At Gorgias we recently switched our flask & celery apps from Google Cloud VMs provisioned with Fabric to using docker with kubernetes (k8s). This is a post about our experience doing this.

‍

Note: I'm assuming that you're somewhat familiar with Docker.

‍

Docker structure

The killer feature of Docker for us is that it allows us to make layered binary images of our app. What this means is that you can start with a minimal base image, then make a python image on top of that, then an app image on top of the python one, etc..

‍

Here's the hierarchy of our docker images:

• gorgias/base - we're using phusion/baseimage as a starting base image.
• gorgias/pgbouncer
• gorgias/rabbitmq
• gorgias/nginx - extends gorgias/base and installs NGINX
• gorgias/python - Installs pip, python3.5 - yes, using it in production.
• gorgias/app - This installs all the system dependencies: libpq, libxml, etc.. and then does pip install -r requirements.txt
• gorgias/web - this sets up uWSGI and runs our flask app
• gorgias/worker - Celery worker

‍

Piece of advice: If you used to run your app using supervisord before I would advise to avoid the temptation to do the same with docker, just let your container crash and let k8s handle it.

‍

Now we can run the above images using: docker-compose, docker-swarm, k8s, Mesos, etc...

‍

We chose Kubernetes too

There is an excellent post about the differences between container deployments which also settles for k8s.

‍

I'll also just assume that you already did your homework and you plan to use k8s. But just to put more data out there:

‍

Main reason: We are using Google Cloud already and it provides a ready to use Kubernetes cluster on their cloud.

‍

This is huge as we don't have to manage the k8s cluster and can focus on deploying our apps to production instead.

‍

Let's begin by making a list of what we need to run our app in production:

• Database (Postgres)
• Message queue (RabbitMQ)
• App servers (uWSGI running Flask)
• Web servers (NGINX proxies uWSGI and serves static files)
• Workers (celery)

‍

Why Kubernetes again?

We ran the above in a normal VM environment, why would we need k8s? To understand this, let's dig a bit into what k8s offers:

• A pod is a group of containers (docker, rtk, lxc...) that runs on a Node. It's a group because sometimes you want to run a few containers next to each other. For example we are running uWSGI and NGINX on the same pod (on the same VM and they share the same ip, ports, etc..).
• A Node is a machine (VM or metal) that runs a k8s daemon (minion) that runs the Pods.
• The nodes are managed by the k8s master (which in our case is managed by the container engine from Google).
• Replication Controller or for short rc tells k8s how many pods of a certain type to run. Note that you don't tell k8s where to run them, it's master's job to schedule them. They are also used to do rolling updates, and autoscaling. Pure awesome.
• Services take the exposed ports of your Pods and publishes them (usually to the Public). Now what's cool about a service that it can load-balance the connections to your pods, so you don't need to manage your HAProxy or NGINX. It uses labels to figure out what pods to include in it's pool.
• Labels: The CSS selectors of k8s - use them everywhere!
• ‍

There are more concepts like volumes, claims, secrets, but let's not worry about them for now.

‍

Postgres

We're using Postgres as our main storage and we are not running it using Kubernetes.

‍

Now we are running postgres in k8s (1 hot standby + pghoard), you can ignore the rest of this paragaph.

The reason here is that we wanted to run Postgres using provisioned SSD + high memory instances. We could have created a cluster just for postgres with these types of machines, but it seemed like an overkill.

The philosophy of k8s is that you should design your cluster with the thought that pods/nodes of your cluster are just gonna die randomly. I haven't figured our how to setup Postgres with this constraint in mind. So we're just running it replicated with a hot-standby and doing backups with wall-e for now. If you want to try it with k8s there is a guide here. And make sure you tell us about it.

‍

RabbitMQ

RabbitMQ (used as message broker for Celery) is running on k8s as it's easier (than Postgres) to make a cluster. Not gonna dive into the details. It's using a replication controller to run 3 pods containing rabbitmq instances. This guide helped: https://www.rabbitmq.com/clustering.html

‍

uWSGI & NGINX

As I mentioned before, we're using a replication controller to run 3 pods, each containing uWSGI & NGINX containers duo: gorgias/web & gorgias/nginx. Here's our replication controller web-rc.yaml config:

‍

apiVersion: v1
kind: ReplicationController
metadata:
 name: web
spec:
 replicas: 3 # how many copies of the template below we need to run
 selector:
   app: web
 template:
   metadata:
     labels:
       app: web
   spec:
     containers:
     - name: web
       image: gcr.io/your-project/web:latest # the image that you pushed to Google Container Registry using gcloud docker push
       ports: # these are the exposed ports of your Pods that are later used by the k8s Service
         - containerPort: 3033
           name: "uwsgi"
         - containerPort: 9099
           name: "stats"
     - name: nginx
       image: gcr.io/your-project/nginx:latest
       ports:
         - containerPort: 8000
           name: "http"
         - containerPort: 4430
           name: "https"
       volumeMounts: # this holds our SSL keys to be used with nginx. I haven't found a way to use the http load balancer of google with k8s.  
         - name: "secrets"
           mountPath: "/path/to/secrets"
           readOnly: true
     volumes:
       - name: "secrets"
         secret:
           secretName: "ssl-secret"
And now the web-service.yaml:apiVersion: v1
kind: Service
metadata:
 name: web
spec:
 ports:
 - port: 80
   targetPort: 8000
   name: "http"
   protocol: TCP
 - port: 443
   targetPort: 4430
   name: "https"
   protocol: TCP
 selector:
   app: web
 type: LoadBalancer

‍

That type: LoadBalancer at the end is super important because it tells k8s to request a public IP and route the network to the Pods with the selector=app:web.
If you're doing a rolling-update or just restarting your pods, you don't have to change the service. It will look for pods matching those labels.

