How this map works

PLC Pulse is a public prototype that maps how UK-listed companies and their leaders communicate with the market. It covers 78 companies and 315 tracked leaders across 8 sectors. The graph, the company and people panels, the organisation board, peer comparisons and Ask are all computed from the dataset described below — there is no separate editorial layer, and nothing is fetched from a live feed.

The leader count is a count of roles: four chairs hold a tracked role at two companies, so the 315 roles belong to 311 people.

Two kinds of data sit behind every view. Identity is checked; communication is modelled. Panels mark the difference with the same two symbols used here.

Companies

78

Tracked leaders

315

Modelled posts

12,299

Announcements

616

News items

1,044

Sectors

8

The window runs from Mon 16 Mar 2026 to Sun 13 Sep 2026 — 26 whole weeks, Monday to Sunday — and the dataset is labelled as of 13 Sep 2026. Times are UK wall-clock. Posts are modelled on LinkedIn and X. Each company has between 3 and 6 tracked leaders — its senior public voices, not the full board. A few people hold a tracked role at two companies, so the 315 tracked leaders are 311 different people.

By index:

FTSE 100 (52)

·FTSE 250 (19)

·AIM (7)

By role:

CEO (78)

·CFO (78)

·Chair (78)

·Founder (1)

·Investor relations (42)

·Communications (37)

·Executive (1)

Posts include the company account and every tracked leader. Results, trading updates and AGMs follow the typical UK reporting pattern for each company’s financial year end; they are not the companies’ real 2026 dates. Unscheduled announcements are generic modelled updates, not real events.

Every figure is calculated from the records for the 26-week window. Company figures combine the corporate account with every tracked leader unless the label says otherwise. Medians are used wherever a single unusual post could distort an average.

Posts per week

All posts in the window divided by 26. Corporate and leadership cadence are the same count split by author.

posts per week = posts ÷ 26

Leader status

Active: at least 0.5 posts a week, about one a fortnight. Occasional: at least 0.1, about one every 10 weeks. Silent: anything less. A company’s active leaders are the tracked leaders at active status. Across the dataset, 136 leaders are active, 58 occasional and 121 silent.

Role comparison

A leader’s posts per week against the median for their role (CEO, CFO, chair and so on). The median only counts people in that role who posted at least once in the window, so it describes those who post, not the whole role. No comparison is shown for a role with fewer than 5 such people.

median( posts per week of people in the role with at least one post )

Leadership engagement share

The share of a company’s engagement that lands on leadership posts rather than the company account.

leadership engagement ÷ (leadership engagement + corporate engagement)

Median engagement per post

The middle value of engagement (interactions) across a set of posts. One exceptional post does not move it.

Announcement follow-up

The share of a company’s announcements where at least one tracked leader posted about that announcement within 48 hours of it. The same test applied to the company account gives corporate coverage.

announcements with a leadership post within 48h ÷ announcements

Response time

For announcements that were followed up, the median number of hours from the announcement to the first such leadership post. Announcements with no follow-up are left out rather than counted as slow.

Event concentration

The share of all posts published in the 72 hours after any of the company’s announcements. A high share means output bunches around the company’s own news.

posts within 72h after an announcement ÷ posts

News-linked lift

Each post is compared with its own account’s usual post: the median engagement of that account’s posts that are not linked to news coverage. The lift is the median of those ratios across news-linked posts, so 1.0× means no difference. It is left blank when too few news-linked posts can be compared.

median( post engagement ÷ same account’s median non-news engagement )

Topic lift

The same ratio, grouped by topic. Comparing each post with its own account stops one prolific executive from inflating a topic.

Consistency score

How steadily a company communicates. Active weeks are weeks with at least one post from any account. Dispersion is about 1 when posts arrive steadily at random and rises as they bunch together; above 1, evenness starts to fall.

score = 100 × (active weeks ÷ 26)² × evennessevenness = clamp(1.15 − 0.15 × dispersion, 0.1, 1)dispersion = weekly variance ÷ weekly mean

Scores of 72 and above read as Consistent, 42 to 71 as Uneven, and below 42 as Sporadic.

