The numbers that
decide if Opcelerate
stays alive.
LTV, CAC, ARR, margins, churn, cash cycle, deferred revenue. Named work only. Bid prices labeled as bids. Books left blank where we have not measured them.
Opcelerate Neural Inc. · 201 Kaska Rd #143, T8A 2J6 · andres@opcelerateneural.com · (778) 488-8113 / (825) 459-3324
Three stamps.
Never mix them.
A bid is not a booking. A model is not a book. This plate refuses fake ARR.
Awards recorded as of 24 Aug 2026: 0. Submitted ≠ won. Never print $128,500 as ARR.
Price on a public tender. Not won until awarded.
Invoices, bank, ledger. If HQ has no export, it stays unmeasured.
Formula + the live ledger. You type, it computes.
Four named lines.
Nothing invented.
MSA Growth. Quote engine, pipeline, inventory, AI agent. sageprint.netlify.app
Human-in-the-loop tender hunt. Discovery, not a fake CMS.
Private-AI appliance. Hardware + model, on-prem feel.
Train the team. Closer to a course/retainer than a seat SaaS.
Bid floor $350,000 CAD. Prefer $1M+. Hourly card: $175 / $150 / $125 / $110 / $95. GST extra unless the buyer is GST-free.
The print shop is the
proof. Not a mock.
ON-MSA-2026-001 · Growth Plan · effective 25 Mar 2026 · Shaheen · hello@sageprint.ca · live at sageprint.netlify.app
Quote engine
5-step wizard, DPI check, file upload. 8 quotable products, 44+ in catalog.
Admin + pipeline
New → Quoted → Paid → In Progress → Done. Inventory, low-stock, Stripe, QuickBooks export.
AI agent Sage
Chat trained on products, pricing, file rules. Tokens on Opcelerate. Same-day bugs in the $350.
99.9%
Uptime target. Critical same business day. Two free upgrades/quarter. Code ownership after 24 paid months.
Hours we put in.
Cash that hasn’t come back.
| Line | CAD | |
|---|---|---|
| Deploy (Growth) + GST · INV-2026-001 · 27 Mar | $8,265.60 | Unpaid on PDF |
| Retainers Apr–Aug · 5 × $350 + GST | $1,837.50 | AR |
| 225 extra hours at $37.50 (half leftover; MSA is $75) | $8,859.38 | AR |
| INV-SP-2026-08 due on receipt (no deploy line) | $10,696.88 | AR |
Proposal frame 150–200 hours. Logged about 380–420. Leftover ~180–270. Other half of extra hours stays in the platform, not billed.
$16,272
Deploy $7,872 + $350 × 24 months, ex-GST, then they own the code. Plus extras if the Aug settlement clears. That is one logo. Churn here is non-payment, not a cancelled seat.
DSO is already five months on $350.
Do not count $350 × 12 as ARR while five invoices sit unpaid.
Cash in. AR sitting.
Not the command-center bar.
$8,880
ProShield × CK Condo kickoff, GST-included. Four invoices 25–28 Aug 2026, remaining $0. Owner confirms paid.
$10,697
Sage Print pack 28 Aug: five $350 retainers Apr–Aug + extra hours at $37.50. GST-in $10,696.88. Mar deploy $8,265.60 still unpaid on the PDF.
$128,500
Hardcoded “validated ARR” on command-center.html. No Stripe, QuickBooks, or invoice ledger behind it. Pitch, not books. Not ARR.
The message thread shows the dated PDF invoices for 25 and 26 Aug 2026.
The screenshot supports the billing trail; the $8,880 cash status remains sourced from the underlying book record.
ProShield remaining setup $21,721.60 (quarterly $2,715.20 from 15 Dec 2026) + $667.49/mo if they keep ops. Ads $513.45+/mo only if approved.
ARR is the subscription
that repeats.
Implementation is not ARR. A one-time website is not ARR. A three-year $126k software line is.
MRR × 12 = ARR
Do not annualize one-off professional services.
$126,000 / yr
HSS Ambient Cat 1 and RRC 2026-004 both bid a stand-alone software subscription at $126,000 per year, years 1–3, 3% escalation cap on RRC. Implementation sat beside it ($264k HSS / $359k RRC), which is not ARR.
Contracted, if they pay Sage $350 × 12 = $4,200/yr ex-GST (five months already AR). ProShield $667.49 × 12 = $8,010/yr GST-in, only if they keep the ops retainer. Combined ~$12,210 is not $1M and is not the $128,500 bar.
What a customer is
worth if they stay.
For Sage Print, LTV is the years of live AI plus expansion. For Academy, it is the course plus the next build. For a municipal roster, it is call-ups, not the RFSQ itself.
Better: discounted cash from that logo, including expansion.
Churn 0 → formula blows up. Use a finite horizon (5–7 yr) instead.
If one logo pays the $126k software line, 75% gross margin, 10% annual logo churn:
$945,000
= 126,000 × 0.75 / 0.10. That is a model. We have not measured churn, so we do not print this as fact.
