River
Y CombinatorBacked by Y Combinator

InvestorsFree

Founder Update Analysis for Investors

Send this month's update and the prior months, and get back every metric that stopped being reported and what it would now be.

Start here

Tarnbrook Systems has sent an investor update every month for fourteen months. The August one reports ARR of 6.42 million dollars, up 4.9 per cent on July and 92 per cent on the year, cash of 4.68 million, and headcount of 88. Nothing in it is untrue. Read on its own it is a good month. Read against the thirteen updates before it, eleven of the eighteen metrics this company has reported at some point are no longer in it.

Two of the eleven can be recomputed rather than guessed, because their inputs are still being printed. The March update gave cash of 8.41 million and net burn of 560,000 a month. Five months later cash is 4.68 million and burn is not stated. Add back the 500,000 tax credit the June update mentioned in prose and the company spent 846,000 a month, 51 per cent above the last figure it published. Headcount went from 62 to 88 over the same five months, which is 11,000 a month per added head.

Runway follows. At 846,000 a month the 4.68 million runs out on 15 February 2027. The last runway figure Tarnbrook ever printed was 15.0 months as of 31 March 2026, which pointed at 30 June 2027, 135 days later. Public filers are held to something close to this. Material information left out of a report must be added when its absence makes what was said misleading. A metric whose calculation changes between periods carries its own disclosure. A private update carries neither obligation, so the portfolio-level version has to reconstruct it.

Eleven metrics that stopped, five of them in the same two months

Fourteen updates by eighteen metrics is 252 cells, and 142 of them are filled. Seven metrics are in the latest update and eleven are not. The median silent one was last seen eight months ago. Five of the eleven went dark in the same two-month window, November and December 2025: net dollar retention, gross margin, CAC payback, sales cycle and pipeline coverage. Weekly active accounts appeared in January. Metrics do not go quiet one at a time at random, and a cluster like that is the finding.

Each silent metric with three or more points gets its own trend extended to the current month, then checked against the watchlist trigger you set in advance. Eight of the eleven have one. All eight were on the safe side when last printed. Seven now project past it: retention to 86 per cent against a 100 floor, gross margin to 57 against 65, churn to 2.1 per cent a month against 1.5. The eighth, net promoter score, projects the right way.

A projection is a question, not a finding, and it says so on the row. Pipeline coverage extends to 0.4 times, which cannot be true of a company still growing, so the trend broke and the update does not say when. The composition test is firmer. Customers were last printed at 214 in February against ARR of 4.72 million, or 22,056 dollars each. ARR is now 6.42 million and the prose claims 31 new logos, which puts 60 to 84 per cent of the increase somewhere other than new customers.

How it works

  1. Paste the updates

    This month's update and as many prior months as you have, oldest first, in one block.

  2. Build the ledger

    Every metric named in any update becomes a row, every update a column, and blanks get dates.

  3. Recompute, then project

    Anything still derivable from reported inputs is recomputed. The rest carries its own trend, labelled as projected.

  4. Write the questions

    One question per silent metric, naming the month it stopped and the figure you are asking for.

What you get

  • A ledger of every metric ever reported against every update, with first seen and last seen
  • Months silent per metric, counted from its own last appearance rather than from the newest update
  • Burn and runway recomputed from two cash balances the updates printed, not from a trend
  • Each silent metric's own last three points extended forward, labelled projected on every row
  • Projections checked against the watchlist triggers you set, so a crossing is flagged by rule
  • Metrics that went dark in the same window grouped, because they rarely stop one at a time
  • The questions to send, each naming one metric, the month it stopped and the value asked for

Common questions

What if I only have this month's update?

Then the run reports what the single update states and says the continuity ledger is empty, which is honest rather than useful. The mechanism needs at least three prior updates to build a trend, and six to make the silence meaningful. Forward the thread. Most investors already have every update sitting in one mail folder, which is all the history this needs.

Is a projected value not just a guess?

It is an extrapolation of the metric's own last three reported points, labelled as one on every row, and it is there to make the question precise rather than to stand in for an answer. Asking what net dollar retention is now gets a vague reply. Asking whether it is near 86 per cent gets a number or a refusal, and both are informative.

Does it say the founder is hiding something?

No, and it is written not to. Every output is a statement about the documents: this metric appeared in these months and not in those. A reporting standard nobody set, a finance hire who changed the template, and a bad quarter all produce the same gap. Which of the three it is comes out of asking, not out of the ledger.

How is this different from a metrics dashboard?

A dashboard shows the metrics that are still arriving, which is the set this looks at last. Its columns are fixed, so a metric that stopped being sent shows as a blank cell rather than as an event with a date. The same test runs across a whole portfolio in the portfolio monitoring pack, and against board minutes in board meeting prep.

The update announced a bridge round. Does that break the cash bridge?

Yes, and the bridge has to be adjusted for it. Any financing, credit drawdown or one-off receipt inside the window is an inflow, and the derived burn is wrong without it. Tarnbrook had one, a 500,000 tax credit mentioned in prose, and it moved the cash-out date 22 days. A round also changes what you own, which the waterfall settles.

Where does this sit relative to the diligence we did at entry?

The definitions were set then, in the investment committee memo, which is the only record of what ARR meant when the check was written. The deck that arrived before it is pitch deck analysis, and turning a target's own statements into an earnings figure is add-back review.

What comes back, and in what format?

A document and a sheet. The document lists what changed this month, what disappeared and when, and the questions to send. The sheet is the ledger, one row per metric per update, with first seen, last seen, months silent and the projected value. Word and CSV, so both open in Excel, Numbers, Sheets, Word, Pages and Google Docs.

Founder Update Analysis for Investors

Fill in the form and your workspace opens with the work already underway.