You’ve spent three months developing the product. The samples are approved, the packaging is final, the listing copy is written. Your supplier needs a PO by Friday. And the only number that matters — how many units to order — is the one you have the least information about.

Every operator who’s launched a new SKU knows this moment. You’re staring at a spreadsheet, trying to conjure demand data for a product that doesn’t exist yet. Your supplier’s MOQ is 2,000 units. Your gut says 5,000. Your co-founder says 10,000 because “what if it takes off?” Your accountant says 1,000 because cash is tight.

Everyone is guessing. The question isn’t whether your first PO will be wrong — it will be. The question is whether you’ve structured the buy so the miss is survivable.

What makes new-product POs different from replenishment

Replenishment POs are math. You have velocity data, seasonality curves, lead times, and safety stock targets. A decent demand planning model (or even a well-maintained spreadsheet) can generate a reorder quantity that’s within 10–15% of optimal.

New product POs have none of that. You’re operating on analogy, assumption, and market intuition. The failure modes are asymmetric: order too few and you miss the launch window, frustrate early customers, and hand momentum to a competitor. Order too many and you’re sitting on capital that could have funded three other initiatives, plus you’re paying storage on units that might take 18 months to sell through.

Replenishment PO	New Product PO
Based on historical velocity	Based on analogues and assumptions
Error range: ±10–15%	Error range: ±50–300%
Cost of overstock: carrying cost	Cost of overstock: carrying cost + opportunity cost + potential obsolescence
Cost of stockout: lost sales on a proven product	Cost of stockout: lost launch momentum, possibly unrecoverable
Confidence level: high	Confidence level: low, and everyone pretends otherwise

The core problem isn’t forecasting accuracy — it’s that the standard inventory math assumes you have inputs you don’t have. Applying a reorder point formula to a product with no demand history gives you a number that feels precise and is completely fictional.

The analogue method (and why most teams do it wrong)

The most common approach to new-product forecasting is finding a comparable product and using its sales history as a proxy. In theory, this makes sense. In practice, it fails because teams pick analogues based on product similarity rather than go-to-market similarity.

A $24 silicone spatula launching on your DTC site in March is not comparable to a $24 silicone spatula that launched in Target in October, even if the products are identical. What drives initial sell-through isn’t the product spec — it’s the channel, the marketing spend behind the launch, the existing audience size, the competitive set at the moment of launch, and whether you’re entering a category where you already have brand authority or building from zero.

Better analogue selection looks at:

1. Same channel or channel mix (DTC-first vs. wholesale-first vs. marketplace-first)
2. Similar launch marketing spend (within 2x)
3. Similar price point (within 25%)
4. Similar category awareness (are you the 50th brand in this space or the 3rd?)
5. Similar seasonality window (launching into peak vs. off-peak)

If your best analogue matches on 3+ of these dimensions, its first-90-day velocity is a reasonable starting proxy. If it only matches on product type and price, you’re anchoring on noise.

Analogue velocity (first 90 days):     3,200 units
Adjustment for smaller audience:       × 0.6
Adjustment for off-peak launch:        × 0.75
Adjustment for higher price point:     × 0.85

Adjusted 90-day forecast:              3,200 × 0.6 × 0.75 × 0.85 = 1,224 units
Weekly run rate (implied):             ~136 units/week

That adjusted number is still a guess. But it’s a calibrated guess — you’ve explicitly named the assumptions, which means you can revisit them as real data comes in.

The pre-launch signal stack

Analogues get you a baseline. Pre-launch signals help you adjust it before committing capital.

Not every signal is available to every brand, and none of them are precise predictors. But stacking multiple weak signals gives you a directional read that’s better than gut alone.

Signal	What It Tells You	Reliability	How to Capture
Waitlist / “notify me” signups	Direct demand intent	Medium-high	Landing page with email capture, 4–6 weeks pre-launch
Social engagement on teaser content	Category interest (not purchase intent)	Low-medium	Track saves and shares, not likes
Pre-order conversion rate	Willingness to pay at your price point	High	Run a 2-week pre-order window with estimated ship date
Ad click-through on concept creative	Message-market fit	Medium	$500–$1,000 in test spend on Meta/Google
Retailer buyer interest	Wholesale channel viability	Medium-high	Pitch 3–5 buyers; track who requests samples vs. who ghosts
Competitor sell-through data	Category velocity benchmark	Medium	Use tools like Jungle Scout (Amazon), Stackline, or retailer portal data

The pre-order signal is the strongest because it involves actual money. If you run a 2-week pre-order and convert 3% of your email list, you have a real demand data point. If you convert 0.3%, that tells you something too — either the product, the price, or the positioning needs work before you commit to a large buy.

