Trending topics: what your market is talking about

What this screen is

Trending topics is a word counter with a calendar attached. Words come from two places, the subject line and the body text, and every newsletter also belongs to the week it was sent. That gives you one tally for the whole window and a second tally week by week. The word cloud shows the window total, the timeline shows the weekly counts, and the arrows on the chips compare the last two weeks.

That is all it does, and the limits follow from it. Words are never reduced to a common root, never gathered into themes, and a rare word earns no extra credit for being rare. The word sale and the word sales are two different keywords. A word that appears twelve times inside one long email contributes twelve to the count, exactly as if twelve senders had each used it once.

Brand names, product names and campaign slogans all count as ordinary keywords, which is why a retail cohort usually produces a cloud full of commerce vocabulary rather than themes. Filter down to a single brand and that brand's own name will often become the top keyword. Read the output as market vocabulary, not as a list of topics somebody curated.

One panel works differently. The scatter plot of body word count against estimated open rate has an AI estimate up its vertical axis. Newsletrix receives competitor newsletters as an ordinary subscriber and can never see another sender's real opens, so there is no measured open rate to plot. Treat that chart as a sanity check on a common argument about email length. It is not evidence.

Open it when

• You are planning next month's editorial calendar and want to see which themes competitors are pushing right now.
• A keyword suddenly appears with a sparkle badge and you want to know whether a competitor has started a new campaign narrative.
• You need to decide whether a topic is a one-off spike or a recurring beat before you commit to a long-form piece.
• You suspect the market has drifted away from a theme you still write about every week.
• Someone claims longer emails get better opens and you would rather look at the chart than argue.
• A client asks what their competitors have been saying for the last quarter and you need a defensible summary in one screen.

Screen anatomy

The screenshot below shows the middle of the screen on a small sample account. The annotated map after it walks the current layout in order.

1. Filter card Two dropdowns, Time period and Brand. Time period offers Last 7 days, Last 30 days and Last 90 days, and starts on Last 30 days. Brand lists every tracked brand with its newsletter count next to it. Changing either one redraws every panel below.
2. Count tiles Total newsletters, Number of weeks, Unique keywords and Top keyword. They describe the sample the rest of the page is built from, so read them first. If Total newsletters is small, everything below is anecdotal.
3. Unique keywords Worth calling out separately. It is the size of the word-cloud list and that list stops at 30, so on any healthy account it reads 30 forever. It is not the vocabulary size of your market.
4. Word cloud - Top keywords The 30 most frequent words in the window, drawn as coloured chips. Size and weight grow with the total count, and the colour shows direction. Each chip carries its total count, a change arrow and, for brand-new words, a sparkle.
5. Word count vs. Est. open rate One dot per newsletter, body length across the bottom and estimated open rate up the side, with a green trend line, a dashed industry reference line and a correlation badge. This is the only panel that ignores the time period filter.
6. Keyword trend timeline Up to 10 keyword lines across the week buckets, Frequency on the vertical axis and Week along the bottom. This is where you separate a real trend from a single campaign burst.
7. Page header and help icon The title Trending topics & Keywords with its subtitle, and a round help icon on the right that opens the Stay Ahead of Trends explainer. The explainer is fixed guidance text and never reflects your data.
8. Tab bar Two tabs, Overview and Seasonal Patterns. Overview is the default. The choice is not remembered, so every visit starts on Overview.
9. Seasonal Patterns tab A short intro card, three guidance cards, the Historical pattern analysis list of your five most frequent keywords with their weekly average and peak, a fixed tips box, and a four-step preparation workflow. Only the keyword list is built from your data.
10. Empty states With nothing to count, the tiles show 0, the Top keyword tile shows a dash, the cloud says No keyword data available, the timeline is not drawn at all, and the seasonal tab says No trend data available yet. Check back after collecting more newsletter data.