‍

Celery

Also a replication controller that runs 4 pods containing a single container: gorgias/worker, but doesn't need a service as it only consumes stuff. Here's our worker-rc.yaml:

‍

apiVersion: v1
kind: ReplicationController
metadata:
 name: worker
spec:
 replicas: 2
 selector:
   app: worker
 template:
   metadata:
     labels:
       app: worker
   spec:
     containers:
     - name: worker
       image: gcr.io/your-project/worker:latest

‍

‍

Some tips

• Installing some python deps take a long time, for stuff like numpy, scipy, etc.. try to install them in your namespace/app container using pip and then do another pip install in the container that extends it, ex: namespace/web, this way you don't have to rebuild all the deps every time you update one package or just update your app.
• Spend some time playing with gcloud and kubectl. This will be the fastest way to learn of google cloud and k8s.
• Base image choice is important. I tried phusion/baseimage and ubuntu/core. Settled for phusion/baseimage because it seems to handle the init part better than ubuntu core. They still feel too heavy. phusion/baseimage is 188MB.

‍

‍

Conclusion

With Kubernetes, docker finally started to make sense to me. It's great because it provides great tools out of the box for doing web app deployment. Replication controllers, Services (with LoadBalancer included), Persistent Volumes, internal DNS. It should have all you need to make a resilient web app fast.

At Gorgias we're building a next generation helpdesk that allows responding 2x faster to common customer requests and having a fast and reliable infrastructure is crucial to achieve our goals.

If you're interested in working with this kind of stuff (especially to improve it): we're hiring!

We've released a new version of the Chrome Extension, with sharing features and a new navigation bar. We hope you'll love it!

‍

Share templates inside the extension

Before, the only way to share templates with your teammates was to login on Gorgias.io.

If you're on the startup plan, when you create a template, you can choose who has access to it: either only you, specific people, or your entire team.

‍

‍

The account management section is now available in the extension, under settings.

‍

New navigation

Tags are now available on the left. It's easier to manage hundreds of templates with them.
You can also navigate through your private & shared templates. Shared templates include templates shared with specific people or with everyone.

‍

‍

We hope you'll enjoy this new version of our Chrome Extension. As usual, your feedback & questions are welcome!

Today, we’re thrilled to announce that we’ve raised a $1.5 million Seed round led by Charles River Ventures and Amplify Partners, to help build our new helpdesk.

‍

We’re incredibly grateful to early users, customers, mentors we’ve met both at and Techstars.

‍

We started the journey with Alex at the beginning of 2015 with our Chrome extension, which helps write email faster using templates. We’ve been pleased all along with customers telling us about how helpful it was, especially for customer support.

‍

While building the extension, we’ve realized that a big inefficiency in support lies in the lack of integration between the helpdesk, the payment system, CRM and other tools support is using. As a result, agents need to do a lot of repetitive work to respond to customer requests, especially when the company is big.

‍

That’s why we’ve decided to build a new kind of helpdesk to enable customer support agents to respond 2x faster to customers. You can find out more and sign up for our private beta here.

‍

When a company has a lot of customers, support becomes repetitive. We want to provide support teams with tools to automate the way they treat simple repetitive requests. This way, they have more time for complex customer issues.

‍

We'll now focus on this helpdesk and on growing the team, oh, and if you'd like to join, we're hiring! We're super excited about this new helpdesk product. If you’re using the extension, don’t worry.

‍

Romain & Alex

We've been busy, but not deaf!

Last few months we got lots of feedback about our extension and found to our delight that most people are satisfied, but still a few recurrent issues came up:

• The HTML/WYSIWYG editor sucks.
• No support for Outlook.com.

‍

We listened and now we're presenting:

• A brand new editor
• Support for outlook.com
• More on the Rich-Text editor

‍

WYSIWYG editors for the web are notoriously buggy and are just difficult to develop.

‍

I have yet to see one that is bug free. There are few venerable editors that do a good job like TinyMCE, FKEditor or CKEditor.. but they are big and all have edge cases that break the intended formatting and add a lot of garbage html.

‍

There are newer good quality editors in town such as Redactor. The one that got my attention and finally landed in Gorgias is this wonderful editor called which is super lightweight, uses modern content-editable (no i-frames) and 'just works' most of the time. That's not to say it's perfect, but it's good enough and I'm satisfied with it's direction in terms of development.

‍

Enjoy it and as always send us bug-reports or feedback on: support@gorgias.com

TL;DR:

• Managing support across multiple channels without a system is the core problem, not the volume of tickets.
• A knowledge base of your most common questions is the foundation for everything else, from templates to AI tools.
• Automate in layers: start with auto-acknowledgment, then AI-assisted drafts, then full automation only for zero-judgment replies.
• When free tools stop keeping up, a helpdesk connects all your channels and gives you full customer context in one place.

If you're a small business owner handling support solo, you’re familiar with the following: questions coming in from every direction, hours spent typing out replies, and customers waiting days for an answer. What you’re missing is a system that holds it all together.

Below, we’ll walk you through how to build a reliable customer support operation solo, starting with free tools and simple processes, with options to scale as your business grows.

{{lead-magnet-1}}

Why customer support feels overwhelming for small businesses

Running support solo is hard for a few specific reasons:

• Every channel is its own inbox. Instagram DMs, email, WhatsApp — a customer's history is spread across different tools, so every interaction starts with zero context.
• Your best answers live in your head. Until they're written down, no new hire can help without escalating back to you.
• The same questions come in on repeat. Most small businesses find that nearly all of their support tickets are variations of the same handful of questions. Without a system, you're writing the same replies from scratch every time.
• There's no triage. Urgent issues and FAQs land in the same inbox with equal weight, so prioritizing takes judgment and time.

Most of the fixes in this guide address one of those problems directly. The fastest wins come from consolidating where support happens and getting your knowledge out of your head and into a format that works for you, and eventually, for the tools that can help you.

Related reading: Why consolidated doesn’t mean compromised: Top 3 myths debunked

Step 1: Pick one channel and own it

Before adding any tools or automation, identify where the majority of your inquiries come from. Pull a rough count over the last 30 days across your active channels. If one channel tops the list, that's your starting point.

