Metrics

Every number on this page is zero or absent today. That is the point of writing it down now. Demostatics has no users, no revenue, nothing deployed and no analytics, so the standard startup dashboard — MRR, DAU, churn, CAC — measures nothing and will keep measuring nothing for months. What this page defines instead is a set of leading indicators that move month to month before any customer exists, and that an incubator reviewer can check without taking anything on trust.

It is the companion to the report template, which is the format each report takes, and to the August 2026 baseline, which is the first report. This page is the measuring instrument; those two are the readings.

Where these numbers come from

Everything in the Baseline today columns below was measured on 2026-08-02 by running the command in the Source column against the working copies of the five repositories. No figure here is an estimate unless the word ESTIMATE appears next to it, and the two figures that could not be measured say so instead of guessing.

Re-measure, do not copy

The value of these metrics is entirely in the delta between reports. Copying a baseline forward because it “probably has not changed” destroys the only thing the page is for. Every number is a shell command; running all of them takes about ten minutes.

Group 1 — The data asset

This is the most important group on the page and the one that makes the company legible to an outsider. Demostatics is not, today, a product company with a slow start; it is a data company that has not begun accumulating its asset. Every other group measures work. This group measures worth.

The standout metric is days of history accumulated. It starts at zero, it increases by one every single day once collection starts whether or not a line of code ships that day, and it cannot be bought back later. A quantitative buyer’s first question about a data set is how far back it goes, and there is no answer to that question other than having waited.

Metric	Definition	Baseline today	Source of the number	Why it matters
Days of history accumulated	Whole days between the earliest object in the raw archive and today	0	Oldest key prefix in the archive store	Increases daily with no engineering effort; the only metric here that cannot be recovered by working harder later
Days of history, best-covered source	Same, for the single source with the longest unbroken record	0	Oldest key prefix under one source tag	A buyer evaluates the deepest series, not the average
Sources live	Distinct sources that wrote to the archive on their intended cadence in the last 24 hours	0 of 10 candidates	Count of distinct source tags with a fetch in the last 24h	Breadth is the second question after depth; the ten candidates are named on The Minimum Viable Data Set
Cumulative observations collected	Total rows in the observations table, never decreasing	0 — no table and no writer exist	SELECT count(*) once the table exists	The raw size of the asset; the number that goes in a data-room summary
Raw archive size	Bytes stored in the archive, all sources, all time	0 B	du -sb or the bucket’s own size metric	Proxy for asset scale, and the input to the storage line of the cost model
Collection uptime	Scheduled fetches that succeeded, as a percentage of scheduled fetches, last 30 days	Not measurable — nothing is scheduled	Fetch log	A gappy archive is worth much less than a complete one, and gaps are also unrecoverable
Plausibility pass rate	Observations passing range, monotonicity and unit checks, as a percentage of observations ingested	Not measurable — no gates and no observations	Quality-gate counters	Distinguishes a real data set from a pile of files; needed before anyone will pay for it

Every row in this group is zero, and that is the finding

The first non-zero value in this table is the single most meaningful progress signal available to this company. Not the first deployment, not the first test suite, not the first customer conversation — the first day of history. Report 2 either has a non-zero number in the first row or it does not, and everything else in it is secondary to that. This is the same conclusion What To Do Next reaches under “the history clock has not started”.

Group 2 — Product and engineering

The site already has a status vocabulary — Shipped / Partial / Planned / Not built — applied to every capability on the component pages, the roadmap and Where Things Stand. Rather than invent a second scheme, the metric is a count of those labels. It is grep-able, it is unambiguous, and it moves the moment something genuinely ships. It undercounts slightly, by design: a cell that mixes two statuses in a sentence — as the Reports row on the web component page does, where the page is Shipped and the producer behind it is Not built — is not counted at all. The command is the definition, and it is the same command every month.

