Diagram demand index: deeper use, not broader reach, on ChatDiagram

The short answer: ChatDiagram had fewer artifact-producing accounts but deeper use during the 30 full Pacific Time days ending September 6, 2026. The cohort fell 18.8% from the previous 30-day window, from 9,250 accounts to 7,511, while saved artifacts rose 4.4%, from 22,027 to 23,004. That lifted average output from 2.38 to 3.06 artifacts per account.

Flowcharts still reached the most people: 1,466 accounts saved 4,776 flowchart artifacts. Circuit diagrams were second at 1,051 accounts, followed by general SVG diagrams at 712 and single-line diagrams at 680. This is a ranking of activity on ChatDiagram, not a claim about worldwide diagram use.

Observation window: August 8, 2026 00:00 through September 7, 2026 00:00 America/Los_Angeles, covering 30 complete days and excluding the partial update day. The comparison window was July 9 through August 7 on the same basis. The query ran September 7, 2026 at 03:12 Pacific Time. The internal account [email protected] was excluded. No prompt, diagram content, title, or other customer content was selected or returned; email was evaluated only inside the database exclusion predicate and was never returned.

The 15 most-used diagram types

“User reach” is the share of all 7,511 artifact-producing accounts that saved at least one artifact of that type during the window. “Previous reach” uses the same rule for the 9,250 accounts in the previous window. The last column is the percentage-point change, which controls for the different cohort sizes.

The aggregate CSV contains all 44 types that met the displayed cohort rule of at least 20 unique accounts. Counts below that threshold remain in the database but are not used for type-level conclusions here.

Site plans gained reach, but search did not explain it

Site plan was the largest positive reach-share mover: 181 accounts saved 667 site-plan artifacts, compared with 77 accounts and 223 artifacts in the previous 30 days. Its reach rose from 0.83% to 2.41%. That is an observation about the current ChatDiagram product and cohort, not evidence that worldwide site-plan demand grew by the same amount. A renderer launch, routing change, template coverage, acquisition mix, or repeat behavior could move the count.

Google Search Console gives the necessary cross-check and does not show matching conversion behavior. In the complete August 30–September 5 Pacific Time window, query strings containing site plan or site-plan produced 1,936 impressions and 28 clicks for ChatDiagram pages; August 23–29 produced 1,522 impressions and 47 clicks. Exposure rose while clicks fell, so this study does not claim that search caused the product shift.

The same separation matters elsewhere. Flowchart reach rose 1.35 percentage points even as flowchart-family search impressions fell from 10,050 to 8,468 and clicks fell from 94 to 73. Circuit reach fell 0.47 points while circuit-family impressions stayed nearly flat at 3,349 versus 3,394 and clicks rose from 89 to 120. Single-line reach rose 1.11 points while its search family moved from 1,446 impressions and 48 clicks to 1,477 impressions and 73 clicks. Product usage and search visibility answer different questions; neither is a proxy for the other.

Flowcharts widened their lead to 19.52% reach. Block diagrams posted the largest decline among the leading formats, from 8.62% to 6.34%, while general SVG fell from 11.11% to 9.48%. These are prioritization signals for templates, documentation, and renderer QA—not market-share estimates or evidence of a causal product change.

Artifact volume measures iteration, not just reach

The window contained 4.4% more saved artifacts than the previous 30 days, even though the number of artifact-producing accounts was 18.8% lower. Average artifacts per account therefore rose 28.6%, from 2.38 to 3.06.

That gap is why this index leads with unique accounts and publishes medians in the CSV. Network diagrams illustrate it: 170 accounts saved 1,208 artifacts, or 7.11 artifacts per account on average, but the median network account saved two and the 75th percentile saved 3.75. Genogram accounts averaged 5.31 artifacts, while their median was two. Circuit and flowchart both had medians of one. A few deep sessions can lift the mean without indicating that the format reached more people.

These counts do not measure quality. An artifact row records a saved diagram payload keyed to a chat and message. It is not a render-success event, approval, export, or completed engineering deliverable. Client hard-render errors live in a different event table with a different denominator, so this study does not divide error events by saved artifacts or label the result a failure rate.

Worked example: one pump project, five diagram jobs

Suppose a team is adding a duty/standby pump set to a process area. “Make a pump diagram” is not enough to choose a format. Start with the reader’s decision:

1. Explain material movement: use a process flow diagram showing major equipment and streams.
2. Define piping and control intent: use a P&ID with valves, tags, signals, alarms, and equipment boundaries. ISA’s current ISA-5 series provides a common framework for instrumentation and control documentation.
3. Show electrical distribution: use a single-line diagram for sources, protective devices, feeders, and loads.
4. Show motor-control wiring: use a circuit or breadboard-style drawing only if that representation matches the actual review job.
5. Define controller behavior: use ladder diagram or function block diagram when the PLC implementation requires it; IEC 61131-3:2025 specifies those graphical languages.

BPMN is different again: it is for a business-process contract, not equipment connectivity. The OMG BPMN 2.0.2 specification defines its process notation. Popularity cannot make one of these formats substitute safely for another.

How the index was reproduced

The query used artifact metadata only: timestamp, opaque user and chat identifiers, artifact type, and engine label. Engine prefixes such as schematex-, echarts-, and econ- were removed to form display labels. The query joined users only to exclude the internal account; it did not return any email value.

with normalized as (
  select
    regexp_replace(
      coalesce(nullif(a.engine, ''), nullif(a.type, ''), 'unknown'),
      '^(schematex-|echarts-|econ-)',
      '',
      'i'
    ) as diagram_type,
    a.id,
    a.user_id,
    a.chat_id
  from public.artifacts a
  left join public.users u on u.id = a.user_id
  where a.created_at >= $1::timestamptz
    and a.created_at < $2::timestamptz
    and coalesce(lower(u.email), '') <> '[email protected]'
)
select
  diagram_type,
  count(*) as artifacts,
  count(distinct user_id) as unique_users,
  count(distinct chat_id) as chats
from normalized
group by diagram_type
having count(distinct user_id) >= 20
order by unique_users desc, artifacts desc;

The previous-window comparison repeats the query for July 9 00:00 through August 8 00:00 Pacific Time. User reach divides each type’s unique-account count by all artifact-producing accounts in that window. The type shares overlap because one account can create several diagram types.

The GSC cross-check queried the Search Console query dimension separately for August 30–September 5 and August 23–29, 2026. Case-insensitive families were site plan / site-plan, circuit, flowchart / flow chart, and single line / single-line / one line / one-line / sld; impressions and clicks were summed within each family. GSC reports search activity, not accounts or saved artifacts, so no rate combines the two sources.

Limitations and next step

• Accounts are the privacy-safe proxy for users; one person with multiple accounts can be counted more than once.
• The index observes saved artifacts, not every request, abandoned generation, export, or downstream use.
• Renderer launches, retirements, routing changes, and acquisition mix can change the available opportunity for a type.
• Engine labels are ChatDiagram’s operating taxonomy. They do not replace definitions in applicable standards or project specifications.
• The cohort threshold reduces noisy small-sample claims, but it does not make the ranking representative of diagram use outside ChatDiagram.

Use the index to decide what deserves better templates, documentation, and reliability work—not to choose a notation by popularity. For an actual project, write down the reader, the decision, the required entities, and the review authority first. Then use ChatDiagram’s online diagramming workspace to draft the format that matches that job and verify the result against the applicable standard and project rules.