If ticket volume is roughly the same across channels, choose whichever works best for your needs. Which channel lets you respond most effectively? Which one can keep detailed and retrievable records? Where do customers like to receive answers?

A quick overview of the top support channels: 

• Email supports longer answers, searchable history, and batching replies, meaning you can respond to 10 emails at once.
• DMs are faster but naturally become back-and-forth exchanges that can turn into lengthy message threads. Most platforms also don’t offer deep search beyond keyword matching.
• Live chat sets response time expectations that are hard to meet without a dedicated support agent.

For most small businesses, email is the stronger primary channel. Once you've chosen yours, redirect everyone to it: a link in your Instagram bio, a WhatsApp auto-reply pointing to your email, or a contact form confirmation that sets response time expectations.

You don't have to abandon your other channels. Just choose one as the place where conversations get resolved.

Related reading: How to implement an omnichannel customer service strategy

Step 2: Build your support knowledge base

The single most helpful thing you can do for your support operation is write down your best answers. Everything else — templates, AI tools, VA training, and customer-facing FAQs — depends on this foundation.

How to build it:

1. For one to two weeks, save every reply you send. Paste them into a running Google Doc or Notion page as you go.
2. Look for the patterns. Shipping timelines, return policies, pricing questions, order status, and product fit will come up again and again.
3. Turn the best version of each recurring answer into a clean, reusable write-up.

Efficiency tip: Use AI to speed this up. Export past support emails, chat histories, or even a rough doc of notes, and paste them into Claude or ChatGPT. Ask it to group recurring questions into categories and draft answers for each. What might take an afternoon of manual work takes just 20 minutes.

Put the knowledge base to work in two ways

Internally: This doc becomes the main reference for anyone or any tool, answering on your behalf. It's the foundation for the canned responses in the next section, and the training material if you ever bring on a new teammate.

Externally: A customer-facing FAQ reduces inquiries before they happen. If a customer can answer their own question at midnight without waiting for you, that's one less ticket in your inbox. Add FAQs to your website footer, a dedicated FAQ page, or directly on product pages for questions specific to that item.

Tip for social-heavy brands: Your FAQ content doesn't have to live only on your website.

• On Instagram, save FAQ content as dedicated Highlights (e.g., "Shipping," "Returns," "Sizing") so customers can find answers without DMing you.
• A pinned post on Instagram or Facebook can address your single most common question before it gets asked.
• On TikTok, pin your most-viewed FAQ-style videos to your profile so new visitors see them immediately.

Step 3: Set up canned responses and templates

Once your knowledge base exists, turn it into templates. This is the fastest operational win available to a one-person support operation.

In Gmail, canned responses let you insert a full reply with two clicks. Set them up for your top five to ten questions, and you'll cut your average reply time significantly. Most other email clients have an equivalent feature.

A few tips for making templates work:

• Start with your highest-volume questions, not your most complex ones. The goal is to reduce time spent on repetitive replies, not to script every possible scenario.
• Treat templates as a starting point. One sentence of personalization, using the customer's name or referencing their specific order, goes a long way toward making your pre-written replies feel human.
• Keep a master doc as your source of truth. Even if you're using Gmail templates, maintain a doc with all your answers in one place. It's easier to update, share, and eventually hand off.

You don’t need a formal template system yet. A Google Doc you copy from is still a real improvement over writing from scratch. Don't wait for the perfect setup.

Step 4: Add automation gradually, in the right order

The instinct when you're overwhelmed is to automate everything at once. The better approach is to layer it in, starting with the lowest-risk options and working your way toward AI tools only once you trust the outputs.

Layer 1: Automate acknowledgment

Set up an immediate reply on every channel that confirms you received the message and gives an expected response time. This one change reduces follow-up messages significantly. Customers don't need an instant answer, they need to know you're there. Most email clients and helpdesks offer this for free.

Layer 2: AI-assisted drafts

Tools that draft a reply for you to review before sending are the right entry point for AI in a small business support operation. You edit instead of write from scratch, which is meaningfully faster, and you stay in control of what goes out. This addresses the most common concern about AI in support: not that it can't help, but that it might say something wrong without you knowing.

Layer 3: Automated replies for zero-judgment tickets

Order confirmations, shipping notifications, and return receipts are fully automatable. The answers are always the same, no context is needed, and customers expect them to be automated. Start here before moving to anything more complex.

Layer 4: AI auto-responses for FAQs

At this point, you trust the accuracy of your knowledge base and are comfortable with how your AI tool responds to customers. Now, you can consider letting it automatically respond to straightforward FAQs. 

Be transparent about it. A simple label like "Hi, I'm a support assistant" sets the right expectation and reduces frustration if the answer misses. Only apply this layer when the previous three are working reliably.

What to look for when you're ready for a helpdesk

Free tools can only do so much. When you spend so much time managing customer support that the other parts of your business get neglected, it’s time to consider paid tools and services.

A helpdesk solves a specific problem that free tools can't: it connects your channels so every conversation, regardless of where it started, lives in one place with full context.

What a helpdesk gives you that a shared inbox doesn't:

• A unified inbox that pulls in email, social media DMs, WhatsApp, phone, and contact forms together
• Customer history attached to every ticket, so you can see past orders and past conversations before you reply
• Rules and routing that automatically tag, assign, or respond to certain ticket types
• Shared access for a VA or team member, without the risk of two people replying to the same ticket

Signs it's time to make the move:

• You're spending more than two hours a day on support
• Customers are following up because their original message got missed
• You're considering bringing on help, and you need a system they can work within

Helpdesks with free tiers or trials:

When evaluating any of these, prioritize integration with your ecommerce platform, ease of setup, and whether the AI features assist your replies or send automatically without review. For most small businesses just getting started, assisted drafting is the right fit before committing to full automation.