Worked example · Babcock

Posted in 24 of 26 weeks. Weekly mean 4.6, variance 5.9, so dispersion is 1.28 and evenness 0.96.100 × (24 ÷ 26)² × 0.96 → 82 · Consistent

Quiet periods

The longest run of days with no post from any account, and the number of gaps longer than two weeks.

Momentum

Average weekly posts in the last 4 weeks (17 Aug – 13 Sep) against the average of the 8 weeks before (22 Jun – 16 Aug), as a percentage change. Left blank when the earlier period had no posts.

(mean of last 4 weeks − mean of previous 8) ÷ mean of previous 8

Voice concentration

The share of leadership posts written by the single most active leader.

Communication styles

Each company gets one measured style, assigned by rules over the metrics above. The rules are checked in order — quiet, event-led, one main voice, leadership-led and corporate-led — and a company that matches none is “mixed”. Styles describe the records, not the generator’s inputs.

Filter bands

Filters dim companies that fall outside a band; they never change the underlying figures. A topic filter matches a company when the topic is among its three most frequent, leaving routine results posts aside.

A line between two companies is always explainable. Each pair is scored on the reasons below, and every link keeps its reasons so the panels can say why two companies sit together. A link can carry several reasons at once.

Keeping the graph sparse

• Each company nominates its strongest links — about three — and keeps its curated peers on top. Links nominated by the other company count too, which is why a few companies have more.
• A link across sectors needs a very similar topic mix and the same style. Curated peers aside, each company nominates at most one.
• Every company keeps at least one link; if none qualifies, it is joined to its closest company in the same sector.
• The map currently draws 131 links between 78 companies: a median of 3 per company, at most 6, and 8 across sectors.
• In People mode, two leaders at different companies are linked only when both post often enough to compare, they hold the same role, and they either work in the same sector or talk about very similar topics — 181 links in all.

Peers for comparison

Compare and the peer figures in each panel use the same ordered list, taking the first five:

1. 01Curated peers, in either direction.
2. 02Graph neighbours in the same sector, strongest link first.
3. 03The remaining companies in the same sector, closest topic mix first.

Peer figures are medians and are only shown when at least two peers have a value.

Insights are generated, not written. Each one is a rule over the metrics above, and every number in an insight is a computed value from the same records.

1. 01Calculate the company’s metrics and those of its peers.
2. 02Test each rule against its threshold. A rule only fires when the evidence clears it — for example, “leadership is silent” needs every tracked leader to fall short of active status.
3. 03Score each insight: a base weight for how telling it is, plus how far past its threshold the company is.
4. 04Show the strongest, at most one per theme (who speaks, engagement, announcements, rhythm, topics, peers) and up to four. If that leaves fewer than three, the next strongest fill in from any theme; a plain cadence line is the last resort.
5. 05Choose the “pattern” sentence by rules over the same metrics — for example, event-led output with falling momentum, or a single leader carrying the voice.

Across 78 companies, 23 different insight rules fire and panels show 3–4 insights each, with 13 distinct pattern sentences in use.

Every company and its leadership list were checked against public sources in Sep 2026: mostly the company’s own website and its regulatory announcements, with some filings and reference pages. All of them were confirmed by a second, independent pass. Company names open that company on the map. Where several pages on one site were checked, the extra pages are numbered after the site name. Source links open in a new tab.

Sites marked Secondary source are neither the company’s own website nor a regulatory news service or filing register: reference pages, market-data sites and trade press. They supported the check but are not the company’s own statement (53 of 467 source pages).

78 companies · 467 source pages

• Communication records are modelled. Use them to explore patterns and questions, not to judge a real company’s performance.
• Coverage is 78 companies and 315 tracked leaders, not every UK PLC or every director. A company’s real public output will include channels and people not shown here.
• Figures are fixed for the window ending 13 Sep 2026 (generator version 1). They change only when the dataset is regenerated.
• Identities can change after the Sep 2026 check. 9 recent or interim roles are marked with a note on the person’s panel.
• Engagement is a single modelled count per post. It does not separate reactions, comments and reposts, and it is not comparable with any platform’s own analytics.
• Market moves are context during the same session, never a measure of what an announcement or a post caused.
• Ask is a rule-based parser, not a language model. It handles questions about cadence, leadership, sectors, topics and announcements, shows what it applied, and declines anything else — including share-price forecasts and investment advice.