Metric	Definition	Baseline today	Source of the number	Why it matters
Capabilities Shipped	Table cells whose entire value is Shipped, across the docs site excluding reports/	88	grep over src/content/docs, command below	The honest surface area of working software
Capabilities Partial	Same, for Partial	12	Same command	Work in flight; a rising number with a flat Shipped number is a warning
Capabilities Planned	Same, for Planned	61	Same command	The size of the promise still outstanding
Capabilities Not built	Same, for Not built	36	Same command	The gap between the pitch and the tree
Roadmap phases complete	Phases on Roadmap marked complete	0 of 11 — one Partial (Phase 0), ten Planned	Status column of the phase table	The plan’s own unit of progress
Gap register, open	Entries on Gap Register not marked closed	86 open, 0 closed	Summary table of the register	The list of things that get more expensive with time
Gap register by severity	Open entries split Existential / High / Medium / Low	5 / 42 / 35 / 4	Same table	Severity mix matters more than the total; closing four Low items is not progress
Gaps due Now still open	Register entries whose When is Now	17	Same table	These are the ones whose cost is rising today
Automated tests, web	Tests executed by the Laravel suite	211 passing, 922 assertions, 2.38 s	./vendor/bin/pest	The only suite on the platform that is actually run and green
Automated tests, desktop	#[test] plus #[tokio::test] functions in the Rust workspace	121 (115 + 6), across 18 crates, bin/ and examples/	grep over crates/ bin/ examples/	Counted statically, not executed — see the caution below
Automated tests, mobile	Jest tests executed	102 passing in 6 suites, 0.60 s	npx jest	Small but real and green
Repos with CI	Repositories with a working CI configuration	1 of 5 — only demostatics-pc_application	.github/workflows in each repo	Four repositories have no automated check of any kind on any commit
CI pass rate, desktop	Successful runs of ci.yml, last 20	4 of 20 green; the latest run on main is red	gh run list --workflow=ci.yml	A red default branch is a stronger signal than a test count
API endpoints	Routes registered under api/	45 (of 105 routes total)	php artisan route:list --path=api	The contract surface other clients depend on
Schema size	Migrations / Eloquent models / controller files in demostatics-web	22 / 23 / 36	ls and find, commands below	Cheap proxy for how much domain the display tier already models
Tracked lines of code	Tracked non-vendor, non-lockfile, non-binary lines across all five repos	53,853 — web 17,378, desktop 16,747, mobile 10,098, docs 9,628, backend 2	git ls-files piped to wc -l	Context for the test and CI numbers; not a goal in itself
Commits, all repos	git rev-list --count --all summed	114 — web 44, desktop 57, mobile 5, docs 7, backend 1	git rev-list	Activity, not achievement; useful only alongside the rows above
Docs pages	Markdown files under src/content/docs	41 at time of measurement; this page makes 42	find … -name '*.md' | wc -l	The documentation is currently the most complete artifact the company has
Environments deployed	Reachable environments serving the real application	0	Direct HTTP check, below	The gate in front of every metric that needs a running system

Two of these are weaker than they look

The desktop test figure is a count of test attributes found by grep, not a count of tests that passed — cargo test --workspace on this tree is slow enough that it was not run for this baseline, so 121 means “121 functions are annotated as tests”, nothing more. And demostatics.com currently resolves to 94.73.151.29 and returns HTTP 200 with the title “DemoStatics is coming soon”, over a TLS certificate that does not match the hostname; docs.demostatics.com has no DNS record at all despite being the configured canonical host. A placeholder page is not a deployment, which is why the environment count is zero.

Group 3 — Business and regulatory

Licensing is the longest-lead-time item in the company and it does not have a percentage. There is no such thing as 40% licensed. It is a ladder, and the only honest report is which rung the company is standing on.

Stage	Meaning	Reached
1. Not started	No counsel, no jurisdiction, no entity	Current position
2. Counsel engaged	A qualified adviser is retained and has the questions from Regulatory Posture	No
3. Jurisdiction decided	The regulator and regime are chosen in writing	No
4. Entity formed	A legal person exists that can hold a licence	No
5. Application prepared	Policies, capital plan and compliance function drafted to the regulator’s shape	No
6. Application filed	Submitted, clock running with the regulator	No
7. Licensed	Authorisation granted	No

Stage 1 of 7 is verified from Regulatory Posture, which states the position as “Not built, and not started” — no legal entity, no application, no compliance function, no counsel engagement. The metric for a report is the stage number and the date it last changed.