Quick wins you can implement today

These are the changes that take under an hour and make an immediate difference:

• Set an auto-reply on every active channel with your expected response time
• Start a Google Doc with your top 10 most common questions and your best answers to each
• Set up Gmail templates for your five most frequent replies
• Add a FAQ section to your website, even a short one
• If Instagram or TikTok are primary channels for your audience, repurpose your FAQ content as Highlights or pinned posts
• Pick one secondary channel and add a redirect message pointing customers to your primary one

None of these require a paid tool or a technical setup. They just require an hour of focused work and the recognition that a small amount of structure now saves a large amount of time later.

Get the foundation right, and the tools will follow

Good support at small scale comes down to sequence, not software. Consolidate your channels, document your best answers, build templates, and layer in automation only once the basics are working. Each step makes the next one easier.

When free tools stop keeping up, Gorgias connects your channels, integrates with Shopify and BigCommerce, and includes AI features that assist your replies rather than replace your judgment. 

Start a free trial now →

TL;DR:

• AI agent ROI is about business impact, not just ticket deflection, so your report should connect automation to cost savings, team capacity, customer experience, and revenue.
• Executives care about five core metrics: cost savings, automation success rate, customer experience impact, team capacity, and revenue impact.
• Quality matters as much as speed, because faster replies only prove ROI when AI resolves issues accurately and avoids repeat contacts or unnecessary handoffs.
• A strong AI ROI report explains the math, including AI-handled interactions, time saved, cost per human-handled ticket, AI tool costs, and revenue attribution rules.
• AI ROI reporting should become a regular workflow, with daily performance checks, weekly handover reviews, and a monthly business summary for leadership.

Your AI agent is answering tickets, but leadership wants proof that it’s paying off.

That proof can’t stop at ticket deflection or faster replies. To show real AI agent ROI, you need to connect automation performance to cost savings, team capacity, customer experience, and revenue impact.

This guide breaks down the metrics that matter, how to calculate them, and how to turn AI reporting into a business case executives can understand.

Why proving AI agent ROI is harder than it sounds

AI agent ROI is hard to prove because most teams measure activity, not impact.

Ticket deflection doesn’t always mean resolution: A deflected ticket is not always a solved problem. A customer may abandon the conversation, ask the same question later, or contact your team through another channel.

Automation rate needs context: A high automation rate can look impressive in a report. But it needs to be paired with metrics like CSAT, handover rate, repeat contact rate, and resolution time to show whether AI is handling the right tickets well.

Speed can hide quality issues: AI can reduce FRT and resolution time quickly. But fast answers only prove ROI when they’re accurate, helpful, and complete.

Cost savings need a clear calculation: Leadership needs to know how your team calculated savings. That means connecting automated interactions to agent time saved, average handle time, cost per ticket, and AI tool costs.

Revenue impact is easy to miss: AI agents can influence purchases, recommend products, or recover carts. Those results are harder to prove when AI reporting, support data, and ecommerce data live in separate tools.

ROI needs a complete view: No single metric proves AI agent ROI. The strongest reports connect efficiency, customer experience, team capacity, and revenue impact.

The 5 metrics executives actually care about

To prove AI agent ROI, focus on metrics that connect AI performance to business outcomes.

Leadership does not need every AI stat in your dashboard. They need to know whether AI is lowering costs, helping the team scale, protecting customer experience, and contributing to revenue.

1. Cost savings

Executives care about whether AI is reducing the cost of support without creating more work somewhere else.

Track:

• automated interactions
• time saved by agents
• cost per human-handled ticket
• cost per AI-handled interaction

Cost savings show how much money your AI agent saves by handling customer interactions instead of a human agent.

Show how many interactions AI handled, what those interactions would have cost your team, and what it costs for AI to handle them instead.

AI ROI is not just about cutting costs. It’s about helping the business handle more volume without increasing support costs at the same pace.

2. Automation success rate

Is your conversational AI actually giving customers correct, high-quality answers?

Track:

• AI automation rate
• success rate
• handover interactions
• repeat contact rate 

Automation success rate shows whether your AI agent is actually resolving customer interactions without human help.

A high automation rate with high escalations may indicate poor AI quality.

A lower automation rate with strong CSAT and fewer repeat contacts may show that AI is handling the right tickets well.

The best AI programs optimize for successful resolution, not maximum automation.

3. Customer experience impact

AI should improve efficiency without hurting the customer experience.

Track:

• customer satisfaction score (CSAT)
• first response time (FRT)
• resolution time
• human response time after AI handoff
• repeat contact rate

Customer experience metrics show whether customers are getting faster, helpful support from AI.

Speed is not the same as quality.

AI can reduce FRT and resolution time, but those gains only matter when customers still get accurate, complete answers.

The strongest AI reports show that customers got help faster and still had a good experience.

4. Team capacity

Executives care about whether AI helps the team scale.

Track:

• time saved by agents
• automated interactions
• closed tickets
• coverage rate
• workload handled by AI

Team capacity shows how much repetitive work AI removes from the queue.

This matters because human agents can spend more time on complex issues, high-value customers, retention risks, and revenue-generating conversations.

Team capacity is not the same as headcount reduction.

A stronger story is that AI helps the same team handle more customer demand without adding the same amount of cost or pressure.

5. Revenue impact

Executives care about whether AI contributes to revenue, not just cost savings.

Track:

• revenue influenced
• orders influenced
• revenue per interaction
• AOV
• purchase rate from AI-assisted conversations

Revenue impact shows whether AI helps shoppers choose products, get answers before purchase, use discounts, or recover carts.

Revenue attribution needs a clear window.

Explain how your team defines an AI-influenced purchase, such as an order placed within a set number of days after an AI-assisted conversation.

AI agents are becoming part of the shopping experience, not just a way to reduce support tickets.

How to build an executive AI ROI report

An executive AI ROI report should show what changed because of your AI agent.

Start with the outcome leadership cares about most, then add the proof underneath.

Step 1: Lead with the headline result

Start with the clearest business impact. That might be cost saved, time saved, revenue influenced, or tickets resolved without human help.

For example: “Our AI agent resolved 8,000 interactions this month and saved the team 420 hours.”

This gives leadership the answer before they have to interpret the data.

