Ops & Supply

How to forecast product demand as CPG Founders

Ops & SupplyFor CPG Founders3 apps12 steps~24 min to set up

Your demand forecast is a Shopify export pasted into a spreadsheet, adjusted by whatever your gut says about the upcoming Whole Foods reset. You've over-produced twice this year — once on a seasonal SKU that aged out before it sold through — and stocked out during a regional promotion that your broker didn't warn you about until it was too late. Your co-packer needs a 6-week lead time, which means you're guessing demand two months out with month-old data. You don't have a demand planner on staff. You have yourself, a Google Sheet, and the memory that Q4 always spikes but you can never remember by exactly how much.

Build this on Starch → See the apps used

Ops & SupplyFor CPG Founders3 apps12 steps~24 min to set up

Outcome

What you'll set up

An AI-powered demand plan that pulls your actual Shopify sales velocity, POS sell-through data, and historical seasonal patterns — so your forecast is based on what's actually selling, not last month's number plus a guess

Automated reorder triggers wired to your co-packer lead times, so you're not manually reverse-engineering production schedules every time you run low

A unified view of forecasted demand against real inventory across your co-packer, 3PL, and warehouse locations — so you can catch a potential stockout before it happens instead of after

The Starch recipe

Apps, data, and prompts

The combination of Starch apps, the data sources they pull from, and the prompts you use to drive them.

Apps used

Demand Planner Inventory Planner Retail Analytics

Data sources & config

Starch connects directly to Shopify from its integration catalog — the agent queries your order and inventory data live when the demand plan runs. POS data from retail accounts (Whole Foods, Sprouts, regional chains) is pulled through browser automation where retailer portals don't expose an API — no API needed. Co-packer and 3PL inventory feeds are ingested via CSV import or browser automation against your partners' portals. Starch syncs your Stripe data on a schedule to cross-reference revenue against forecasted sell-through.

Prompts to copy

Build me a demand forecast for my top 8 SKUs using the last 18 months of Shopify orders and my Whole Foods POS sell-through data. Factor in a 6-week co-packer lead time, flag any SKU where I'll stock out within 10 weeks at current velocity, and show me what a 20% promotional lift would do to those numbers.

Show me inventory across my co-packer, 3PL in Dallas, and my FBA reserve — grouped by SKU, with days-on-hand calculated at current sell rate and a red flag on anything under 45 days.

Pull my POS sell-through by store for the last 8 weeks at Sprouts and compare it to the same period last year. Highlight any stores where velocity dropped more than 15% week-over-week.

Run these in Starch → or paste them into your favorite agent

Walkthrough

Step-by-step

1 Connect Shopify from Starch's integration catalog so the agent can query your live order history, SKU-level sales volume, and current inventory positions whenever a forecast or replenishment plan runs.

2 Wire in POS data from your retail accounts — Whole Foods, Sprouts, Target, or regional chains. Where a retailer provides a portal (like Whole Foods' supplier hub), Starch automates it through your browser — no API needed. Where they send weekly reports via email, route those to Starch via Gmail, which Starch syncs on a schedule.

3 Set your co-packer lead times and MOQs inside Demand Planner — currently in beta, request access — so every forecast automatically accounts for how far out you need to commit before you can actually get product.

4 Tell Starch your SKU roster and any planned promotions: 'I have a 15% trade promotion running at Sprouts in weeks 8-10 next quarter. Build the lift assumption into the forecast for that region.' The agent incorporates the promotional period as a demand multiplier.

5 Demand Planner surfaces a rolling 12-week forecast by SKU, flagging where you're at risk of stocking out given current inventory plus in-production quantities — and showing you the production order you'd need to place today to avoid the gap.

6 Inventory Planner — also in beta — gives you one dashboard of stock across your co-packer, 3PL, and FBA reserve. Shelf-life tracking ensures FEFO rotation is flagged automatically, so you don't find out a pallet is 60 days from expiry only when your 3PL calls you.

7 Retail Analytics — in beta — pulls POS sell-through velocity by store so you can see whether the demand signal at shelf is tracking to plan or diverging. If Whole Foods Central region is running 20% above forecast, you know to pull forward your next production run.

8 Set automated reorder alerts: 'Notify me via Slack when any SKU drops below 6 weeks of forward cover based on current sell rate.' Starch syncs your Slack workspace and sends the message directly to your ops channel when the threshold is hit.

9 Run a scenario: 'What happens to my Q3 inventory position if I land the Costco road-show and sell 4,000 units in 3 weeks?' The agent recalculates safety stock requirements and flags which SKUs need an emergency production run.

10 Before every buyer meeting or QBR, pull a one-page sell-through summary: 'Show me the last 12 weeks of POS velocity at [Retailer], compare to prior year, and highlight stores where I have distribution but velocity under 1 unit per week per store.' Walk in with your own data instead of waiting for the retailer to hand you a spreadsheet.

11 Tie the demand forecast back to cash planning: tell Starch 'Estimate my COGS commitment for the next two production runs based on this forecast, and show me how that lands against my current Plaid cash balance.' Starch syncs your Plaid data on a schedule and calculates the runway impact of the production spend.

12 Revisit and recalibrate weekly — the forecast updates automatically as new Shopify orders come in, so you're not manually refreshing a spreadsheet. You review the exceptions, not the data entry.

See this running on Starch

Connect your tools, describe what you want, and the agent builds it. Closed beta is free.

