Customer Support

How to respond to online reviews as CPG Founders

Customer SupportFor CPG Founders2 apps12 steps~24 min to set up

Your brand gets reviews on Amazon, your DTC Shopify site, Google, and maybe Walmart.com — all in different dashboards, all requiring manual log-ins to check. A one-star review about a broken seal or a wrong flavor hits your Amazon listing on a Friday night and sits there unanswered until Monday because you were running a production run all weekend. You know response time and tone affect both your seller rating and whether the next shopper converts. But you're also the person doing co-packer calls, filing deduction disputes, and managing FBA shipments. You don't have a customer service rep. You have a notebook and a phone.

Build this on Starch → See the apps used

Customer SupportFor CPG Founders2 apps12 steps~24 min to set up

Outcome

What you'll set up

An automated review-monitoring workflow that surfaces new reviews from Amazon, Google, and your DTC site every morning in a single digest — no manual log-ins required

Draft replies generated for each new review, matched to your brand voice and product context, ready to send or edit in one step

An escalation rule that flags reviews mentioning food safety, contamination, or allergic reactions so they never get auto-replied and always hit your personal inbox within minutes

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

Customer Support Agent

Data sources & config

Starch automates log-in to Amazon Seller Central through your browser — no Amazon reviews API needed. Google Business reviews are pulled through browser automation the same way. Starch connects directly to Gmail so review notification emails land in your triage queue; the Email Triage starter app is wired to Gmail via scheduled sync. Your Notion brand voice doc is connected via Starch's scheduled sync so the agent always drafts replies against your current guidelines, not a cached version. Customer Support Agent (coming soon) will handle multi-channel reply routing at scale; in the meantime, the Email Triage app and browser automations cover the same ground.

Prompts to copy

Every morning at 7am, pull all new reviews from our Amazon Seller Central account, our Google Business profile, and our Shopify store from the past 24 hours. Group them by rating (1-2 star, 3 star, 4-5 star). For any review mentioning 'allergy,' 'sick,' 'mold,' 'recall,' 'contamination,' or 'wrong product,' flag it as urgent and send me a Slack DM immediately. For everything else, draft a reply using the brand voice doc in our Notion workspace and queue it for my review.

Draft a reply to this 2-star Amazon review: [paste review text]. We're a better-for-you snack brand. Tone should be warm, direct, and apologetic without being defensive. Offer to make it right with a replacement or refund. Do not include discount codes. Keep it under 120 words.

Summarize last week's review sentiment across all channels. Flag any recurring complaint themes — packaging issues, shipping damage, flavor inconsistency — and tell me how many reviews mentioned each one.

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

Walkthrough

Step-by-step

1 Connect your Amazon Seller Central account — Starch automates browser log-in to your Seller Central dashboard, no API credentials required. The agent navigates to the Customer Reviews section and pulls all reviews posted in the last 24 hours.

2 Connect your Google Business profile the same way — Starch automates your browser session to pull new reviews from Google Maps/Business, capturing star rating, reviewer name, review text, and timestamp.

3 Connect Starch to Gmail via scheduled sync so any review notification emails (from Amazon, Shopify, or third-party review aggregators you use) are automatically ingested alongside the browser-pulled data.

4 Set up the morning digest automation: tell Starch 'Every day at 7am, compile all new reviews from the past 24 hours across Amazon, Google, and Shopify into a single summary, grouped by channel and star rating, and send it to my Gmail.'

5 Configure the food-safety escalation rule: 'If any review contains the words allergy, allergic, sick, illness, contamination, mold, foreign object, or recall, skip the draft queue and send me a Slack message immediately with the full review text and a link to respond.'

6 Wire your Notion brand voice document via Starch's scheduled sync so the draft-reply agent always references your current tone guidelines, product descriptions, and standard response policies.

7 For each non-urgent review, Starch drafts a reply using your brand voice doc as context. 1-2 star reviews get a draft that acknowledges the issue, apologizes, and offers a resolution path. 4-5 star reviews get a short, warm acknowledgment.

8 Drafts land in a review queue — a lightweight Starch app you build with a prompt like 'Show me all pending review replies in a table with the original review, the draft response, the channel, and an Approve / Edit / Skip button for each row.'

9 You work through the queue in one sitting — approve the ones that look right, edit the ones that need a personal touch, skip anything that needs follow-up. Approved replies post automatically back to the platform via browser automation.

10 At the end of each week, run the sentiment summary prompt: Starch reads all reviews from the past 7 days, identifies recurring complaint themes (e.g., 'pouch seal failing in transit' shows up 6 times), and outputs a prioritized list you can hand to your co-packer or operations lead.

11 Optional: pipe the weekly complaint themes into your Notion ops database or a Starch-built issue tracker so packaging and formulation issues are logged with the review evidence attached.

12 When Customer Support Agent launches (currently in development — request beta access), you'll migrate this workflow into a fully autonomous reply system that handles tier-1 responses 24/7 and escalates based on rules you define.

