2026-07 · thirty metrics in five sections, every one
with its definition on the page · synthetic data
Reporting period2026-07
SourceGoogle Sheet export (monthly_metrics.csv)
Rows19 months
ValidationPASS
Last refresh (UTC)2026-11-05 19:38
Source hashfb8bff7fc829
Management commentary — supplied by the preparer, not a
calculated figure
July added 1.01 million dollars of new ARR against 668 thousand of expansion, so growth this month came from new logos rather than the installed base - the reverse of the June pattern. Nine new logos closed and three customers left. Ending ARR is 39.8 million dollars.
Gross retention softened. Churn of 232 thousand and contraction of 145 thousand together exceeded expansion, so the installed base contributed negatively before new logos are counted. The board should read the retention line as the weaker half of an otherwise strong month.
A 180 thousand dollar inference overage true-up from one customer was billed in July and is excluded from ARR. It is real revenue and appears in the revenue line; it is not a subscription commitment and would misstate both ARR and NRR if treated as expansion. Sales Operations classified it as an upsell in the CRM and the classification has been corrected for board reporting under FIN-POL-021 section 2.1. Commission and retention reporting owned by Sales Operations still treats the amount as expansion, and that difference is known, documented, and expected to persist until their next reporting cycle - it is a definitional difference, not an error in either system.
AI cost as a percentage of revenue continued to fall, which is the margin story the board asked to see tracked each month. Cost per thousand tokens benefited from a committed-use rebate received this month.
Runway is reported on trailing three-month burn. A single strong collections month would flatter the figure if it were computed on the latest month alone.
Growth
Metric
Value
Definition — the one this company uses
ARR
$39.8M
Annualized run-rate of contracted recurring revenue at month end. Excludes one-time services.
Net new ARR (LTM)
$12.6M
Ending ARR minus ARR twelve months prior: new + expansion − contraction − churn.
ARR growth YoY
46.0%
Ending ARR over ARR one year ago, minus one. · ≥40% strong at this scale
Bookings (Q)
$5.6M
Total contract value signed this quarter (TCV). Leading indicator; not revenue.
Billings (Q)
$10.4M
Amounts invoiced: revenue + change in deferred revenue. The cash-collection precursor.
Pipeline coverage (next Q)
3.2x
Open pipeline over next quarter's new-ARR target. · ≥3.0x planning floor
Customer Health
Metric
Value
Definition — the one this company uses
NRR (LTM)
106.0%
Chain-linked monthly: (expansion − contraction − churn) ÷ beginning ARR, installed base only — new logos excluded. · ≥110% strong
GRR (LTM)
87.8%
Same base, but expansion cannot offset losses. Ceiling 100%. · ≥90% strong
Average revenue per account: ending ARR ÷ customer count.
Avg ACV (new, Q)
$101K
New ARR this quarter ÷ new logos: the annual value of a newly signed contract.
Efficiency
Metric
Value
Definition — the one this company uses
Burn multiple (LTM)
0.48
Net burn ÷ net new ARR: dollars burned per dollar of new recurring revenue. · <1.5 good, <1.0 strong
Rule of 40
28 pts
YoY ARR growth % + FCF margin %. · ≥40 strong
Magic number (Q)
0.92
Net new ARR this quarter ÷ prior quarter's S&M. ARR is already annual — not annualized again. · ≥0.75 lean in
CAC payback
21.8 mo
All S&M ÷ (new ARR × GM ÷ 12): months of gross profit to repay acquisition. Conservative: charges all S&M to new logos. · <18mo acceptable
LTV : CAC
9.7x
Formulaic (ARPA × GM ÷ churn) ÷ CAC per logo. Stated as formulaic; treat with the suspicion it deserves — see revenue-cohorts for the empirical version. · >3x conventional
ARR per FTE
$241K
Ending ARR ÷ headcount. · ≥$200K strong
Financial Performance
Metric
Value
Definition — the one this company uses
Cash
$25.5M
Month-end cash and equivalents.
Runway
49.6 mo
Cash ÷ trailing 3-month average net burn. · ≥18mo comfortable
Gross margin (LTM)
77.9%
(Revenue − COGS) ÷ revenue. COGS includes AI inference, infra, support. · ≥75% strong for AI-native
EBITDA margin (LTM)
-15.1%
Revenue − COGS − opex, ÷ revenue. Before capex, SBC treated as cash here.
FCF (LTM)
$-6.0M
EBITDA − capex. Negative = burning.
Net burn (t3, monthly)
$0.5M
Trailing 3-month average of −FCF. The quotable burn number.
AI-Specific
Metric
Value
Definition — the one this company uses
AI cost, % of revenue (LTM)
9.1%
Inference + model API spend ÷ revenue. The line that decides whether usage growth is good news. · watch the trend, not the level
GM after AI costs
77.9%
Gross margin with AI inference fully loaded in COGS — not adjusted out. · the honest GM
Cost per inference
$0.0375
Total AI cost ÷ inference calls, trailing quarter. · must fall as volume grows
Cost per 1K tokens
$0.0110
Total AI cost ÷ tokens processed, trailing quarter.
Tokens processed (mo)
25.9B
Total tokens through the platform this month.
AI cost trend (12mo)
-1.1%
Change in AI cost as % of revenue, first vs last quarter of the LTM window. Negative = margin expanding. · negative is the story
Why the definition column exists: NRR, magic number, and
CAC payback each have several defensible definitions, and a scorecard that
doesn't state which one it uses will quietly disagree with someone else's
deck within two quarters. LTV here is flagged as formulaic on the page
itself — the empirical version lives in the revenue-cohorts project.
July added 1.01 million dollars of new ARR against 668 thousand of expansion, so growth this month came from new logos rather than the installed base - the reverse of the June pattern. Nine new logos closed and three customers left. Ending ARR is 39.8 million dollars.
Gross retention softened. Churn of 232 thousand and contraction of 145 thousand together exceeded expansion, so the installed base contributed negatively before new logos are counted. The board should read the retention line as the weaker half of an otherwise strong month.
A 180 thousand dollar inference overage true-up from one customer was billed in July and is excluded from ARR. It is real revenue and appears in the revenue line; it is not a subscription commitment and would misstate both ARR and NRR if treated as expansion. Sales Operations classified it as an upsell in the CRM and the classification has been corrected for board reporting under FIN-POL-021 section 2.1. Commission and retention reporting owned by Sales Operations still treats the amount as expansion, and that difference is known, documented, and expected to persist until their next reporting cycle - it is a definitional difference, not an error in either system.
AI cost as a percentage of revenue continued to fall, which is the margin story the board asked to see tracked each month. Cost per thousand tokens benefited from a committed-use rebate received this month.
Runway is reported on trailing three-month burn. A single strong collections month would flatter the figure if it were computed on the latest month alone.