Sending platform: which ESP each brand uses

What this screen is

The app titles it Sending platform analysis, under the sub-line Analyze email sending platforms (ESP) used by newsletters. One question runs through the whole screen: which email service provider actually put each tracked newsletter into your inbox, and how certain that label is.

Every newsletter carries one platform name and one confidence level. A vendor's own header, or a vendor domain inside Return-Path or Message-ID, gives high confidence. A recognised domain in the List-Unsubscribe header alone gives medium. When the headers identify no platform, the newsletter gets the name unknown and a confidence of none. No AI is involved in any of it.

Five panels use those labels. A doughnut gives market share across your tracked set, and a line chart follows adoption month by month. The brand table puts each brand's primary ESP next to the share that ESP holds across everything you track. A migration timeline flags the month a brand switched. Last is a matrix comparing SPF, DKIM, DMARC and BIMI pass rates per platform, and under all of it a plain counts table backs the doughnut with exact numbers and the confidence split.

Two things to keep straight. Everywhere the word industry appears on this screen it means your own tracked set, so twelve tracked brands produce a twelve-brand market share. And the authentication columns describe how the messages arrived at your tracking address, which depends as much on each sender's own DNS setup as on the platform. It is not a verdict on the ESP's deliverability.

Open it when

• You are shortlisting an ESP and want to see what the brands you benchmark against actually use, with real send counts instead of vendor claims.
• A competitor's email suddenly looks different, with a new footer or a new tracking domain, and you want to confirm whether they migrated ESP and in which month.
• You are writing a competitive or category report and need a line such as 31% of the senders we track run on Klaviyo, with a defensible sample behind it.
• Your own SPF, DKIM and DMARC setup is under review and you want to see which platforms in your sample show the cleanest authentication.
• You want to check how reliable the platform labels are before quoting them, by reading the high versus medium confidence split.
• You spotted a change in sending cadence and want to know whether an infrastructure change went with it.

Screen anatomy

The map below shows the layout top to bottom with sample numbers. The markers match the numbered list underneath.

1. Total newsletters How many collected newsletters in the current filter carry a platform label of any kind, including the ones labelled unknown. This is the sample size behind every panel below.
2. High confidence A count, not a percentage, of the sends whose ESP was proven by a vendor-specific header or a vendor domain in the delivery headers. Divide it by the first tile yourself to judge how solid the labelling is.
3. Platform distribution One doughnut slice per detected platform, sized by send count and ordered from largest to smallest. Senders whose ESP could not be identified are still charted, as an unknown slice.
4. Platform usage over time One line per platform across the months in your window, plotted as absolute sends per month rather than shares.
5. Brand ESP vs. industry share A rail of platform shares across your whole tracked set, then one row per brand: its primary ESP, how much of its own volume goes through it, and how that compares with the tracked set. The green pill is the brand's own share minus the platform's share of everything you track.
6. Platform change timeline One row per brand, one cell per month, coloured by the ESP that dominated that month. A cell with the red marker is the month the dominant ESP changed; hover it for the from and to platforms and the send count.
7. Platform x deliverability Authentication pass rates per platform, with a winner banner above and a colour key below. BIMI is shown but is excluded from the Overall score.
8. Platform distribution details The exact counts behind the doughnut plus the confidence split for each platform, sorted by count descending.
9. Unique platforms and Medium confidence tiles Unique platforms counts distinct labels in the filtered set, and treats unknown as one of them. Medium confidence counts the sends identified from the List-Unsubscribe domain alone. There is no tile for unidentified sends, so those are only visible in the last table.
10. Filter card Brand, From date and To date, side by side. There is no Apply button; changing any control reloads the whole screen immediately.
11. Page header and help button The title, the sub-line, and a round help icon that opens a read-only explainer called Platform Performance Insights. It changes no data.
12. Loading and empty states While the screen loads, everything below the filters is replaced by a spinner and the line Loading sending platform data. With no labelled newsletters the four tiles read 0, the two charts draw an empty plot area, and the three lower cards each print their own message. The grey line No data available appears in a chart only when the data could not be loaded at all.

Every panel explained

The four tiles

Total newsletters, Unique platforms, High confidence and Medium confidence describe the size and the reliability of the sample everything else is built from.

