We've cut our peak demand almost in half. It's still more than triple what we have.
This is our concurrency plan for BF/CM 2026: the demand we're facing, every guardrail we're putting up to shrink it, and the honest number we're left with. The short version is on the right, in checkout lanes.
Peak demand, no action
88
lanes needed at peak · 5.9× our limit
After every guardrail →
~50
the most we can shrink it to
What we actually have
15
concurrency lanes provisioned
Think of a concurrency lane like a checkout lane: each one rings up about 20 orders a minute, so 15 lanes handle ~300 orders/min. On Black Friday, orders arrive far faster than that, and the ones that can't get a lane queue — late shipments, marketplace penalties, oversells, cancellations. We've done everything possible to send fewer shoppers to the registers. We still have ~50 lanes' worth of rush and 15 registers open.
Left alone, BF/CM 2026 would demand about 88 lanes at peak. Here's every lever we're pulling to bring that down — and where it lands. The goal isn't to fix the shortage with settings; it's to prove we've done everything possible before asking for more lanes.
From 88 lanes of demand down to ~50
Each step is a real change we control · the dashed line is what we have (15)
What the guardrails buy: capping Amazon FBA (Amazon ships it, nothing waits on us) and parking Global (Pull Doc) (sampling orders, off that week) remove almost half the demand outright. B2B is throttled for the week because those orders arrive weeks early. That's the ~40 lanes of intake we can't avoid, plus the loop tax below.
Why it can't go below ~50: the remaining B2C + TikTok orders don't cost one lane each — they travel a round trip to the warehouse and back, and the warehouse export leg competes with new orders coming in for the same lanes. That coupling is the +11 on the chart. See The loop.
One B2C or TikTok order isn't one draw on our lanes — it's a round trip. Both channels run through the same Celigo / integrator.io connector, and every order has to go out to the warehouse and come back before the customer is served. Four legs, sharing 15 lanes.
The order loop: Shopify ↔ NetSuite ↔ warehouse
Blue = Celigo real-time · purple = ADS warehouse batch (every 2 hours, 24/7 at peak)
Celigo — real-time, always on ADS warehouse — batched every 2h, holds lanes until done where they collide
What broke in 2024: the ADS batch fires, grabs almost every lane to clear its backlog, and holds them until it finishes — so Celigo's real-time order import slowed to a crawl waiting for a lane. The bigger the backlog, the longer ADS holds on, the longer new orders starve. In 2025 the two "battled more fairly," but the whole system slowed. That collision, not raw volume, is the failure mode.
Two ways out, and we're doing the settings one: (1) de-conflict the flows — give Celigo import a floor it can always reach, and/or run ADS more often in smaller bites so it grabs less per burst; (2) add lanes so there's enough for both. Setting changes soften the collision; they don't end it while two flows share 15 lanes.
We have run every Black Friday on these 15 lanes. 2024 and 2025 were the two worst peaks we've had — and the reason is simple: demand was already 2–3× what 15 lanes can serve. We survived, but "survived" meant queues, slow imports, and a system-wide crawl. Built from actual November order volume.
Peak lane demand by year vs the 15 we have
Estimated peak lanes from November order actuals · dashed line = 15-lane limit
How this is estimated: peak lanes are scaled from one measured point — June 2026 hit 41 concurrent on ~35k monthly orders. Prior Novembers are projected from that ratio, so treat them as directional. The pattern is not directional: every Black Friday has run well over 15, and each has been worse than the last.
Here are the clean numbers for the two channels that actually matter at peak — B2C and TikTok through the Celigo/ADS loop — after every guardrail is in place.
The honest verdict
Even after everything we can do, 2026 is on track to be our worst Black Friday yet.
After capping FBA and B2B, parking Global (Pull Doc), throttling the batch integrations, and tuning the loop, peak demand still lands near ~50 lanes. That's higher than 2025's ~48 — the worst peak we've ever run — against the same 15 lanes.
The guardrails are the proof we did our part. They are not a substitute for capacity: no arrangement of 15 lanes serves a 50-lane peak. It only decides which orders queue first.
2024 → 2025 → 2026 (after guardrails) vs 15 lanes
Peak lane demand · 2026 is the mitigated number, not the raw 88
The ask
Approve added concurrency now — on the February concession, not at Black Friday rush.
To clear the ~50-lane loop peak with any headroom for the Celigo/ADS collision, we need to move from 15 lanes to the 45–50 range. NetSuite sells this as SuiteCloud Plus licenses (and, above 45, the Enterprise tier). Here are the three quoted options, mapped to the lanes each one actually buys.
still short
Option 1
+2 SuiteCloud Plus licenses (Premium tier)
35 lanes
below the ~50 peak — still queues
$44k/yr
after −$10k Feb concession
recommended
Option 2
+3 SuiteCloud Plus licenses (Premium tier)
45 lanes
meets the loop peak · thin headroom
$56k/yr
after −$25k Feb concession
most headroom
Option 3
Enterprise tier +3 SuiteCloud Plus licenses
50 lanes
clears peak + processor headroom
$79k/yr
after −$103k Feb concession
Lanes per option are read from NetSuite's SuiteCloud scaling table, assuming we're on the Premium base today (15 lanes at 0 licenses). Option 1 = Premium+2 = 35 · Option 2 = Premium+3 = 45 · Option 3 = Enterprise+3 = 50. Prices are the quoted February EOQ-concession figures and are budgetary, subject to final NetSuite executive approval — confirm the resulting tier with our rep before signing.
Why the timing is the whole decision
The same lanes cost far more — and arrive too late — if we wait and buy them during peak
Approve now (Feb concession)
$56k–$79k / yr
Locks the concession pricing. Lanes are provisioned, tested, and stable weeks before peak. Guardrails + capacity both in place.
Buy reactively at Black Friday
~$105k–$238k / yr
Rush upgrade adds ~30%, and the Feb concession is gone. Worse: provisioning can take up to 3 days to hit our account — by then the queue has already formed, and orders don't un-queue until the backlog clears the following week. You pay more for lanes that arrive after the damage is done.
The recommendation: approve Option 2 (45 lanes, $56k) at minimum, or Option 3 (50 lanes, $79k) for real headroom against the loop collision. Decide on the concession timeline — not in a war room on Nov 27. This is capacity planning, not an emergency purchase; treat the number as decision-support and confirm final pricing and resulting tier with our NetSuite rep.
Volume by channel, so nothing's hidden. B2C and TikTok are the peak-week story (same connector, on the 3-day clock). FBA and B2B are capped for the week; Global (Pull Doc) is parked. Monthly order counts, Jan 2024 – Jul 2026.
Jul 2026 is a partial month (data as of Jul 13).
Peak-week status: B2C & TikTok — protected, floating. FBA & B2B — capped to 1 (Nov 24–Dec 1), removed after. Global (Pull Doc) — not running that week. "Marketing/Other" (<0.1%) is in totals only.