next expected column The last-seen date plus the topic's average gap. The small line underneath is relative to today: red when the date has passed, amber for today and anything inside the next seven days, muted grey beyond that.
9. Header and help icon The title, the subtitle and a question-mark button on the right that opens a static explainer called "Content Uniqueness Management".
10. Recurring topics tile The middle tile. It counts topic categories seen in two or more newsletters. With "All brands" selected it counts brand and topic pairs, so one topic shared by three brands adds three.
11. Empty and loading states With nothing to show, the tiles read 0 and 0.0, the chart says "No trend data available", the open bucket shows a dashed box such as "No exact duplicate rows for this brand and period.", and the cadence panel hides its distribution strip. There is no loading spinner, so the previous numbers stay on screen until the new ones arrive.

Every panel explained

Total duplicate links

How many distinct repeated destinations exist for the current brand and period.

How it is calculated. A group is one brand, one destination type and one URL or path, and it counts only when that destination turned up in more than one newsletter from the brand inside the window. The groups with the most newsletters behind them come first, and the tile stops at 100, so it can never read higher. The count includes both the exact-duplicate and the template-re-use rows; near duplicates are excluded.

How to read it. It counts repeated destinations, not emails and not clicks. Read 40 as 40 distinct link groups reused across at least two newsletters. Exactly 100 means 100 or more.

Recurring topics

How many topic categories appeared in more than one newsletter inside the period.

How it is calculated. Topics are grouped by brand and category, and a topic counts only when two or more distinct newsletters carry it. This tile stops at 100 as well.

How to read it. Higher means the brand keeps returning to the same themes. It is a raw count, not a share of their content, and rows are per brand and topic, so one theme shared by three brands adds three to the tile.

Freshness score

A single 0 to 100 figure for how varied and how recent one brand's content mix is.

How it is calculated. Two halves worth 50 points each. Variety is distinct topics divided by total topic mentions, multiplied by 50, so a brand that never repeats a topic takes the full 50. Recency is 1 minus the average age of its newsletters in days divided by 365, floored at 0 and multiplied by 50, so content sent today takes the full 50 and content a year old takes none. The two halves are added, capped at 100 and rounded to one decimal. If the brand has no analysed newsletters or no topic mentions in the window, the score is 0. Colour bands run green at 70 and above, yellow from 40 to 69.9, and a red badge below 40.

How to read it. A rough content-mix indicator for one sender, not a ranking. A brand publishing many distinct topics very recently approaches 100; a brand hammering three themes in older sends approaches 0.

Topic repetition trend

A weekly line of average mentions per distinct topic, labelled "Topic repetition rate".

How it is calculated. For each brand and each calendar week, all topic mentions are divided by the number of distinct topics in that week and rounded to two decimals. Points run newest week first. The x labels are the date of the first newsletter inside each week.

How to read it. 1.0 means every topic that week was mentioned once, which is maximum variety. 2.5 means each topic was hit roughly two and a half times on average, so the week leaned repetitive. Read the shape, not the individual point.

Severity breakdown

The same repeated links, sorted into three buckets with a suggested action on each.

How it is calculated. For every repeated destination, coverage is the number of newsletters containing it divided by the total number of that brand's analysed newsletters in the window, shown to one decimal. Coverage of 70% or more files the row as Template re-use; anything below 70% is filed as Exact duplicate. Near duplicate works on query strings instead. Links that are identical up to the first question mark and differ only after it appear as one base path, and a base path needs two or more different full URLs behind it before it shows up at all. The rows holding the most variants come first, each one lists up to five of them, and the remainder is summarised as a "+N more variants" line.

How to read it. Work Exact duplicate first: those are content links being reused, which is why the action line says "Replace next send". Near duplicate is a tracking-hygiene finding rather than a content one. Template re-use is normal behaviour and only deserves a link-health check.

Clicking Near duplicate replaces the row list with the variant view. Each row shows the base path once, then the individual full URLs indented underneath it.

Recurring topic cadence

Every recurring topic with its brand, mention count, cadence label, last-seen date and predicted next appearance.

