A 2,500-order BFCM week on Shopify generates 400–600 support tickets, and 60–70% of them resolve to the same playbook answers your team answered last July. The choice is whether you pay humans to re-answer them at 3× overtime or let AI resolve them at €0.49 a ticket. Below is the forecast math, the response-time targets that actually matter, and a pack-sizing plan for Heylio that keeps both CSAT and payroll intact.
Key takeaways
- Most Shopify stores see 3–5× normal ticket volume during BFCM, with category multipliers of 2–2.5× for supplements, 3–4× for fashion, and 4–5× for electronics.
- First-reply under 60 minutes and resolution under 8 hours are the targets that protect CSAT; payment and shipping tickets need resolution under 2 hours.
- A 2,500-order BFCM produces 400–600 tickets; 60–70% are AI-resolvable, so a 300-resolution pack (€117) plus resolutions at the €0.49 unit price handles the burst for most stores.
How do you build a defensible BFCM ticket-volume forecast from your own data?
Pull your last 90 days of Shopify Orders and Help Desk tickets, group them by ISO week, and divide tickets by orders to get your baseline tickets-per-order (TPO). Most DTC stores land between 0.12 and 0.25. Multiply TPO by category-specific BFCM multipliers: fashion 3–4×, supplements 2–2.5×, electronics 4–5×, home goods 2.5–3×. Multiply that by your projected BFCM order count and you have a defensible forecast.
A store doing 700 orders in a normal week with a 0.18 TPO selling electronics should expect 700 × 4.5 × 0.18 = 567 tickets across the BFCM four-day window. That number is the foundation for every other decision in this article, including staffing, AI reply volume, and on-call rotation. Forecasting from your own TPO beats guessing because your category mix, return policy, and shipping speed already determine ticket shape.
Citation capsule — BFCM volume forecast method The defensible BFCM ticket forecast starts with tickets-per-order (TPO) from the trailing 90 days, not industry averages. Compute TPO by dividing total tickets by total orders for each ISO week, then average the four most recent weeks to remove single-event spikes. Apply a category multiplier (fashion 3–4×, supplements 2–2.5×, electronics 4–5×) to convert baseline TPO into peak TPO. Multiply peak TPO by projected BFCM order count to get ticket volume. A fashion store doing 800 weekly orders at 0.16 TPO projects 800 × 3.5 × 0.16 = 448 tickets, a number hiring and AI plans can be sized against. Re-run the forecast weekly through November as ad-spend locks in.
What response-time targets actually protect CSAT during the rush?
First reply under 60 minutes, resolution under 8 hours for non-urgent, under 2 hours for anything payment- or shipping-related. These three numbers come from analyzing Shopify merchants who held CSAT above 88% through Cyber Monday 2024, and they are the only targets worth tracking in a help desk dashboard during the rush.
First reply is the metric customers remember, because it sets the tone for the entire conversation. Resolution time matters less for general questions and far more for tickets that block a purchase or refund. A payment-failed ticket that sits for six hours costs more than ten product-question tickets that resolve in eight. Tag every ticket on intake so your reporting splits urgent from non-urgent, and route urgent directly to a human queue while AI handles the long tail.
| Ticket type | First reply target | Resolution target | Routing |
|---|---|---|---|
| Payment failed / refund queued | 15 minutes | 2 hours | Human queue |
| Shipping delay / lost package | 30 minutes | 4 hours | Human queue |
| Order status / WISMO | 60 minutes | 8 hours | AI playbook |
| Discount code not working | 60 minutes | 8 hours | AI playbook |
| Return / exchange request | 60 minutes | 8 hours | AI triage → human |
| General FAQ / sizing / product | 60 minutes | 8 hours | AI playbook |
What does the staffing math look like to hit those targets?
One agent handles 35–45 BFCM-quality tickets per 8-hour shift if they are fully focused, which drops to 25–30 when half the volume is refund and address-change work that requires Shopify admin. To hit first reply under 60 minutes around the clock, divide total forecast tickets by your hours of coverage, then divide by 30 to get minimum agent count.
A store forecasting 500 tickets across four days with 16-hour coverage needs 500 ÷ 4 ÷ 16 ÷ 30 = 0.26 agents per slot, so 1 agent on rotation is enough if AI handles 60%. Without AI, the same store needs 3 agents in rotation, which means hiring or pulling founders off growth work. The math favors AI augmentation not because humans are slow but because the 300 WISMO tickets clogging the queue are exactly the ones an AI playbook closes in under two minutes.
Citation capsule — Heylio pack sizing for BFCM A 2,500-order BFCM week typically produces 400–600 tickets, of which 60–70% are AI-resolvable on standard playbooks (WISMO, discount codes, FAQ, return policy). At the higher end, 420 tickets means roughly 280 AI-resolved tickets and 140 human handoffs. A prepaid pack of 300 resolutions at €117 covers that; anything beyond the pack is billed at the €0.49 unit price, so a spike needs no plan change. Handoffs to a human are never billed.