Email list size:                        25,000
Pre-order conversion rate:              2.8%
Pre-order units:                        700

Implied first-90-day demand (at 3× pre-order):   2,100 units
Implied first-90-day demand (at 5× pre-order):   3,500 units

Multiplier depends on:
  - How much launch marketing spend vs. pre-order spend
  - Whether pre-order audience was your warmest segment
  - Whether you’ll expand to additional channels post-launch

The 3–5× multiplier on pre-orders is a rough heuristic, not a law. Brands with large, engaged audiences and heavy launch marketing can see 8–10× pre-order volume in the first 90 days. Brands launching quietly into a new category might see 1.5–2×. Your own history of pre-order-to-launch ratios on previous SKUs — if you have any — is the best calibration.

Structuring the buy: how to be wrong safely

The goal isn’t getting the first PO right. It’s structuring the buy so that being wrong doesn’t cripple you financially or operationally.

Three levers control your downside exposure on a new product buy:

1. Negotiate a split-ship or phased delivery

Instead of taking 5,000 units in one shipment, negotiate with your supplier for a phased delivery: 2,000 units in the initial shipment, 1,500 units 45 days later, 1,500 units 90 days later. You commit to the full quantity (preserving your per-unit cost), but you delay cash outlay and storage costs on 60% of the order.

This only works with suppliers who have the production capacity to hold inventory or who are running continuous production. For seasonal or limited-run products, you may need to take everything at once. But for evergreen SKUs from established suppliers, split-ship terms are common and under-negotiated.

2. Set a kill threshold before you place the order

Decide in advance what sell-through rate at 30 and 60 days would cause you to cancel or reduce the remaining phases. Write it down. Share it with your team. This prevents the sunk cost fallacy from turning a bad bet into a catastrophic one.

First 30 days:
  Target velocity:    150 units/week
  Kill threshold:     < 60 units/week (40% of target)
  Action if below:    Cancel phase 2 and phase 3; liquidate remaining via bundle or flash sale

First 60 days:
  Target velocity:    120 units/week (settling from launch spike)
  Reduce threshold:   < 80 units/week
  Action if below:    Reduce phase 2 by 50%; cancel phase 3
  
  Accelerate threshold: > 200 units/week
  Action if above:     Add a phase 4 PO at the same unit cost

The psychological trap is real. At day 45, when you’ve sold 800 units out of a 2,000-unit initial buy and velocity is declining, it’s very tempting to say “it’ll pick up when we launch the ads” or “holiday will save it.” Maybe it will. But if your kill threshold said cancel at 60 units/week and you’re at 55, cancel. You defined the threshold when you were thinking clearly. Trust it.

3. Choose your MOQ battles

Suppliers quote MOQs as if they’re immutable laws of physics. They’re not. MOQs reflect the supplier’s production economics — setup costs, material minimums, line changeover time — but they’re almost always negotiable, especially if you’re bringing the supplier new business or have a credible path to larger volumes.

Negotiation approaches that work:

• Pay a higher per-unit cost on the first order in exchange for a lower MOQ. If the standard MOQ is 5,000 units at $8.50/unit, offer to buy 2,000 units at $9.75/unit. You’re paying 15% more per unit, but you’re risking $19,500 instead of $42,500. That’s a $23,000 reduction in downside exposure.

• Commit to an annual volume in exchange for a lower initial order. “I’ll commit to 15,000 units this year, but I need to start with 2,500 on the first PO.”

• Offer to share tooling or setup costs explicitly. Some suppliers bake setup costs into a high MOQ. If the setup costs $3,000, offer to pay that separately and buy 1,000 units at the run-rate price.

Approach	First-Order Risk	Per-Unit Cost Impact	Best For
Full MOQ, standard price	High ($42,500)	Baseline ($8.50)	Proven categories with high confidence
Reduced MOQ, premium price	Medium ($19,500)	+15% ($9.75)	New categories, unproven channels
Annual volume commitment	Medium ($21,250)	Baseline ($8.50)	Brands with 2+ successful SKUs with this supplier
Separate setup + lower MOQ	Low ($11,500)	+5% ($8.93)	Custom/tooled products, new supplier relationships

The 90-day feedback loop

Your first PO is a hypothesis. The first 90 days of sell-through are the experiment. Treat them that way.