What it costs to
win the logo.
For Opcelerate that is not Facebook ads. It is Andres’s hours writing bids, portal fees, travel, and any paid MERX add-on.
Include owner time at the rate card if you want an honest CAC.
A lost bid still sits in the numerator.
No HQ export of ad spend or hours-per-win. Honest range to start tracking:
- Bid pack labour: 20–60 hrs × $150 = $3k–$9k per pursuit
- Paid MERX add-on: parked unless you say buy
- Wins so far in this file: awards = 0 on Bonfire as of 24 Aug 2026
One target.
No actual yet.
LTV:CAC asks whether the lifetime gross profit expected from a new customer justifies the full cost to win that same customer cohort.
Compare the same cohort · count every pursuit, won and lost
Use the same tax basis on both sides.
3 : 1
Target at least $3 of modeled lifetime gross profit for every $1 spent to win the customer. Not reported performance.
Missing a complete pursuit-cost and win ledger, plus verified retention and gross margin by cohort.
What’s left after
the work to deliver.
Software subscription should be high-70s to 90s. Implementation and on-site days sit much lower. Mix matters more than a single percentage.
COGS: hosting, inference, HITL reviewers, implementation labour, appliance BOM.
Owner drawings are not COGS. Neither is bid time (that is CAC / opex).
- Software line $126k/yr — high margin if Canadian hosting stays lean
- Implementation $110–$175/hr — margin is rate minus the person-hour cost
- Extra on-site day $2,500 (RRC) — travel eats this fast
- Mac mini appliance — hardware BOM, not SaaS COGS
Book blended GM unmeasured.
Gross margin after
the costs that scale with a customer.
Support, success, payment fees, HITL hours on Neural Scout. If contribution is thin, growth makes you poorer.
Contribution margin % = contribution / revenue
Use this, not gross, when you decide whether to spend CAC.
Opcelerate variable stack to track:
- Inference + Canadian host
- HITL review time (Neural Scout)
- Academy live-session hours
- Warranty / 24×5 support we bid to HSS
Who leaves.
Count logos and dollars.
Logo churn is accounts. Gross dollar churn is revenue that walked. Public retainers churn at renewal, not monthly.
Gross dollar churn = lost ARR ÷ starting ARR
Monthly × 12 is a lie if you have three annual contracts. Use annual.
Sage Print is live — that is a retained logo until it isn’t. Bid awards this summer: none recorded. Do not print 0% churn on a company that still has almost no recurring base. Small n makes the percentage violent.
The same logos,
a year later, paying more.
NDR over 100% means expansion beat churn. That is how a tiny firm gets to $1M without a new logo every month.
World-class SaaS: 110–130%. Services shops often sit under 90% unless they land retainers.
Opcelerate expansion paths that are real:
- RRC-style 3% cap — price, not true expansion
- Call-ups off Strathcona / SCE PQR / AMSA if awarded
- Sage Print: more lines, more plants, not a second website
- Academy → private-AI appliance on the same logo
Months until the
new logo has paid for itself.
Use contribution, not revenue. A $9k bid pursuit against $126k × 75% contribution pays back in about a month of that software line — if you actually win and they pay.
Target: under 12 months for cash-tight. Under 18 is acceptable if NDR is strong.
~1.1 mo
on a $126k/yr line at 75% contribution. The risk is win-rate, not the math: unpaid bids sit in CAC forever.
How long cash is
stuck in the work.
Software with prepaid annual invoices can print a negative CCC (customers fund you). Government net-30/60 plus implementation in arrears does the opposite.
DSO days sales outstanding · DIO inventory (appliances) · DPO days payable
Negative CCC is a feature. Slow public AP is the default risk.
Opcelerate mix:
- Bids / PS: high DSO, DIO ≈ 0
- Academy prepaid: DSO low, CCC can go negative
- Mac mini: real inventory, DIO shows up
- HITL contractors: DPO if we pay them after the client pays
Days to collect.
You wrote ESO next to DPO.
In a cash-cycle stack that pair is DSO (days sales outstanding). If you meant employee stock options, that is equity dilution — not the cash cycle. Both are below.
Public bodies: budget 45–75 days, not 15.
ESO (equity): options outstanding × strike vs. current value. Separate book.
Sage AR $10,696.88 against a $350/mo retainer is brutal DSO: months of work sitting in paper. ProShield kickoff cleared in three days (25–28 Aug) — that is the DSO you want on every build deposit.
ESO / equity: Opcelerate Neural is owner-operated. No option pool on file. Treat ESO next to DPO as DSO plus this note. If you later grant ESO, it hits dilution and possibly IFRS 2 expense — not CCC.
How long we keep
other people’s cash.
Pay inference, contractors, and hardware suppliers after the client has paid, when the contract allows. Stretching DPO is free float. Stretching past terms is a reputation cost.
Raise DPO without breaking suppliers. Never hold payroll there.
