Methodology
HOW THE DEW POINT WORKS
The Dew Point is a single number (currently 34, ladder reviewed Sep 2026) that says how far humanity has climbed toward living on Mars, the way the Doomsday Clock says how close the world stands to catastrophe, pointed the other way. It is the weight of the rungs already climbed on a published ladder of28 milestones worth 100 points: 13 done,2 in progress. The weights are editorial and open to argument; the statuses are dated, sourced facts; the number moves only when a rung flips, and every flip is recorded on the scorecard. It will not reach 100 until people are living sustainably on Mars.
Where this stands today, stated plainly: the ladder's weights are editorial judgments and say so. The conviction scores are computed from the site's own feeds by the formula below, with the four sub-scores shown on every company page; a company whose execution cannot be measured keeps a reviewed editorial number, labelled as one. Quotes, market caps, launches and signals are live data with sources. We would rather show you an honest opinion than a fake calculation, and now that there is a calculation, we show its working.
The full ladder with every source, and the ledger of every move, is on the scorecard.
THE DEW PULSE
The Dew Pulse is the number of launches that have reached orbit, anywhere in the world, so far this calendar year, with the change against the same dates of the year before.
The Dew Point is the ladder: it moves only when a rung flips, a few times a year, and says how far we have climbed. The Dew Pulse is the heartbeat beside it: every launch that has reached orbit anywhere in the world since January 1, counted from Launch Library and refreshed every six hours. It climbs through the year and starts again at zero every January 1, like a season's tally. It is a measurement, not a judgement. It says how fast the world is getting to orbit, which is the floor every rung on the ladder stands on. Suborbital flights are not counted.
- A launch counts when Launch Library records it as a success, a failure or a partial failure, and its mission was orbital. Only successes are the headline number; attempts and failures are shown beneath it.
- Suborbital flights are not counted, so New Shepard passenger hops and Starship test flights are out. Starship's progress is scored on the Dew Point ladder instead.
- The window is the calendar year in UTC, from January 1 to the latest refresh. The comparison is the same stretch of the year before, January 1 to the same date and time, because a count that restarts every January can only be fairly compared with the same point in another year. The monthly bars show this year's complete months; the current month is named separately.
- It only climbs during the year. It can fall back only when Launch Library corrects a record, such as a launch first filed as a success and later reclassified. On January 1 it starts again at zero, and the full card shows where the year before finished.
- Every provider and every country counts equally. This is the world's cadence to orbit, not any one company's.
- It refreshes every six hours. If the two-year read from Launch Library fails, the last good figure is carried forward with its own timestamp, and the site's monitor nags after thirty hours so a stale pulse cannot pass for a live one.
- It is not an input to the Dew Point, and the world total is not an input to any conviction score. The same feed's per-provider counts do feed the launch lens of a covered launch company's score; those cover the trailing 365 days, not the calendar year, so a score does not reset every January.
- Anyone can check it: the feed is public at spacedew.com/data/launches.json, field flown.thisYear.successes.
Source: Launch Library 2 (The Space Devs). The live number, with its monthly bars and the year before, is on the scorecard and on the home page.
HOW A SCORE IS COMPUTED
Since September 8, 2026 each conviction score is computed: four sub-scores, each 0 to 100, read from the site's own feeds, frozen in a dated snapshot and weighted as shown. Every company page shows its four numbers, the evidence behind each, the arithmetic and the date. Today 31 of 50 scores are computed; the other 19 could not be measured and keep a reviewed editorial number, labelled as one.
The mean of the lenses that apply. Wins in the Feed: sourced items tagged as a win in 90 days, one per story, 3 wins is a full lens and an item tagged as a setback cancels one; neutral coverage alone leaves this lens unmeasured. Orbital launches flown in 365 days, 40 a year is a full lens on a log scale, scaled by the success rate. Revenue growth, the lower of the trailing year and the latest quarter, on a published curve that tops out at 80: growth alone cannot carry Execution to a full mark, a win or a launch can. A lens that does not apply (nothing directional in the Feed, not a launch provider, no public financials) is left out and named, never counted as zero.
Half from the ladder: the rungs the company led or took part in, weighted by the ladder's own weights, done in full, in progress at the ladder's credit, pending at 25% for a contracted programme (a part role counts half), on a square-root scale so being on the ladder at all shows. Half from its own timelines: the execution velocity, an editorial label that is dated and recorded on the scorecard when it changes.
50 is a quiet quarter. Each positive item adds 10 (up to 4), each neutral item 3 (up to 5), each negative item takes 15 (up to 3). An open-market purchase adds 6, a sale outside a 10b5-1 plan takes 3, within 18 either way. Form 4 is marked not applicable for a company that files none, which is not the same as no buying.
Breadth, 60%: one sector 50, two 75, three or more 100. Scarcity, 40%: fully thin at 2 covered names in a sector, not thin at 10, averaged over the company's sectors. The Polymath tag is not a sector.
Not inputs, on purpose: the share price (a score that moved with it would be a changelog of the tape) and analyst ratings (a different question, shown beside the score instead). The same contract win counts under Execution and under Signals, because the framework asks both questions of it.
Editorial judgement survives as an adjustment of at most 10 points either way, with a reason and a source, printed on the company page and recorded on the scorecard. None is in force today. A disagreement larger than that means an input or a mapping is wrong, and the fix belongs there.
When execution cannot be measured (nothing in the Feed tagged as a win or a setback this quarter, no orbital launches to count, no public revenue figure) the score is not computed and the reviewed editorial number stands, labelled. One tagged item in the Feed makes a name computable at the next snapshot. Today: Relativity Space, Stoke Space, Rocket Factory Augsburg, Oklo, Nano Nuclear Energy, Helion Energy, Rolls-Royce SMR, Axiom Space, Sierra Space, Thales Alenia Space, Varda Space, ispace, D-Orbit, The Metals Company, AstroForge, K2 Space, Impulse Space, True Anomaly, Astranis.
The record: every input is frozen in the site's repository with the date it was read, and the site cannot be built if a published score no longer follows from that record.
CONVICTION SCORES
HOW OFTEN SCORES ARE REVIEWED
Every 90 days. Earnings is the natural beat for a judgement about execution, and a quarter is short enough that a score cannot quietly go stale while being long enough that the record fills with decisions rather than share-price noise. A company is pulled forward when something happens that a quarterly rhythm would sleep through: a move of 30% or more, a cluster of insider filings, a new sourced signal, or a passed earnings date. A review means the snapshot is taken again: the inputs are read fresh, the formula runs, and every score that moves lands on the scorecard with its four sub-scores. The date beside each score is the date of its inputs.
A score older than 4 months is marked in amber next to the number, with its age. That is deliberate. An editorial judgement that has not been revisited should not look identical to one made this week.
When a score, signal or velocity does change, the change is recorded with its reason and a source on the scorecard. Entries are never edited after the fact; a correction is a new entry.
WHAT COUNTS AS A PURE PLAY
The screener's pure-play filter answers one question: Would this company still exist, in something like its current form, if space went away?
Some of these are judgement calls and reasonable people would move a name either way. The flag carries the date it was last reviewed, and a name nobody can classify with confidence is marked not-pure until a human decides rather than guessed at.
SIGNALS
Strong conviction: the company is executing well on the multiplanetary thesis. Fundamentals, catalysts, and execution all align.
Moderate conviction: worth watching but not adding. Mixed signals or waiting on a catalyst to confirm direction.
Low conviction: the thesis is weakening. Execution problems, dilution risk, or deteriorating fundamentals.