How it is calculated. Total newsletters counts every collected newsletter in the filter that carries any platform label, and Unique platforms counts the distinct labels among them. High confidence counts the sends identified by a vendor-specific header or by a vendor domain in Return-Path or Message-ID. Medium confidence is the count of sends identified only by the List-Unsubscribe domain. All four are plain counts, and when the filter matches nothing they read 0 rather than disappearing.

How to read it. Read the first tile as the denominator, then read High confidence against it. In the sample above, 1,046 of 1,284 sends are high confidence, so roughly four in five labels are proven rather than inferred. Everything left over after high and medium is a send whose ESP could not be identified at all, and there is no tile for that remainder.

Platform distribution

A doughnut with one slice per detected platform, sized by the number of sends attributed to it.

How it is calculated. Each slice is the platform's total send count, every confidence level together, and slices are ordered largest first. Only six colours are used, so from the seventh platform onward they repeat and two slices can share one.

How to read it. This is share of the detected set, which is the number you want when you quote something like Klaviyo is roughly a third of the senders we track. It is not share of your whole archive, and it is not market share in the wider email world.

Platform usage over time

One line per platform, months along the x axis, number of sends that month on the y axis.

How it is calculated. Months come from the date each email was sent, so a newsletter that never carried a send date does not appear here at all. A platform that sent nothing in a given month has no point for that month rather than a zero, so a line can jump across a gap. Colours repeat after six platforms, as they do in the doughnut.

How to read it. Rising and falling adoption inside your window. Because the axis is absolute sends rather than share, a busy month such as Black Friday lifts every line at once, which is not the same as one platform gaining ground. If you care about share, compare against the doughnut.

Brand ESP vs. industry share

A rail of platform shares across your whole tracked set, and under it one row per brand with its primary ESP.

How it is calculated. Industry uses it is the platform's sends divided by all detected sends in the filter, to one decimal. A brand's primary ESP is simply the platform with its highest send count. Brand uses it is that platform's sends for the brand divided by the brand's own detected sends. Share of ESP takes the same numerator and divides it by that platform's total volume across your set, so it says how much of the platform's presence is this one brand. N is the brand's total detected sends.

Also uses lists the brand's other platforms by count, showing the top three as chips with counts plus a plus-N overflow, or the word Single ESP when there are none. The pill under Brand uses it is the brand share minus the industry share in percentage points, green above +1 pp, red below -1 pp and grey in between. Brands are ordered by N, largest first.

How to read it. One row is a sentence: Glossier sends 88.2% of its mail through Klaviyo, while Klaviyo is 31.4% of everything I track, and Glossier alone is 14.9% of the Klaviyo volume in my set. The pill tells you whether the brand over-indexes or under-indexes on its own platform relative to the set.

Platform change timeline

A brand-by-month grid, a migration log and a colour legend, designed to catch ESP migrations as they happen.

How it is calculated. For every brand and month the platform with the most sends becomes that month's dominant platform, and the cell tooltip gives its share of the month and the total sends. A migration is logged whenever the dominant platform differs from the previous month that has data. The migration log underneath lists one line per switch with the brand, the month and the from and to platforms. Brands are ordered by number of switches first and volume second, so migrators float to the top.

Each platform keeps one colour across the whole screen, so the same ESP always looks the same. Cell labels are the first three letters of the platform in capitals, and the column headers show only the month number. The badge reads N migrations detected, amber when there is at least one and green when there are none, followed by how many of your brands changed ESP.

How to read it. A horizontal band of one colour is a stable ESP. A colour change carrying the red marker is a candidate migration. Read the migration log for the direction, hover the flagged cell for the sends it carried, and hover the months on either side, whose tooltips also give the dominant platform's share, before you write it down.

Platform x deliverability

A matrix of SPF, DKIM, DMARC and BIMI pass rates per platform, with an Overall score bar and a winner banner.

How it is calculated. For each platform and each check, the pass rate is the number of sends whose result was a pass divided by the number of sends where that check could be read at all, to one decimal. A check with nothing measured shows a dash. The Overall score is the unweighted average of whichever of SPF, DKIM and DMARC are known; BIMI is deliberately left out.