Step 2: Show how you calculated it

Explain the math behind the headline result.

If you’re reporting cost savings, show the number of AI-handled interactions, average cost per human-handled ticket, and cost per AI-handled interaction.

This makes the number easier to trust.

Step 3: Add quality checks

Next, prove the AI agent is not just handling volume.

Add success rate, handover interactions, CSAT, and repeat contact rate if available.

This shows whether AI is solving issues well, not just removing tickets from the queue.

Step 4: Connect the result to team capacity

Then show how AI changed the team’s workload.

Use time saved, automated interactions, and queue impact to show whether agents had more time for complex issues, retention risks, or sales-focused conversations.

This turns AI reporting into an operations story.

Step 5: Add revenue impact if AI supports shopping

If your AI agent answers pre-purchase questions, include revenue metrics.

Show revenue influenced, orders influenced, revenue per interaction, and AOV.

Make the attribution window clear, such as purchases made within three days of an AI-assisted conversation.

Step 6: End with what improves next

Close the report with the actions your team will take next.

That might include updating AI instructions, improving handoff rules, filling help center gaps, or reviewing low-CSAT conversations.

This shows leadership that AI performance is being actively managed, not passively monitored.

Common mistakes teams make when reporting AI ROI

AI ROI reporting falls flat when the numbers look impressive but don’t answer the real business question.

Avoid these common mistakes when you’re building your report.

1. Leading with ticket deflection

Ticket deflection is useful, but it doesn’t prove ROI on its own.

A ticket can be deflected without being resolved. Pair deflection with success rate, CSAT, handovers, and repeat contact rate to show whether AI actually solved the issue.

2. Treating automation rate like the goal

A higher automation rate is not always better.

The goal is to automate the right conversations well. If automation rate rises while handovers, repeat contacts, or poor CSAT scores also rise, your AI agent may be creating hidden work.

3. Ignoring the handoff experience

A handoff is not a failure when it gets the customer to the right person faster.

But leadership should know what happens after AI escalates a ticket. Track human response time after AI handoff so you can spot delays, routing gaps, or tickets that need clearer escalation rules.

4. Reporting cost savings without the math

Cost savings need context.

Show how you calculated the number, including AI-handled interactions, average cost per human-handled ticket, agent time saved, and AI tool cost. This makes the ROI story more credible.

5. Leaving revenue out of the report

AI agents can do more than reduce support volume.

If your AI agent helps shoppers choose products, answers pre-purchase questions, recommends SKUs, or offers discounts, include revenue impact. Executives need to see where AI supports both efficiency and sales.

6. Measuring AI in a separate tool

AI reporting gets harder when support data, automation data, CSAT, and revenue live in different systems.

Disconnected reporting makes it harder to prove what changed because of AI. A stronger setup gives your team one place to track AI performance across support, customer experience, and revenue.

How to build AI ROI reporting into your day-to-day workflow

AI ROI reporting works best when it becomes a regular operating habit.

A monthly report can show leadership the results, but your team needs a daily and weekly rhythm to understand what’s improving, what’s breaking, and where AI needs coaching.

• Check performance daily: Review automated interactions, handovers, success rate, CSAT, and any unusual spikes in ticket volume. This helps your team catch issues early, before they show up in an executive report.
• Review handovers weekly: Look at the most common reasons AI escalated tickets to a human agent. Then decide whether the fix is better AI instructions, clearer help center content, more accurate product data, or a new handoff rule.
• Connect AI performance to team workload: Track whether AI is actually giving agents more capacity. Look at time saved, queue volume, AHT, and the types of tickets agents are handling.
• Watch quality, not just automation: Pair automation rate with CSAT, repeat contact rate, handovers, and resolution time. This helps your team avoid chasing a higher automation rate at the expense of customer experience.
• Share a monthly business summary: Turn the daily and weekly signals into a simple leadership update. Show what AI handled, what it saved, how it affected customers, how it affected the team, and what your team is improving next.

See AI agent ROI in one place

Proving AI ROI gets harder when your support, automation, and revenue data live in separate tools.

Gorgias’s AI Agent brings AI-specific reporting into the same helpdesk your team uses every day, so you can track what AI handled, what it saved, and how it contributed to the customer experience.

Book a demo to see how AI Agent helps ecommerce teams measure and improve AI support from one customer experience platform.

{{lead-magnet-1}}

TL;DR:

• Ticket volume is your support workload: It counts every customer inquiry across every channel in a given time period.
• High volume signals friction in your business: Spikes usually point to unclear policies, product issues, or gaps in your website experience.
• Every ticket has a real cost: Agent time, tooling, and overhead add up fast — and they compound during peak seasons.
• Automation reduces volume without reducing quality: AI tools and self-service deflect repetitive questions while keeping customers satisfied.
• Measurement drives improvement: Tracking volume by channel, category, and time period reveals exactly where to focus your efforts.

Your ticket volume number is probably wrong. If customers are reaching you through email forwards, Slack DMs, or channels that bypass your helpdesk, those tickets aren't being counted, and your SLA reporting is built on incomplete data. This guide covers how to get an accurate count, break it down by channel and category, and use your vertical benchmark to figure out whether your volume is actually a problem or just normal for your industry.

What is ticket volume?

Ticket volume is the total number of customer inquiries your support team receives across all channels — email, live chat, phone, social media, and contact forms — within a specific time period. It is the most direct measure of your team's workload.

Do not confuse it with contact rate. Contact rate = tickets ÷ orders (or customers). That normalized number is more useful for benchmarking and planning because it accounts for business growth. Raw ticket volume tells you how busy your team is. Contact rate tells you whether support demand is outpacing your business.

How to calculate your ticket volume

Start by looking at the last 30 days of customer conversations, no matter where they currently live.