Metric	Definition	Baseline today	Source of the number	Why it matters
Licensing stage	Position on the seven-rung ladder above	1 of 7 — Not started	Regulatory Posture	Lines 2, 3 and 4 of the business cannot earn a lira until this moves; it is measured in months to years
Days since licensing stage last changed	Calendar days at the current rung	0 — the ladder is defined on this page today	Date arithmetic	A stage that has not moved in six months is the report’s most important sentence
Infrastructure spend, actual	Money actually paid for hosting, storage, APIs and inference last month	Not tracked — no billing artifact exists in any repository, so this could not be measured	Provider invoices	Without an actual there is no COGS, and without COGS there is no price
Infrastructure spend, estimated	The Scenario A figure from Cost and Funding	ESTIMATE $500–$5,000/month (Scenario B: ESTIMATE $50,000–$250,000/month)	Cost and Funding	The two scenarios differ by roughly 100x and describe two different companies
Actual as a share of estimate	Actual divided by the Scenario A midpoint	Not computable until an actual exists	Arithmetic	The point at which the cost model stops being a guess
Cost drivers with a real quote	Of three — LLM inference, managed database, object storage	0 of 3	Cost and Funding	Self-serve pricing pages; days of work that convert the whole cost table from estimate to fact
Cost scenario chosen	A or B, in writing	Undecided	Cost and Funding	Architecture, price and whether investors are needed at all follow from this one line
Runway, months	Cash divided by monthly burn	Not computed — no burn figure and no funding figure exist anywhere	Founder’s own accounts	The roadmap’s “9 to 18 months for one full-stack engineer” is stated purely in time and has never been priced
Paid dependencies with a spend cap	Paid services with a hard budget alarm configured	0 of 0 — no paid dependency exists yet	Provider consoles	The cap has to exist before the first API key does, not after the first bill

Group 4 — Commercial, pre-revenue

Revenue metrics do not apply and will not for months. These are the leading indicators that exist before money changes hands, and every one of them can be moved by a founder with a phone and no deployment.

Metric	Definition	Baseline today	Source of the number	Why it matters
Design-partner conversations, this month	Substantive calls with a named prospective user about their actual problem	0	A markdown log, one line per call	The only Group 4 metric that is fully under the founder’s control today
Design-partner conversations, cumulative	Same, all time	0	Same log	Ten conversations before writing ingestion code changes what gets ingested
Pilot agreements signed	Written agreements, paid or unpaid, to trial the product	0	Signed documents	The first external commitment of any kind
Expressed interest	Named people or organisations who have asked to be told when it launches	0 — no waitlist form exists	A markdown list, or a form once one is deployed	Cheapest possible demand signal; costs one afternoon to start collecting
Inbound enquiries	Unsolicited approaches about the product	0	Email	Distinguishes pull from push
Weekly active accounts	Distinct accounts with a session in the last 7 days	Not measurable — nothing is deployed	Application database, once deployed	The first metric that becomes real the week after Phase 0 lands
Paying accounts and MRR	Self-explanatory	0 — no billing code, no plans or subscriptions tables exist	Billing provider, once one exists	Listed here to be explicit that it is deliberately out of scope until Roadmap Phase 3

Group 4 is the cheapest group to move

Nothing in this table needs code. Ten design-partner conversations and a list of twenty names who want to be told when it launches are achievable before the next report, and they would make the second report qualitatively different from the first without a single commit.

How to collect these

Most of this is a shell command, not a dashboard. The commands below produce the Group 2 figures exactly as measured for this page, run from ~/Desktop with the five repositories checked out alongside each other.

# Commits per repo
for d in demostatics-web demostatics-pc_application demostatics-mobile_application \
         demostatics-docs demostatics-backend; do
  printf "%-34s " "$d"
  (cd "$d" && git rev-list --count --all)
done

# Tracked lines of code per repo, excluding vendor/node_modules/target/dist, lockfiles, binaries
for d in demostatics-web demostatics-pc_application demostatics-mobile_application \
         demostatics-docs demostatics-backend; do
  printf "%-34s " "$d"
  (cd "$d" && git ls-files \
     | grep -vE '(^|/)(vendor|node_modules|target|dist|build)/' \
     | grep -vE '(package-lock\.json|composer\.lock|Cargo\.lock|pnpm-lock\.yaml|yarn\.lock)$' \
     | grep -vE '\.(png|jpg|jpeg|gif|svg|ico|woff2?|ttf|otf|webp|pdf|zip|wav|mp4)$' \
     | tr '\n' '\0' | xargs -0 -r wc -l | tail -1)
done

# Capability status counts, the site's own vocabulary.
# reports/ is excluded: the reports quote these counts back, so including them
# would make the metric count itself.
cd demostatics-docs/src/content/docs
grep -rhoE '\| *\*{0,2}(Shipped|Partial|Planned|Not built)\*{0,2} *\|' --exclude-dir=reports . \
  | tr -d '|* ' | sort | uniq -c
# Shipped 88, Partial 12, Planned 61, Not built 36 — 197 total