Every panel explained

The four count tiles

Total newsletters, Number of weeks, Unique keywords and Top keyword. Together they describe the sample, and each one has a quirk worth knowing.

How it is calculated. Total newsletters counts collected newsletters that have body text, a send date inside the selected window and a finished analysis. Anything collected but not yet analysed is invisible here. Number of weeks snaps every send date back to its Monday and counts the distinct Mondays that contain at least one newsletter, so a 30-day window normally produces 5 or 6 partial weeks rather than a tidy 4. Unique keywords is simply the length of the word-cloud list, and that list is cut at the 30 most frequent words. Top keyword is the first entry of the same list.

How to read it. Total newsletters is the denominator for everything else. Number of weeks is the resolution of the timeline: a 1 there means there is no week-over-week comparison at all, so no arrows and no sparkles appear anywhere on the screen. Top keyword tells you the dominant vocabulary of the set, which for a retail cohort is usually a commerce word rather than a theme.

Word cloud - Top keywords

The 30 most frequent keywords in the window, each shown with its total occurrence count, its week-over-week change and, when relevant, a new-keyword sparkle.

How it is calculated. Counting covers the subject line and the body text together. Case is ignored, so Sale and sale are the same keyword, and nothing shorter than three characters ever appears. Words containing digits or accented letters are eligible. Two kinds of word never appear at all. One is a fixed list of about 80 stopwords, which covers ordinary English filler plus email noise such as newsletter, unsubscribe, subscribe, click, here, view, read, email, link and the common URL fragments. The other is any word made of ten or more plain lowercase letters, a blunt rule aimed at random tracking identifiers. Every surviving occurrence counts, repeats inside one email included. Chip size and weight grow with the total count and level off once a word is frequent enough, so the two loudest words can end up the same size. The arrow compares the last week bucket with the one before it: the change is the difference divided by the previous week, times 100, to one decimal place. A word with zero occurrences in the previous week and some in the last one is marked new instead.

How to read it. Size is the window total; the arrow compares only the last week bucket with the one before it. Big and rising is an established theme still gaining. Small and rising is where new narratives show up first. A sparkle means the word was completely absent from the previous week bucket, which is the clearest signal on this screen that a sender has started something new.

Word count vs. Est. open rate

A scatter plot with one dot per newsletter. Body word count runs across the bottom and estimated open rate up the side, with a green straight-line fit through the cloud and a dashed grey industry reference line.

How it is calculated. The chart plots the 500 most recent newsletters that have both a body word count and an estimated open rate, restricted to the selected brand if you set one. Word count is the number of words in the body text. Newsletrix estimates the open rate from the content of the email, and every estimate carries a source note explaining what it leans on. The green line is an ordinary least-squares fit. The badge reports the Pearson correlation and labels its size regardless of sign, so 0.5 or above is strong, 0.3 to 0.5 moderate, 0.1 to 0.3 weak, and anything below that no correlation. Below three points, or with no variation on one of the axes, you get no line and no correlation at all. The dashed line is the published average open rate for your selected benchmark industry, which defaults to retail at 15.68 percent.

How to read it. A rising green line means longer emails tend to carry higher estimated opens in your sample, and a flat one means length is not doing much. Check the sample size under the badge before you take the shape seriously, and remember that a weak label means the line is essentially noise.

Keyword trend timeline

A line chart with one series per keyword across the week buckets. The vertical axis is titled Frequency and the horizontal axis Week.

How it is calculated. The ten most frequent keywords in the window each get a line in their own colour, with one point per week bucket. A week where the keyword is missing sits at zero. Only weeks containing at least one newsletter become points on the axis.

How to read it. A line that climbs across the last two or three weeks is a live trend. A line with one tall spike and a flat remainder is a campaign, not a trend, and copying it a month later means arriving after the market has moved on. Compare the shapes rather than the heights, because a keyword's absolute height depends on how wordy its senders are.