Cell tints follow four bands. 95% and above, then 80 to 94%, then 50 to 79%, then under 50%, with grey for anything unmeasured. The Overall bar is green at 90 and above, amber from 75 to 89 and red below 75. The winner banner picks the highest Overall score among platforms with at least 3 sends in the sample, and its row is tinted green. A platform with nothing measured gets no row, and the table is sorted by Overall descending.

How to read it. Higher is better in every column. Read N first: a platform with a handful of sends can top the table on noise. A weak column points either at the platform's setup or at that sender's own DNS records, and this screen cannot tell you which.

Platform distribution details

The counts table at the bottom: one row per platform with its send count, its percentage and the split across confidence levels.

How it is calculated. Count is the platform's total sends, all confidence levels merged. The confidence columns split those same sends by the confidence of their label, with unidentified sends falling into None confidence. Percentage divides the platform's count by every newsletter matching your filters, including the ones with no platform label at all, and prints two decimals. Rows are sorted by count descending.

How to read it. This is where you check a label before quoting it. A platform whose volume sits mostly in High confidence is safe to name; one dominated by Medium confidence was inferred from a single header and deserves a spot check against the actual emails.

Metric definitions

Filters and controls

Some things are deliberately missing, so you can stop hunting for them. There is no Apply button and no Reset button. Nothing exports to CSV or PDF, there is no search box, and none of the four tables sorts by column. There is no drill-down either: clicking a doughnut slice or a table row does not open the underlying newsletters. To inspect the individual sends behind a label, go to Newsletters and filter there.

How to use it

1. Set the window, then judge the sample Pick a From and To date wide enough to hold several sends per brand, leave Brand on All brands, and read the tiles. If High confidence is a small fraction of Total newsletters, treat everything below as provisional.
2. Take the market picture from the doughnut, the numbers from the table The doughnut gives you shares of the detected set. Scroll to Platform distribution details for the exact counts and the confidence split, and remember that its Percentage column uses a wider base.
3. Find who runs what In Brand ESP vs. industry share, scan the Primary ESP column, then the pills. A large positive pill means the brand is far more committed to that platform than your set as a whole. Check Also uses before calling anyone a single-platform sender.
4. Check for migrations Read the badge on Platform change timeline. If it is amber, work through the migration log, then hover the months either side of each flagged cell to confirm the dominance was decisive rather than a one-send month.
5. Compare authentication hygiene In Platform x deliverability, read the N column first, then the Overall bars. Treat mid-range SPF or DMARC values as partly unreadable rather than partly failing, and ignore BIMI when comparing overall scores.
6. Confirm before you publish Filter to the single brand you plan to quote, note the label and its confidence, then open that brand in Newsletters and check one recent email against the claim.

Limits and caveats

• Newsletrix recognises a fixed set of vendors: Mailchimp, SendGrid, Mailgun, Amazon SES, Constant Contact, Campaign Monitor, ActiveCampaign, Klaviyo, Braze, Customer.io, HubSpot, Marketo, Pardot, Sendinblue (Brevo), ConvertKit, Bloomreach and Exponea, plus dotdigital, which never gets past medium confidence. A sender on any other platform, or relaying through an intermediary, is labelled unknown.
• The label comes from the delivery headers and nothing else. A tracking domain or a link in the body of the email never changes it, so a brand whose headers carry no vendor trace is labelled unknown even when it is on a mainstream platform.
• Industry on this screen means your own tracked set. Twelve tracked brands produce twelve-brand market share, and adding or removing a brand moves every percentage.
• Authentication results are what the receiving mailbox saw. Results that could not be determined still count in the denominator, so a mid-range pass rate often means unreadable rather than failing. The Overall score is an unweighted mean of up to three checks, so a platform with only DKIM measured is compared directly against one with all three, and the winner banner qualifies a platform on as few as three sends.
• The brand table and the migration timeline both need a newsletter to have been analysed and a brand name identified, so newsletters you collected but never analysed count in the tiles and never appear in those two cards. No migration can be flagged until a brand has at least two months with sends.
• The brand table, the timeline and the authentication matrix can lag up to five minutes behind a fresh import, and the list of brands in the filter up to ten, while the tiles, the two charts and the counts table update immediately. So just after an import the middle of the screen can briefly disagree with the top and the bottom.
• Shares are rounded to one decimal in most panels and printed to two decimals in the details table, so small rounding differences between the two views are expected and are not a bug.

FAQ

Related screens