# Docs page count
find . -name '*.md' | wc -l

# Web: tests, routes, schema
cd demostatics-web
./vendor/bin/pest                      # 211 passed (922 assertions), 2.38s
php artisan route:list --path=api      # 45 routes
ls database/migrations | wc -l         # 22
ls app/Models | wc -l                  # 23
find app/Http/Controllers -name '*.php' | wc -l   # 36

# Desktop: static test count (cargo test --workspace is slow; this is the cheap proxy)
cd ../demostatics-pc_application
grep -rE '^\s*#\[(test|tokio::test)' crates/ bin/ examples/ | wc -l   # 121

# Mobile: executed test count
cd ../demostatics-mobile_application
npx jest --silent                      # 102 passed in 6 suites

# CI health (only demostatics-pc_application has CI at all)
cd ../demostatics-pc_application
gh run list --workflow=ci.yml --limit 20 --json conclusion

# Environments: is anything actually serving the application?
curl -sSL -o /dev/null -w '%{http_code}\n' http://demostatics.com   # 200, but a placeholder page
getent hosts docs.demostatics.com                                    # no record

The two product tables deserve their own note, because they are the ones an outsider will assume contain data. reports and database_items in demostatics-web have no producer in production. The only code anywhere that writes to them is database/seeders/ReportSeeder.php, which inserts 12 demo rows, and database/seeders/DatabaseItemSeeder.php, which inserts 7 — 19 rows of hand-written demo content in total. Both are invoked only by DemoSeeder, never by DatabaseSeeder, so a plain php artisan migrate --seed creates none of them. The production row count for both tables is therefore zero and will stay zero until the worker tier exists. Once there is a database to point at, the metric is one query:

SELECT count(*) FROM reports;
SELECT count(*) FROM database_items;

Which metrics need something built first

Metric group	Blocked on	Available today
Group 1, all rows	The archiver — a cron job writing raw source files to object storage with a timestamp and a source tag	Nothing; this is the one thing worth unblocking first
Group 1, plausibility pass rate	Quality gates on top of the archiver	No
Group 2, all rows	Nothing	Yes, all of it, today
Group 3, licensing stage	Nothing — it is a judgement recorded by hand	Yes
Group 3, actual spend and runway	A billing account and a habit of recording the invoice each month	Partly; the estimate exists, the actual does not
Group 4, conversations, pilots, interest	Nothing — a text file and a phone	Yes
Group 4, weekly active accounts	A deployment and a session table	No

The cheapest possible version is a markdown file

Do not buy a BI tool, do not stand up Metabase, do not instrument analytics. Create reports/2026-09.md from the report template, paste the numbers in, and commit it. Twelve of those files is a year of trend data, it is diffable, it is reviewable in a pull request, and it costs nothing. Revisit this decision when the numbers are being read by someone other than the founder and the incubator.

What not to measure yet

Being deliberate about what is excluded is what keeps the list above short enough to actually run each month.

• Anything requiring users who do not exist. DAU, MAU, WAU, retention cohorts, session length, funnel conversion, churn, NPS, activation rate. Zero users produces zero of all of them, and a table of zeros crowds out the rows that are genuinely moving.
• Revenue and revenue derivatives. MRR, ARR, ARPU, CAC, LTV, payback period, net revenue retention. There is no billing code and no plans, subscriptions or invoices tables anywhere; first revenue is gated on Roadmap Phase 3 at the earliest.
• GitHub vanity. Stars, forks, watchers, contributors. This is a one-person private project; these measure attention the company is not seeking.
• Lines of code as a goal. The figure is in Group 2 as context for the test and CI numbers, not as a target. Deleting the desktop client’s mock-server coupling would reduce it and be an improvement.
• Velocity, story points, burndown. One engineer, no sprint boundary, no estimate history. The roadmap phase and the gap register already carry the sequencing.
• Uptime, latency percentiles, error rates, MTTR. All require a deployment. They become real the week Phase 0 lands, and instrumenting them earlier measures a laptop.
• Test coverage percentage. Reaching a trustworthy coverage figure across PHP, Rust and TypeScript means three separate toolchains wired into three CI setups, four of which do not exist. The absolute test counts and the CI pass rate carry the same signal for far less work.
• Model accuracy, precision, recall, F1. There is no model, no training data and no ground truth. These arrive with the observations, not before them.
• Anything that would take longer to instrument than the decision it informs is worth. The test for a candidate metric: if collecting it costs more than an hour a month, and no decision changes based on its value, it does not go on the page.

Next

The format each report takes is on the report template. The first reading of every metric above is recorded in the August 2026 baseline. The work that would make the second reading different from the first is sequenced on What To Do Next.