Pull these four numbers:

• Total customer questions received across all channels
• Breakdown by channel (email, chat, social DMs, phone, contact forms, etc.)
• Breakdown by category (shipping, returns, product questions, account issues)
• Tickets or conversations per order during the same period — this gives you your contact rate baseline

Here’s how to pull that data depending on your setup:

Gmail or Outlook

Open your inbox or Sent folder and filter by the last 30 days. Count how many customer conversations came in during that period. You can also copy subject lines into ChatGPT or Claude to group conversations by topic.

Shopify Inbox

Go to Inbox > Conversations and review your recent conversations. Count how many messages you received and look for repeated themes or questions.

Any helpdesk (Gorgias, Zendesk, Freshdesk, Help Scout, etc.)

Most helpdesks have ticket reporting or exports built in. Search “export tickets” or “ticket report” in your platform’s help center. From there, you can pull:

• Total tickets
• Channel breakdown
• Top ticket categories
• Tickets over time

If a large portion of customer questions are still happening in untracked places like Slack DMs, personal inboxes, or Instagram comments, your reporting is incomplete. Before optimizing support operations, route customer conversations into one shared system so you can accurately measure volume, response times, and recurring issues.

Why your volume breakdown matters more than the total

A raw ticket count tells you how busy your team is. The breakdown tells you what to fix.

Run this categorization for your last 30 days. Your top two or three categories are your highest-leverage targets.

Track volume alongside these KPIs

Ticket volume only tells part of the story. Track it alongside:

• Contact rate (tickets ÷ orders) — so you know if volume is growing faster than your business
• First response time (FRT) — volume spikes show up here first
• Average handle time (AHT) — high AHT + high volume = a capacity problem
• Cost per ticket — total support costs ÷ total tickets, the clearest financial measure
• Backlog size — a growing backlog is the earliest warning sign that volume is outpacing capacity
• Deflection rate — tickets resolved through self-service or automation without agent involvement

How to reduce ticket volume without reducing quality

Once you know what is driving your volume, address each category at the source. The goal is to eliminate unnecessary tickets.

Automate the highest-volume, lowest-complexity tickets first. WISMO inquiries, order status checks, and basic return initiations require no agent judgment. An AI Agent connected to your ecommerce platform can handle these end-to-end without a human stepping in. When a question is too complex, the AI escalates it with full context attached.

Build self-service content around your top categories. A Help Center that directly addresses your most common ticket types is the highest-leverage tool for sustained volume reduction. Start with your top five categories. Write one article per category. Surface those articles on relevant product pages, in checkout, and in post-purchase emails — before customers need to search.

Send proactive messages at the moments that generate the most tickets. Post-purchase is the single highest-value touchpoint: an order confirmation that includes a tracking link, estimated delivery window, and a clear link to your return policy eliminates a large share of inbound questions before they are ever submitted.

Measure deflection, not just volume. Deflection rate, the percentage of issues resolved through self-service or automation, is the metric that tells you whether your volume reduction efforts are actually working. Track it weekly alongside CSAT for automated interactions to make sure quality is holding.

Ticket volume benchmarks

The all-industry average is not your benchmark. Ticket volume per 100 orders varies 2.4x across verticals, so comparing yourself to a cross-industry number will either make you complacent or create false urgency.

According to Gorgias platform data from March 2026 across 14 verticals at the $10M GMV band, here is what tickets per 100 orders actually looks like by vertical:

Source: Gorgias Ecom Lab, March 2026

High ticket volume is not always a sign of poor CX — it often reflects product complexity. Electronics brands generate nearly one ticket per two orders because customers have more pre- and post-purchase questions about technical products. Food and Beverage brands generate about one in five. That gap is not a performance difference; it is a category difference.

The right question is not "are we below 10 tickets per 100 orders?" It is "are we above or below our vertical peers?" Find your row. That is your baseline. Then use the reduction tactics above to move below it.

How to predict ticket volume if your tool charges per ticket

If your ticketing tool uses usage-based pricing, where your bill scales with ticket volume rather than agent headcount, forecasting volume directly affects your budget.

The core formula is simple:

Projected tickets = projected orders × (tickets per 100 orders ÷ 100)

So if you expect 2,000 orders next month and your vertical median is 22 tickets per 100 orders, your forecast is approximately 440 tickets.

But a flat monthly estimate misses the real risk: peak seasons. A volume spike during BFCM that triples your order volume will also triple your ticket count — and your bill — unless you have guardrails in place.

To build a more accurate forecast:

• Use your contact rate, not raw volume. Divide your tickets by orders for each of the last 12 months. This gives you a stable ratio that accounts for business growth and seasonal swings.
• Apply that ratio to your order forecast. If your marketing team has a sales projection for November, multiply it by your contact rate to estimate support volume.
• Separate your AI-handled tickets from agent-handled tickets. Some platforms bill differently for automated resolutions versus human ones. If you're using an AI Agent to deflect WISMO and returns, those deflected tickets may not count toward your billable volume at all — which changes the math significantly.
• Build in a buffer for peak periods. Your contact rate tends to rise during high-demand periods, not just your order volume. First-time customers generate more tickets than repeat buyers, and BFCM brings a disproportionate share of first-timers.

Before signing any usage-based contract, ask two questions: What counts as a billable ticket? And is there a hard cap on monthly charges? Variable billing only works in your favor if you have clear definitions of what triggers a charge and a ceiling on how high costs can go during an unexpected spike.

If your platform bills per ticket resolved by a human agent (not AI), your deflection rate becomes a financial metric, not just an operational one. Every percentage point of additional deflection directly reduces your bill.

Start reducing ticket volume today

Begin by identifying your top ticket categories, then work backward to find the root cause of each one.

From there, layer in self-service content, automation, and proactive messaging to address those root causes directly. The result is a support operation that handles more customers and a team that spends its time on the work that actually requires human judgment.

Book a demo to see how Gorgias helps ecommerce brands reduce ticket volume and improve customer experience at the same time.