Historical pattern analysis, on the Seasonal Patterns tab

A list of your five most frequent keywords, each with a weekly average, a peak week and sometimes a Pattern detected badge.

How it is calculated. For each of the five keywords the app sums its weekly counts and divides by the number of week buckets to get the average, then takes the highest single week as the peak. The Pattern detected badge appears when that peak is more than 1.5 times the keyword's own weekly average. Nothing else is tested. There is no comparison against the same week last year and no calendar of holidays behind it.

How to read it. Treat the badge as a spikiness flag, not proof of a season. It says one week stood more than 50 percent above that keyword's own norm inside a window of at most 90 days. Over a 7-day window, where there may be only one or two buckets, almost anything trips it.

The guidance panels

Proactive content strategy, the three cards Plan ahead, Timely launch and Quality content, the yellow Actionable insights box and the four-step Content preparation workflow.

How it is calculated. They are not. All of this copy is fixed editorial text. It does not read your newsletters, it does not change when you switch brand or window, and the workflow steps are not clickable or trackable.

How to read it. As framing for the tab, the way an introduction sits at the top of a report. One catch: the Actionable insights box only appears when there is keyword data, so it disappears together with the keyword list on an empty account.

Metric definitions

Filters and controls

You cannot export or download anything from this screen. There is no group or tag selector, no keyword search box, no custom date range beyond the three presets, no minimum-count control and no manual refresh button. Figures here can lag a few minutes behind a fresh import, so a newsletter that arrived moments ago may not be counted yet; reload the page to pick it up. If you need the individual emails behind a keyword, work from Newsletters.

How to use it

1. Set the window before you read anything Start on Last 90 days. It gives the timeline enough week buckets to show a shape, and it makes the seasonal averages less jumpy. Drop to Last 30 days only when you are checking for something new.
2. Sanity-check the sample Look at Total newsletters and Number of weeks. Fewer than two weeks means no arrows and no sparkles anywhere on the screen, and a low newsletter count means one wordy sender is probably driving the whole cloud.
3. Read the cloud by size first, arrows second Size tells you what the market talks about habitually. Arrows tell you what changed last week. Write down the two or three sparkle words and the biggest risers; ignore everything in the middle.
4. Confirm each candidate on the timeline Find the word in the legend and look at its line. A sustained climb over several weeks is worth planning around. One spike and a flat remainder is a campaign that has already happened.
5. Find out who is behind it Set the Brand filter to each of your main competitors in turn and watch whether the keyword survives. If it only exists for one sender, it is that sender's campaign, not a market movement.
6. Commit the winner to your calendar A theme that is big, rising, present across more than one brand and visible for three weeks or more is the one that earns a long-form piece.

Limits and caveats

• These are raw occurrence counts. Words are never reduced to a common root, related words are never grouped and a rare word carries no extra weight, so sale, sales and sell are three separate keywords and one long, repetitive email can dominate a week.
• Nothing here groups words into topics. The screen shows word frequencies, not themes, and it will never tell you that three different words belong to the same story.
• The stopword list is fixed and cannot be edited from the app, so brand names, product names and campaign slogans all compete with real vocabulary for the 30 available slots.
• Everything is capped. The cloud stops at 30 words, the timeline at 10 lines, the seasonal list at 5 entries and the scatter at 500 newsletters. Anything below those cut-offs is invisible rather than shown as small.
• Weeks start on Monday and come from the send timestamp as the sender wrote it, so a Sunday night send can land either side of the boundary depending on the sender's timezone. The first and last weeks of any window are almost always partial, which drags the change arrows down early in a week and up at the end of one.
• Seasonal Patterns is a generous name for a tab whose longest window is 90 days. It cannot see last year, and the Pattern detected badge is only a spike test on a short series.
• The scatter's vertical axis is an AI estimate of open rate, its dashed line is a published industry average rather than your own benchmark, and correlation is not causation in either direction.

FAQ

Related screens