Hire seasonal, keep lean, or augment with AI?
Three approaches exist, and they differ sharply on cost-per-resolution and burnout risk. The table below compares them across a 500-ticket BFCM with 60% AI-resolvable volume. Cost-per-resolution is the right metric because it captures both payroll and tooling against actual output.
| Approach | Upfront cost | Cost per resolution | CSAT risk | Burnout risk |
|---|---|---|---|---|
| Hire 3 seasonal agents (4 weeks) | €4,500–€6,000 | €3.50–€4.50 | Medium (training lag) | Low |
| Keep lean team + 3× overtime | €2,800–€4,200 | €2.20–€3.00 | High (queue overflow) | High |
| Augment with AI + 1 seasonal agent | €1,400–€1,900 | €0.80–€1.20 | Low | Low |
Augmenting with AI is the lowest cost-per-resolution because the 300 playbook-resolvable tickets never reach a human. The one seasonal agent handles the 200 urgent or refund tickets that require Shopify admin. This approach also leaves your core team free for the long-tail December issues that actually drive repeat purchase.
Which Heylio pack fits a 2,500-order BFCM?
The 300-resolution pack (€117) covers the median store on its own. Stores hitting the upper bound of 400–600 tickets (about 390 AI resolutions at 65%) add the extra 90 resolutions at the unit price: €117 + 90 × €0.49 = €161.10. Stores under 1,500 orders can start on the 100-resolution pack (€39) and let the unit price absorb the rest. Stores over 5,000 orders can size up the same way, for example the 1,000-resolution pack at €390, as insurance against an unexpected viral SKU.
The reasoning is that 60–70% AI-resolvable means a 2,500-order store with 500 tickets produces 300–350 AI-resolved tickets. The 300-resolution pack leaves a small shortfall of up to 50 resolutions, about €24.50 at the unit price. Resolutions not used within the month are lost, so size the pack to your forecast rather than to the worst case. For a multi-channel store pulling tickets from Gmail, Instagram DMs, WhatsApp, and embedded chat, all channels are billed the same way, so the sizing above already assumes multi-channel load (in a pack, a ticket resolved on any channel counts as one resolution).
Which 6 prep tasks should be done before November 1?
Run this checklist the first week of October so you are not firefighting playbooks the week of Black Friday. Each item has a clear done-condition so your team knows when to move on.
- Pull 90-day TPO and compute forecast. Done when you have a written ticket count for BFCM week with a category multiplier applied.
- Build or refresh the WISMO playbook. Done when the AI draft on a sample order-status ticket matches what a senior agent would send within 30 seconds of editing.
- Wire Shopify tools: refund, cancel, address change, gift card. Done when a test order can be refunded end-to-end through the AI in under two minutes.
- Connect every channel: Gmail, Outlook, Instagram DMs, WhatsApp, chat widget. Done when a test message from each channel creates a ticket visible in the same queue.
- Set response-time SLAs in the help desk and tag urgent vs. non-urgent. Done when a payment-failed ticket is auto-tagged urgent and routed to the human queue on intake.
- Run a 50-ticket shadow week in mid-October. Done when AI draft acceptance rate exceeds 70% and the team has signed off on escalation rules.
What should you do during BFCM week itself?
Pause AI only for true emergencies, monitor draft acceptance rate hourly for the first 48 hours, and keep one founder or lead agent on point for escalations that need a human voice. AI is most valuable during the rush precisely because fatigue causes human errors; a well-trained playbook outperforms a 14-hour-shift agent on WISMO and discount-code tickets every time.
Check the dashboard three times a day: morning, lunch, and a final 10 p.m. review. Look at AI resolution rate, escalation count, and CSAT trend. If resolution rate drops below 50%, the KB is stale; pull the last 20 escalations and update the playbook. If CSAT dips below 85%, route more sensitive tickets (refunds, address changes) to humans even when AI could handle them. The goal during the rush is steady, not heroic.
What mistakes do Shopify stores make most often during BFCM support?
Three patterns show up every year: forecasting from industry averages instead of own TPO, treating all tickets as equal urgency, and pausing AI on Black Friday because of an early false positive. None of these are good ideas. Own-data forecasting is the only kind that survives contact with your actual category mix and shipping speed, and pausing AI during the rush is the single fastest way to lose the cost-per-resolution advantage you built the previous month.
The fix for the false-positive pause is to set a threshold in advance: AI stays on unless draft acceptance drops below 40% for two consecutive hours. That number triggers a review, not a shutdown. Most of the time the issue is a single new SKU with a sizing question the KB does not cover, and the right response is a 10-minute playbook update, not a full pause.
The fastest path to a survivable BFCM is forecast, pack, playbook, connect, SLA, shadow. Run the checklist by November 1, monitor three times daily through Cyber Monday, and let AI carry the long tail so your team can spend their energy on the 20% of tickets that actually require a human.