{{lead-magnet-2}}

TL;DR

• Eight metrics define support performance: CSAT, NPS, CES, FCR, FRT, AHT, TTR, and SLA adherence are the core benchmarks to track.
• First response time varies 5.5x across ecommerce verticals — your vertical benchmark matters more than any industry average.
• Benchmarking is a six-step process: Define goals, select metrics, source benchmarks, capture data, analyze gaps, and set targets.
• Benchmarks connect support to revenue: Faster response times and higher resolution rates directly affect repeat purchases and customer retention.
• Brands at 30%+ automation respond 10x faster than brands at zero — with the same or better CSAT.

The benchmarks in this article are drawn from the Gorgias Ecom Lab, a research hub that publishes platform-level behavioral data from thousands of ecommerce brands. Where we cite a specific figure, it comes from that data, not generic industry surveys.

What is customer service benchmarking?

Customer service benchmarking is the practice of comparing your support performance to industry standards or peer data to find gaps and set improvement targets.

It means measuring how your team performs on metrics like response time and satisfaction scores — then checking those numbers against what similar businesses achieve. The goal is not just to measure. It's to create a clear picture of where you stand and what to fix first.

Benchmarking has four core components:

• Baseline measurement: Recording your current performance across all support channels before making any changes.
• Peer comparison: Looking at how brands of similar size and type perform on the same metrics.
• Gap analysis: Pinpointing where your numbers fall short of industry targets.
• Target setting: Using benchmark data to define realistic, time-bound goals for your team.

Why customer service benchmarks matter for ecommerce brands

Benchmarks turn "we need to improve support" into a specific, actionable goal. Instead of a vague directive, you get a clear target: reduce email response time from 48 hours to 24, or lift CSAT from 72 to 82 percent.

Here's what most benchmarking guides won't tell you: the all-industry average is often the least useful number in the room.

Ecom Lab data across 14 ecommerce verticals shows that first response time varies 5.5x at the same $10M GMV band — from 1.6 hours in Hardware to 8.8 hours in Apparel. A brand sitting in the middle of that range looks fast against one peer and slow against another. Without vertical context, the comparison tells you nothing.

For ecommerce brands, support performance also connects directly to revenue. Customers who get fast, accurate answers are more likely to complete a purchase and come back again. That shows up across the whole business — from support team efficiency to operations, finance, and marketing.

Core customer service benchmarks to track

These eight metrics are the foundation of support performance measurement. Each one captures a different dimension of the customer experience, from speed to ease to loyalty.

First response time

First response time (FRT) is the time between a customer sending a message and receiving your team's first reply.

This is the metric customers feel most immediately — and the one with the most variation across ecommerce brands. According to the Ecom Lab report Stop Benchmarking Against the Average, FRT varies 5.5x across 14 ecommerce verticals at the same revenue band. CSAT, by contrast, varies by just 0.2 points across those same verticals. If you want to know whether your operation is ahead of your peers, FRT is the metric that will tell you.

General targets by channel:

• Email: Under 24 hours (excellent: under 12 hours)
• Live chat: Under one minute (excellent: under 30 seconds)
• Social media: Under two hours (excellent: under one hour)
• Phone: Under three minutes hold time

AI automation changes what's achievable here. Brands automating close to zero percent of tickets average 736-minute response times. At 30% automation, that drops to 80 minutes. At 40%, 12 minutes.

The gains don't scale evenly — they accelerate. If your team has deployed any AI, your FRT benchmark should reflect your automation rate, not just your channel.

First contact resolution

First contact resolution (FCR) measures the percentage of tickets resolved in a single interaction, without the customer needing to follow up.

A high FCR means your team has the right information and authority to solve problems on the first attempt. The industry target is 70 to 75 percent across all channels.

For teams running AI, there's a more actionable metric to track alongside FCR: AI Resolution Rate — the share of AI-touched tickets that close end-to-end without any human involvement.

Ecom Lab data shows the median ecommerce brand resolves 45% of AI-touched tickets end-to-end. The top quartile reaches 65%. Every point of improvement removes a 10-hour median wait and a one-in-three abandonment risk from a customer's experience.

Read more: First contact resolution rate: Your guide to understanding the metric

Customer satisfaction score

Customer satisfaction score (CSAT) measures how satisfied a customer was with a specific support interaction, typically through a short post-conversation survey.

The standard benchmark is 80 to 85 percent. Ecommerce brands with proactive, personalized support often reach 85 to 90 percent. For a deeper look at moving that number, see How to Improve CSAT: 8 Fixes That Make a Real Difference.

Two things stand out from Ecom Lab data. First, CSAT is remarkably stable across verticals — it varies by only 0.2 points at the same revenue band, so your category matters far less here than it does for FRT.

Second, there is a modest tradeoff early in AI adoption: brands at 20% automation average 87.9% CSAT, versus 90.3% at zero. This reflects AI encountering more diverse ticket types as coverage expands. Brands that move past 30% automation and properly configure their AI bring CSAT back up while keeping response times fast.

Net promoter score

Net promoter score (NPS) asks customers how likely they are to recommend your brand to someone else, scored on a scale of zero to 10.

It reflects the overall customer experience — not just a single interaction. The ecommerce benchmark is between 30 and 50, with anything above 50 considered strong. See How To Calculate Net Promoter Score for the full methodology.

Customer effort score

Customer effort score (CES) measures how easy it was for a customer to get their issue resolved, typically on a seven-point scale.

Lower effort correlates with higher repeat purchase rates. The industry benchmark is 5.5 or higher.

The single biggest driver of high-effort experiences is the handoff wait. According to the Ecom Lab report The Cheapest Ticket Is the One a Human Never Touches, the median wait between an AI handing off and a human responding is 10 hours. At the 90th percentile, that wait hits 71 hours — three full days. And a third of handed-off tickets never receive a human response at all.

That experience is what drives CES scores down. Reducing handoffs, not just handling them faster, is the most direct path to a better effort score.

Average handle time

Average handle time (AHT) is the total time an agent spends on a live interaction, including hold time and wrap-up work.

It measures efficiency without accounting for quality, so it works best alongside CSAT and FCR.