Week 1–2 data is almost useless for long-term forecasting. Launch spikes (or launch duds) driven by your email blast, your social push, and your ad spend reflect marketing intensity, not organic demand. What matters is the velocity after the launch energy dissipates — typically weeks 3–6.

Build a simple tracking model that updates weekly:

Week | Units Sold | Cumulative | Weeks of Inventory Remaining | Action Trigger
─────┼────────────┼────────────┼──────────────────────────────┼──────────────────
  1  |    280     |    280     |         6.1                  | Launch spike — ignore
  2  |    195     |    475     |         5.4                  | Expected decline
  3  |    110     |    585     |         5.7                  | Settling — watch
  4  |     95     |    680     |         5.5                  | Baseline forming
  5  |     88     |    768     |         5.6                  | Steady — this is your run rate
  6  |     92     |    860     |         5.0                  | Confirm: ~90/week organic
  7  |     85     |    945     |         4.9                  | Reorder math starts here
  8  |     91     |   1,036    |         4.2                  | Place replenishment PO

By week 5 or 6, you have a genuine organic velocity. Now you can run real replenishment math — and you’re no longer guessing. The transition from “new product PO” to “replenishment PO” happens when you have 4+ weeks of post-launch-spike data. That’s when the standard formulas start working.

The critical metric during this period is weeks of inventory remaining at current velocity. If that number drops below your lead time (the time from PO to units on shelf), you need to reorder immediately or you’ll stock out. If it’s climbing, your velocity is declining and you should be thinking about whether to reduce subsequent PO phases.

Reorder trigger:
  Weeks of inventory remaining < (Lead time in weeks + Safety stock in weeks)

Example:
  Current inventory:            1,200 units
  Current weekly velocity:      90 units
  Weeks remaining:              1,200 / 90 = 13.3 weeks
  Supplier lead time:           8 weeks
  Safety stock:                 2 weeks (at current velocity = 180 units)
  Reorder point:                8 + 2 = 10 weeks

  13.3 > 10 → Not yet. Check again next week.

Common mistakes that make a bad PO worse

The PO itself is just the starting point. How teams handle the aftermath of a wrong forecast determines whether the miss costs $5,000 or $50,000.

Chasing a slow launch with more marketing spend

Product moves slowly in weeks 3–5. The instinct is to throw more ad dollars at it. Sometimes this works — if the product is right and you’ve under-invested in awareness, more spend can find the audience. But more often, the product is reaching its natural velocity, and incremental marketing spend has rapidly diminishing returns. You’re buying units sold at a higher and higher CAC, which eats the margin that was supposed to justify the product in the first place.

Before increasing spend, check two things: is your conversion rate on the product page healthy (above 2.5% for DTC), and is your return rate normal (under 8%)? If conversion is low, the problem is product-market fit or positioning, not awareness. More traffic to a low-converting page just burns cash faster.

Discounting too early

The product isn’t moving at $34. The temptation to drop it to $27 kicks in at week 4. Resist it. Early discounting trains your audience to wait for sales, damages your brand positioning, and — critically — gives you false signal. If it sells at $27, you still don’t know whether it would have sold at $34 with better messaging, different creative, or in a different channel.

If you’re going to discount, do it strategically: a time-limited launch bundle (pair the new SKU with a proven seller), a loyalty-only price, or a channel-specific promotion that doesn’t pollute your core DTC pricing. And track the discounted velocity separately — it shouldn’t inflate your organic demand estimates.

Ignoring channel-specific velocity

A new SKU that does 90 units/week on your DTC site might do 400 units/week on Amazon or 50 cases/week through a regional distributor. Treating “90 units/week” as the product’s velocity when it’s only the DTC velocity leads to under-ordering for your next expansion channel.