Try it on Starch →

Worked example

Spring 2026 Reorder Planning — Grain-Free Granola Brand, 6 SKUs

Sample numbers from a real run

Blueberry Almond SKU — current 3PL stock	1,840
Blueberry Almond SKU — in production (due week 4)	2,400
Forecasted demand weeks 1-10 at current velocity	3,200
Forecasted demand weeks 1-10 with Sprouts promo lift (20%)	3,840
Projected stockout gap (units)	400
Production run needed to cover gap (MOQ 1,200 units)	1,200

You have 1,840 units of your Blueberry Almond SKU at your Dallas 3PL and another 2,400 coming out of your co-packer in week 4. At your current Shopify plus Sprouts velocity of about 320 units a week, that covers you through roughly week 13 — fine, until Starch's Retail Analytics surface flags that Sprouts Central has been running 22% above your baseline for the past three weeks, likely because a competitor SKU went out of stock. When Demand Planner recalculates with the updated velocity and adds the 20% promotional lift you told it to assume for the Sprouts end-cap you booked for weeks 8 through 10, total forecasted demand through week 10 jumps from 3,200 to 3,840 units. Against 4,240 units of supply, that's a gap of 400 units — and your co-packer needs 6 weeks lead time with a 1,200-unit MOQ. Starch surfaces the production order you need to place this week to avoid the stockout, flags the cash impact against your Plaid balance (about $6,800 in COGS at your current cost per case), and posts a Slack message to your ops channel. You catch it on a Tuesday morning instead of getting a 'sorry, we're out of stock' message from your Sprouts buyer in week 9.

Measurement

How you'll know it's working

Forecast accuracy by SKU (% variance between predicted and actual sell-through, measured weekly)

Days of forward cover by SKU across all locations (target: 8-12 weeks given co-packer lead times)

Stockout events per quarter (number of weeks any SKU hit zero available inventory at any selling channel)

Overproduction write-off ($value of product expired or discounted out due to excess inventory)

POS sell-through velocity (units per store per week by retail account, tracked vs. prior period)

Comparison

What this replaces

The other ways teams handle this today, and how the Starch version compares.

Excel / Google Sheets + manual Shopify export

Free and fully owned, but you're doing all the data wrangling by hand every time — it doesn't update automatically, it doesn't model lead times unless you build the formula, and it breaks the moment you have more than 10 SKUs across 3 channels.

Cin7 or Brightpearl

Purpose-built for multi-location inventory, but expensive per seat, require meaningful setup time to onboard co-packer and 3PL feeds, and don't natively pull POS sell-through from retail accounts — you're still doing that part in a spreadsheet.

Crisp or NielsenIQ Byzzer (syndicated data)

Gives you category benchmarks and distribution analytics, but costs $10k–$50k/year, is oriented toward brands with significant retail footprint, and still doesn't connect your forecast to your production schedule or inventory positions.

NetSuite with Demand Planning module

Enterprise-grade and genuinely powerful, but the implementation alone runs six figures and assumes you have a supply chain analyst to operate it — not a solo founder wearing five hats.

Stocky by Shopify

Free and integrated with Shopify, but only looks at your DTC channel — it has no concept of retail POS sell-through, co-packer lead times, or multi-location 3PL inventory, so it gives you an incomplete picture the moment you're in retail.

On Starch RECOMMENDED

One platform — demand planner, inventory planner, retail analytics all running on connected data. Setup in plain English; numbers stay current via scheduled syncs and live agent queries.

Try it on Starch →

FAQ

Frequently asked questions

Demand Planner and Inventory Planner both say 'currently in development' — when can I actually use them?

Both apps are in beta and available to request access now. You'll get notified when your account is activated. In the meantime, you can connect Shopify and Plaid today, and tell Starch in natural language what you want to track — the agent will build a working forecast surface from your connected data while the dedicated app template is in development.

My retail accounts don't have a formal API — can Starch still pull their POS data?

Yes. For retailer portals where you log in and download a report — like the Whole Foods supplier hub or KeHE Connect — Starch automates that through your browser, no API needed. For accounts that email you weekly Excel sell-through files, Starch reads those via Gmail, which syncs on a schedule. You tell Starch where the data lives and it figures out how to get it.

Does Starch store my sales history so I can run year-over-year comparisons?

Starch is built for live data surfaces, not a long-horizon data warehouse. Historical data that's available through your connected sources — Shopify order history, synced Plaid transactions — is accessible to the agent when it builds your forecast. If you need multi-year archival analytics, you'd want a separate data warehouse; Starch is the operational layer on top, not the archive.

Can I model different scenarios — like what happens if I land a new Costco door?

Yes. You describe the scenario in plain language — 'Assume I add 120 Costco doors in Q3 selling 2 units per store per week. What does that do to my inventory position?' — and Starch recalculates your forward cover and flags any SKUs that need a production run to cover the new demand. No formula-writing required.

My co-packer sends me a weekly production schedule in a Google Sheet. Can Starch read that?

Yes. Connect Google Sheets from Starch's integration catalog — the agent queries it live when your inventory or demand plan runs. Tell Starch the structure of the sheet and it pulls production quantities and expected completion dates into your replenishment model automatically.

Is Starch SOC 2 certified? I'm asking because we share some financial and sales data.

Not yet — Starch is not currently SOC 2 Type II certified. If that's a hard requirement for your legal or investor review process, that's worth knowing upfront. It's on the roadmap.