Float we can actually use:
- Cloud / inference: card monthly, ~15–30 day DPO
- Hardware for Mac mini: deposit vs. net 15 from vendor
- HITL: pay on accepted output, after client invoice is out
Current assets minus
current liabilities.
Positive WC means you are funding the operation. Negative WC means customers and suppliers are. SaaS loves negative. Government implementation often forces positive (you work, they pay later).
Current: cash, AR, inventory, prepaid − AP, deferred revenue, short debt, GST owing
Today’s sketch: cash from ProShield $8,880 versus Sage AR $10,697. The build is funding Sage. Next $350k public implementation without a 30–40% deposit will look the same, at 50× scale.
Get paid before
you do the work.
Deposits, annual prepay, deferred revenue. That is the point of Academy and of a software line billed up front. It is also why a $359k implementation should never start at 0% deposit.
Bid term to steal: 30–40% on kickoff, milestone, holdback.
Capital $359k (RRC) or $264k (HSS) wants a deposit or it becomes a WC hole. The $126k annual wants invoice-on-anniversary in advance. Split the contract so the subscription cannot subsidize unpaid build hours.
Cash now.
Revenue later.
You invoice $126k on 1 September. Cash hits. Accrual books $10.5k/month. The rest sits as a liability — deferred revenue — and burns down as you deliver the year.
Each month: deferred ↓ , recognized revenue ↑
Deferred up + AR down = you are becoming more SaaS.
Flow to watch monthly:
- Opening deferred
- + new prepayments (Academy, annual software)
- − recognized (straight-line unless milestones)
- = closing deferred
If deferred is growing faster than burn, the company can look loss-making on accrual and still be fine on cash.
How many dollars
of cash to buy one dollar of new ARR.
Early, this will look ugly. One $126k win against a month of owner draw and unused inference can print a 2–4× multiple. After a $1M+ award it collapses.
Excellent < 1× · Fine < 1.5× · Repair > 2×
Net burn = cash out − cash in from operations (exclude financing)
Need two numbers from the bank: monthly net burn, and new recurring actually contracted. Until a win posts, the multiple is undefined (divide by ~0). Do not optimize it by booking implementation as ARR.
Two clocks. Use both.
When the bank moves
GST, payroll, inference, deposits. This is survival. A $417k AMSA award that pays on milestone 4 does not pay rent in month 1. Run the household here.
When it’s earned
Invoice sent, deferred burned down, AR aged. This is whether the business is any good. LTV, ARR, NDR, GM all live here. Run the bid machine here.
Never run the bid machine on cash-only: you will refuse slow government work that is the actual $1M path. Never run the household on accrual-only: you will run out of GST money.
Bid figures we can stand behind.
| Source | Line | CAD | Stamp |
|---|---|---|---|
| RRC 2026-004 | Year 1 total (capital $359k + annual $126k) | 485,000 | Bid |
| RRC 2026-004 | Annual software, 3% cap | 126,000 / yr | Bid |
| RRC 2026-004 | Extra on-site day | 2,500 / day | Bid |
| RRC 2026-004 | 5-year cumulative with 3% cap | 1,027,951 | Bid |
| HSS AHS-2025-2170 Cat 1 | Implementation PS | 264,000 | Bid |
| HSS AHS-2025-2170 Cat 1 | Software years 1–3 | 126,000 / yr | Bid |
| AMSA EAL AB-2026-05745 | Fixed price | 417,000 | Bid |
| Strathcona 26.0060 Att. 3 | Hourly seats | 175 / 150 / 125 / 110 / 95 | Bid |
| Calgary APA 25-1728 | Y1 one-time $417k + annual $126k | 543,000 | Bid |
| Saint John Energy SJE-RFP-2026-003 | Y1 excl. tax | 444,000 | Bid |
Awards recorded in the Bonfire dashboard as of 24 Aug 2026: 0. Submitted ≠ ARR. Hourly GST extra.
What we list. Not what we booked.
| Line | Price | Stamp |
|---|---|---|
| Academy private coaching | $50 / hour | List |
| Academy small group lab | $25 / person | List |
| Academy corporate sprint | from $2,500 / 4 hours | List |
| Mac mini private-AI setup | $3,000 + hardware at cost | List |
| Neural Scout standalone | $96,000 / yr ($48k add-on path) | List |
| Platform Starter / Growth / Ent. | $2,500 / $6,500 / from $15,000 per mo | List |
| Sage Print MSA | $7,872 deploy + $350/mo + $75/h extra | Contract |
| ProShield monthly ops | $667.49 / mo GST-in | Contract |
Type the books.
I’ll do the ratios.
Enter figures. Nothing is invented until you type.
Measure these six
every month.
Bank + GST owing. Survival.
Only the repeating lines.
Invoice to cash, by buyer.
Hours + fees per win.
After HITL and host.
Once net new ARR exists.
File: this HTML on the Mini. Update the ledger when Sage Print, Academy, or a bid actually clears. Do not decorate empty books.