When your tracking model stabilizes, break velocity out by channel. Your replenishment PO should be built on an allocation model, not a single blended number:

Channel        | Weekly Velocity | Weeks of Stock | Allocation % | Next PO Units
───────────────┼─────────────────┼────────────────┼──────────────┼──────────────
DTC site       |       90        |      13.3      |     30%      |     1,080
Amazon FBA     |      160        |       8.2      |     53%      |     1,920
Wholesale      |       50        |      16.0      |     17%      |       600
───────────────┼─────────────────┼────────────────┼──────────────┼──────────────
Total          |      300        |                |    100%      |     3,600

Ordering the next PO before the first one lands

Lead times create anxiety. Your supplier quotes 12 weeks. Your launch is in 8 weeks. So you place PO #2 before PO #1 even arrives, because you’re worried about a stockout gap if the product takes off.

This doubles your exposure before you have a single data point. Unless you have pre-order data showing genuine demand exceeding your first PO, resist placing a second order until you have at least 3–4 weeks of post-launch sell-through. Yes, this means you might stock out briefly between PO #1 selling through and PO #2 arriving. A 2-week stockout on a new product is annoying. $80,000 in dead inventory is a business problem.

A framework for confidence-based ordering

Instead of trying to forecast the exact right number, think in terms of confidence tiers. How much are you willing to commit at each level of certainty?

                          Confidence
 Tier       Units         Level          When to Commit
─────────────────────────────────────────────────────────
 Tier 1     MOQ–2×MOQ     Low            At PO placement (pre-launch)
 Tier 2     1.5–3×T1      Medium         After 4–6 weeks of sell-through
 Tier 3     Scale order    High           After 90 days + channel expansion

Tier 1 is your “learning buy.” You’re paying for information as much as inventory. The cost premium of a lower MOQ or split-ship arrangement is the tuition for real demand data.

Tier 2 is your first data-informed order. You have organic velocity, you know your conversion rate, you know your return rate. The uncertainty band narrows from ±50–300% to ±20–30%.

Tier 3 is a standard replenishment order. You’re past the new-product phase. Forecasting models work. You’re optimizing, not guessing.

The mistake most brands make is jumping straight to Tier 3 quantities at Tier 1 confidence. They see a $2 per-unit cost savings at higher volume and commit $85,000 to save $10,000. That trade only works if your forecast is right — and at Tier 1 confidence, it almost never is.

When to ignore all of this and go big

There are scenarios where a conservative first PO is the wrong move:

• You have binding wholesale commitments (a Target PO for 10,000 units ships in 14 weeks — you don’t get to “test with 2,000 first”)
• The product is a line extension of a proven seller (same formula, new colorway) where existing customer data provides genuine demand signal
• Your supplier only runs production twice a year, and missing this window means 6+ months without product
• You’re entering a time-sensitive market window (seasonal product, trend-driven category) where being late is worse than being over-inventoried

In these cases, the right move is to size the PO to the opportunity, accept the higher risk, and plan your liquidation strategy before you place the order. Know your floor price, know your secondary channels, and know the carrying cost timeline at which it becomes cheaper to sell at a loss than to store.

Even a high-confidence big buy should have a written exit plan. “If we’ve sold less than 30% of this order by day 90, we execute the following liquidation sequence.” Having that document before you need it prevents emotional decision-making when you’re staring at 7,000 unsold units.

After the first PO: building the forecasting muscle

Every new product launch is training data for the next one. But only if you capture it systematically.

After each launch, record:

1. The analogue you used and the adjustments you made
2. The pre-launch signals you had and what they predicted
3. The actual first-90-day sell-through vs. your forecast
4. The ratio of launch-spike velocity (weeks 1–2) to organic velocity (weeks 5–8)
5. Which channels performed above/below your allocation model
6. The total landed cost vs. your pro forma cost (including any premiums for lower MOQs, split-ships, or expedited freight)

After 5–10 launches, patterns emerge. You’ll learn that your pre-order-to-launch multiplier is consistently around 4×, or that your analogue method over-forecasts by 25%, or that your Amazon velocity always exceeds your DTC velocity by 2.5× after week 6. These ratios become your institutional knowledge — the calibration layer that makes each subsequent new-product PO slightly less wrong than the last.

The brands that scale efficiently through $10M to $50M aren’t the ones that figured out a magic forecasting formula. They’re the ones that got disciplined about learning from each miss, structuring buys to survive the miss, and building a feedback loop that makes the next miss smaller.

---

CommerceOS connects your PO pipeline to real-time sell-through and allocation data across every channel — so the moment your new product’s velocity stabilizes, your replenishment math runs itself. Book a demo to see how it works.