How it is calculated. For each brand and topic, the day span is the number of calendar days between its first and last appearance in the window. Cadence days is that span divided by the number of appearances minus one, rounded to one decimal, and it is only shown when the topic appeared at least twice and the span is greater than zero. The label bands are fixed: 10 days or fewer is Weekly, up to 17 is Bi-weekly, up to 35 is Monthly, up to 100 is Quarterly, anything longer is Occasional. Next expected is simply the last-seen date plus the cadence days. The table sorts by cadence ascending, with topics that have no cadence last and ties broken by mention count, and it draws at most 100 rows.

How to read it. A topic marked "Weekly · 7d" whose next-expected date has gone red is a theme the brand normally hits by now and has not. That is either a gap you can occupy this week, or a signal the brand has shifted focus. A topic that just shipped is one to rest in your own calendar rather than chase.

Help panel: "Content Uniqueness Management"

The question-mark button in the header opens a fixed explainer with the same text for every account.

How it is calculated. It is not. Nothing in the panel comes from your own newsletters.

How to read it. With a pinch of salt. It mentions Repeated Sections, Similarity Scores and Source Tracking. None of the three are computed on this screen. For structural repetition go to Sections instead.

Metric definitions

Filters and controls

Some controls are deliberately absent, so you can stop hunting for them. No CSV or PDF export, no date pickers, no search box, no sort controls and no way to select more than one brand. Table order is fixed in the product, cadence ascending for topics and most-repeated first for links. One control does reach this screen from outside it. Pick a workspace in the switcher at the top of the sidebar and all three panels count only newsletters delivered to tracking addresses carrying that label.

How to use it

1. Set the period to match the question Leave Brand on "All brands" and pick a real planning cycle. Last 90 days, the default, suits a quarterly review; Last 30 days tells you what a brand is pushing right now. The window clips every span, cadence and coverage figure on the screen.
2. Read the three tiles as a pair plus a caveat Total duplicate links and Recurring topics tell you how much repetition exists at all. The Freshness score is a single brand's figure, so note which brand chips dominate the lists before you quote it anywhere.
3. Open Exact duplicate and scan coverage This bucket carries the content findings. Rows near 40% to 60% coverage are the interesting ones: a link in half a brand's sends is a deliberate push, not furniture. The day span then tells you whether it was a fortnight campaign or a link nobody ever retired.
4. Check Near duplicate for tracking hygiene One base path with five variants is a brand running separate campaign tags on the same page. It tells you how they segment campaigns, and it is the fastest audit of your own tagging when you are looking at your own sends.
5. Turn the cadence table into calendar slots The fastest cadences sit at the top. Anything red in Next expected is a theme the brand normally has covered by now, and those are the weeks where a well-timed piece from you lands in an empty inbox slot.
6. Re-run the same window a month later Compare Total duplicate links and the freshness score with the previous read. Repetition climbing while freshness falls is a brand running out of new material.

Limits and caveats

• The Brand dropdown does not currently narrow this screen. Selecting a single brand returns empty panels, so leave it on "All brands" and read the brand chips.
• Row caps stack. At most 100 duplicate link groups, 50 near-duplicate groups and 100 recurring topics, with each bucket list drawing at most 50 rows and the cadence table 100. A tile reading exactly 100 means 100 or more.
• Every count is newsletter-level. Nothing here measures engagement, and Newsletrix cannot see another sender's opens or clicks.
• Links are read out of the email itself, and not always in full. A destination can appear as a path fragment rather than a full URL, or as a tidied version rather than the literal link. Near-duplicate grouping also only looks past the question mark, so tracking IDs buried in the path, such as /p/abc/track123, stay in separate rows.
• Topic labels are free text, so two spellings of the same theme count as two topics and slightly inflate the variety half of the freshness score.
• Cadence on a small sample is noisy: two mentions produce a cadence from a single gap, and two mentions on the same day produce no cadence at all.
• The period changes the answers. Widening from 90 days to a year can move a topic from Weekly to Monthly and lower the freshness score without the brand changing anything.
• A workspace left selected in the sidebar silently narrows all three panels, and the filter card gives no hint that it is active. If the numbers look thinner than expected, check the switcher before blaming the data.

FAQ

Related screens