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

Week of March 10, 2026 — post-Amazon Prime sale surge

Sample numbers from a real run

New Amazon reviews (7 days)	34
1-2 star reviews flagged for response	11
Food-safety escalations (allergy/contamination keywords)	2
Replies drafted by Starch	32
Replies approved without edits	24
Replies edited before posting	8
Average founder time spent on review responses	22

The week after your Prime sale, 34 new Amazon reviews came in over seven days — more than double your typical volume. Two reviews contained the word 'allergic' and triggered immediate Slack alerts; both turned out to be customers confusing your cashew-based product with a peanut product, not an actual formulation issue, but you responded personally within 40 minutes with the allergen statement and a replacement offer. Of the remaining 32 reviews, Starch drafted replies for all of them using your Notion brand voice doc. The 11 one- and two-star reviews were mostly about pouch seals failing in transit — Starch flagged 'seal' as a recurring theme in the weekly digest, you screenshotted the summary and sent it to your co-packer on Thursday. Twenty-four replies posted without any edits. Eight needed a personal note — mostly situations where a customer mentioned a specific order number or described a genuinely unusual experience. Total time you spent on review management that week: about 22 minutes, spread across two sittings.

Measurement

How you'll know it's working

Average response time to 1-2 star reviews (target: under 4 hours during business days)

Review response rate by channel (Amazon, Google, DTC) — most CPG sellers leave 30-40% of reviews unanswered

Recurring complaint theme frequency by week (packaging, shipping, flavor, wrong item) — early signal for co-packer or 3PL issues

Amazon seller feedback score and whether response behavior correlates with rating changes

Food-safety keyword escalation rate — the one metric you want to stay at zero, and the one where a delay costs you the most

Comparison

What this replaces

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

Yotpo or Okendo (DTC review platforms)

Good for Shopify-native review collection and display, but don't touch Amazon or Google, so you're still checking three dashboards manually and have no cross-channel reply workflow.

Birdeye or Podium (multi-location review management)

Built for service businesses with physical locations — pricing and feature set assume you have a dedicated marketing coordinator, not a founder doing this between co-packer calls.

Manual process — checking Seller Central + Google + Shopify separately

Free, but during a sale surge you'll miss reviews for 48-72 hours and your Amazon response rate metric will drop, which affects search ranking and Buy Box eligibility.

Hiring a VA for review monitoring

Works, but the VA still needs a process doc, you'll pay $15-25/hr for something mostly mechanical, and food-safety escalations still depend on the VA recognizing the keywords on the right day.

On Starch RECOMMENDED

One platform — customer support agent, founder inbox 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

Can Starch actually post replies back to Amazon, or does it just draft them?

Starch automates your browser session — the same way you'd log in and type a reply yourself. So yes, approved replies post directly to Amazon Seller Central through browser automation. No Amazon API or third-party review platform subscription needed. You review and approve before anything goes live.

What about Google Business reviews — do I need to give Starch my Google account password?

No. Starch uses browser automation with your existing authenticated session, the same way it works for any web-based platform. You authorize the browser session once; Starch handles navigation from there.

The Customer Support Agent sounds like what I actually want. When is it available?

Customer Support Agent is currently in development — it's coming soon. You can request beta access to get notified when it launches. In the meantime, the Email Triage app and the browser automation workflow described here cover the same core loop: surface reviews, draft replies, post approved ones. It's more hands-on than the full agent will be, but it's not starting from zero.

I sell on Walmart.com and Target.com in addition to Amazon. Can Starch handle those too?

Yes — any retailer portal you can log into through a browser, Starch can automate through your browser — no API needed. The setup is the same: you tell Starch which sites to check, what section to navigate to, and how to extract review data. If the portal structure changes, the automation may need a quick update, but it's not a hard dependency on an official API.

Is Starch SOC 2 certified? I'm nervous about giving it access to my Amazon account.

Starch is not SOC 2 Type II certified yet — that's an honest limit worth knowing. There's no on-premises option either. If SOC 2 certification is a hard requirement for your brand or your retail partners, that's a real constraint to weigh. Most early-stage CPG founders we work with treat this the same way they treat any SaaS tool — evaluate the risk against the operational gain.

What if a review mentions a food safety issue — will Starch auto-reply before I see it?

No. The escalation rule is designed to bypass the draft queue entirely for any review containing food-safety keywords. Those reviews trigger an immediate Slack alert to you personally and are never auto-replied. You define the keyword list; Starch enforces it. This is the one category where a wrong auto-reply creates regulatory and reputational exposure, so it's deliberately manual.

Can I use my existing brand voice guidelines, or do I have to re-enter everything?

If your brand voice doc is in Notion, Starch syncs your Notion workspace on a schedule and the agent drafts replies against whatever's in that doc. If it's a Google Doc or a PDF, connect Google Drive from Starch's integration catalog and point the agent at the right file. You don't rewrite anything — you just tell Starch where to look.