Targets by channel:

• Phone: 6 to 8 minutes
• Live chat: 5 to 7 minutes

As automation rate rises, AHT on human-handled tickets typically drops — AI absorbs the simple volume and leaves agents with a shorter, more focused queue. Ecom Lab data shows that at 50%+ automation, AI does the equivalent work of 6.3 full-time agents while the human team at that tier averages just 3 people. Those agents handle 29% more tickets per month and spend more time on the complex issues that actually require judgment.

Time to resolution

Time to resolution (TTR) is the total time from when a ticket opens to when it fully closes.

Unlike FRT, which measures only the first reply, TTR captures the entire support interaction. For a closer look at this metric and how to reduce it, see Resolution Time: What It Is and How to Reduce It.

General targets by complexity:

• Simple inquiries: Under 24 hours
• Complex issues: 24 to 72 hours
• Technical problems: 3 to 5 business days

For brands with AI in the mix, Ecom Lab data gives channel-level baselines for tickets that require human involvement. Contact form handoffs resolve in a median of 36 hours with a 42% abandonment rate. Email handoffs resolve in 32 hours and abandon 30% of the time. Chat resolves in 8 hours and abandons 13%, because real-time pressure forces faster responses.

These aren't just benchmarks to optimize — they're the cost of every ticket that doesn't resolve end-to-end.

Service level agreement adherence

A service level agreement (SLA) is a defined commitment to respond to or resolve tickets within a set timeframe.

SLA adherence measures the percentage of tickets where your team meets that commitment. The industry benchmark is 90 to 95 percent compliance. For the tactics that make hitting those commitments repeatable, see SLA Best Practices for Effective Support Ticket Management.

How to benchmark customer service performance

Benchmarking works best as a structured process, not a one-time audit. These six steps take you from identifying what to measure to building a plan for improvement.

Step 1: Define your goals

Start by deciding what you want to improve and why.

Are customers complaining about slow responses? Are agents spending too long on simple tickets? Tying your benchmarking effort to a specific business problem keeps the process focused and the results actionable. Limit your initial scope to three to five metrics — tracking everything at once makes it harder to act on what you find.

Step 2: Choose the right metrics

Pick metrics that match the problem you identified in step one.

If customers are frustrated by how long it takes to get help, FRT and TTR are your starting points. If satisfaction scores are slipping, CSAT and CES will tell you more. Match the metric to the pain point.

If you have any automation running, add AI Resolution Rate to your list. The median brand sits at 45%; the top quartile is at 65%. A gap between your rate and the top quartile almost always comes down to one of four things: limited intent coverage, insufficient action authority (AI can't issue refunds or apply discounts), missing system integrations, or an escalation policy that's routing too much to humans by default.

Step 3: Source your benchmarks

You need external data to compare against. Reliable sources include:

• Industry reports: Published by customer service associations and research firms.
• Helpdesk platforms: Many publish benchmark data from their customer base, including Ecom Lab.
• Peer networks: Non-competing brands in adjacent markets often share performance data.
• Customer surveys: Direct feedback on what customers expect from your support.

One critical caveat: filter by your specific vertical, not just "ecommerce." Ticket volume per 100 orders varies nearly as much as response time. Electronics brands generate about 46 support tickets per 100 orders. Food & Beverages brands generate about 20. If you're in a high-ticket-volume vertical, that's your baseline — not a problem to fix.

Step 4: Measure your current performance

Pull at least three months of data from your helpdesk to establish a reliable baseline.

Shorter windows get skewed by seasonal spikes or one-off events. Make sure you're measuring each metric the same way across all channels so the data is consistent. How to Evaluate the Effectiveness & Impact of Your Customer Service Team is a good companion resource for this step.

Step 5: Identify and diagnose gaps

Compare your numbers to the benchmarks you sourced.

Look for patterns. Are certain channels consistently slower? Do specific ticket types take longer to resolve? The goal is to understand why the gap exists, not just that it does.

Step 6: Set targets and build an action plan

Use your gap analysis to set incremental targets.

If your email FRT is 48 hours and the benchmark is 24, aim for 36 hours first. Assign ownership, set a timeline, and schedule a review date. Benchmarking only drives improvement when it leads to a concrete next step.

What good benchmarking looks like in practice

Benchmarking changes how your team operates day to day. Agents know what "good" looks like and can measure their own progress against it. Managers can identify coaching opportunities using real data rather than observation alone.

For ecommerce brands specifically, the operational benefits compound over time:

• Lower cost per ticket: Efficient workflows reduce the time and resources needed to resolve each issue.
• Higher repeat purchase rates: Customers who get fast, easy support are more likely to buy again.
• Reduced cart abandonment: Quick answers to pre-purchase questions keep shoppers moving toward checkout. See Reduce Cart Abandonment: Proven, Data-Backed Strategies.
• Stronger team accountability: Clear targets give agents and managers a shared definition of success.

The financial picture from the Ecom Lab report Most Brands Are Overpaying for Support is concrete. Even at the lowest automation tier, brands net $73K per year after platform costs.

Nearly 1 in 4 brands (23.5%) reduced their support team after enabling Gorgias AI Agent. Of those, 51% achieved all three outcomes at once: fewer people, same ticket volume, same or more revenue. Brands that reduced by at least one person saw each remaining agent handle 29% more tickets per month while revenue grew 22%.

The adoption gap matters too. Only about 1 in 5 ecommerce brands has deployed AI in customer-facing support today. Brands at near-zero automation average 736-minute response times. Brands at 30%+ automation average 80 minutes.

That's not an incremental improvement. It's a structural shift.

Improve your benchmarks with Gorgias

Benchmarks tell you where to focus. The right tools help you get there.

Gorgias gives ecommerce support teams a unified view of every customer conversation, with built-in reporting that tracks the metrics that matter most. AI Agent resolves routine tickets automatically — across email, chat, and SMS — so your team spends less time on repetitive requests and more time on work that requires a human touch.

Book a demo to see how Gorgias helps ecommerce brands hit and exceed their customer service benchmarks.

{{lead-magnet-2}}