Is your store leaking revenue?
Find out exactly where you're losing sales — takes 2 minutes.
Why the way you score this metric matters more than the number itself
What Is a Customer Effort Score?
One question. Five choices. One number.
Customer effort score (CES) is a support metric measured by asking customers, immediately after a ticket closes, to rate agreement with one statement: "It was super easy to get your issue resolved today." The scale is five-point Likert: strongly disagree, disagree, neutral, agree, strongly agree. The score is top-box only, meaning the percentage who picked "strongly agree." Nothing else counts.
That last sentence is where most stores get it wrong.
Nearly every article on this keyword describes CES as an average on a 1-to-7 effort scale, borrowed from the original 2010 Harvard Business Review research by Matthew Dixon, Karen Freeman, and Nick Toman. That version is real, and it's the one most helpdesk tools ship by default.
But there's a second, stricter version some of the best-run DTC support teams actually use: top-box only. Not "agree or strongly agree" combined. Not an average score. Just the percentage who chose the single most positive option.
The two versions are not comparable. A 90% top-box score is a different animal from a "6.2 average" on a 7-point scale, and if you quote one benchmark against the other, you'll draw the wrong conclusion about your own team.
We audit Shopify stores for a living. The support metric section of that audit is where we see the confusion most: a store proudly reports "85% satisfaction" from a CSAT survey and assumes that's the same conversation as CES. It isn't. Let's separate the three metrics properly, then get into the part that actually changes how you run your team.
Is CES Better Than CSAT or NPS for Measuring Support?
Wrong question gets you a wrong metric.
CES, CSAT, and NPS measure three different things, and answering the wrong one wastes the survey. CSAT measures satisfaction with an outcome and is easy to inflate with a discount or refund. NPS measures brand affection, which has little to do with a single support interaction. CES measures how seamless the fix was when something broke: the actual job description of a support team.
Here's the breakdown, side by side:
| Metric | What it actually measures | Best used for | Easy to game? |
|---|---|---|---|
| CES (top-box) | Effort required to resolve a single issue | Per-ticket support quality | Harder: one binary "strongly agree" answer |
| CSAT | Satisfaction with the outcome | Post-purchase or post-interaction sentiment | Yes, a concession or discount inflates it |
| NPS | Likelihood to recommend the brand or product | Overall brand/product health | Not support-specific, measures the wrong layer |
Source: Qualtrics CES methodology, cross-referenced against WebMedic's client audit data.
The practical rule: keep NPS on the product side, scoped to brand and product quality. Run CES on every closed support ticket. Don't try to make one metric do both jobs. That's how a store ends up staffing decisions off a number that's answering a question nobody asked.
When Should You Send a CES Survey and How Often Should You Review It?
Timing decides whether the number means anything.
Send the CES survey immediately after a ticket closes, while the interaction is still fresh, not the next day and not in a weekly digest. Review the aggregate score monthly, tracked alongside two supporting metrics: ticket volume and repeat-contact rate. A single week of data on a small store is too noisy to act on.
Practically, this means your helpdesk (Gorgias, Zendesk, Gladly, whatever you run) fires the one-question survey the moment a ticket status flips to "resolved." No delay, no batching. The closer the survey sits to the moment of resolution, the more accurately it reflects the actual interaction rather than the customer's mood three days later.
Monthly review is the cadence, not because weekly is wrong, but because a Shopify store doing a few hundred tickets a month needs the larger sample to see a real trend instead of noise from one bad shift.
Here's a rule of thumb we use with clients: if your CES is below 90% top-box, or you're not tracking it at all, fix that first. Every other support initiative (automation, AI agents, upsell mechanics) is premature until you know whether the basic job of resolving the issue is actually happening cleanly. You can't optimize a number you're not measuring, and you can't layer revenue mechanics on top of a support experience that's currently frustrating people.
Does Support Over WhatsApp Change How You Send a CES Survey?
Malaysia and Singapore don't run support over email.
In Malaysia and Singapore, WhatsApp is the dominant customer support channel for DTC brands, which means the CES survey has to work as a single follow-up message inside the same chat thread, not a separate email link. A one-question, top-box CES check sent as the next WhatsApp message after resolution gets meaningfully higher response rates than an email survey sent hours later.
The mechanics are simple. Once an agent marks the WhatsApp conversation resolved:
- The system sends one follow-up message in the same thread: "Quick check: it was super easy to get your issue resolved today. Reply 1 (strongly disagree) to 5 (strongly agree)."
- The reply comes back in the same chat the customer already has open. No link, no new tab, no login.
- Only a reply of "5" counts toward the top-box score. A "4" is filed as a near-miss worth reviewing, not counted in the headline number.
Compare that to the standard email flow: ticket closes, customer gets an email 10 minutes later, has to open it, click through, and complete a form on a separate page. Every extra step loses respondents, which is the same friction principle the metric itself is trying to measure, applied to the measurement tool.
Does this sound like your store? Find out where you're leaking revenue. Take the free Revenue Score. 3 minutes. Free. No pitch.
Can a Customer Support Team Actually Be a Profit Center?
Most founders assume the answer is no before they've checked.
A support team can be classified as a profit center using a two-number test: revenue that can be directly attributed to the support team, minus the cost of running that team. When the first number exceeds the second, the classification holds. When the result is unclear, that means the attribution system is incomplete, not that support failed to create value.
We didn't invent this test, but we did put it in a formula because founders think in formulas. This is WebMedic's own framing of the arithmetic, not a formula stated outright anywhere else:
care-team contribution = care-attributed revenue - care-team operating cost
The underlying two-number comparison is the useful part. Care-attributed revenue means tracked purchases, upgrades, renewals, or referrals that trace back to a support conversation. Care-team operating cost means salaries, tooling (Gorgias, Zendesk, whatever's in your stack), and management overhead.
Here's the part most stores get wrong: if you run this test and the answer comes back murky, and you genuinely can't tell whether support paid for itself, the honest conclusion is that your tracking is incomplete, not that support is a cost center. A support agent who talks a hesitant buyer off the fence, or upsells a plan upgrade mid-conversation, generated real revenue. If nothing in your system tagged that conversation to that purchase, the value existed and your dashboard just didn't see it.
Two guardrails before you run this test on your own numbers:
- There's no universal attribution window, profit threshold, or staffing ratio that applies to every store. Set those from your own records: your average order value, your ticket volume, your team's cost structure. A framework can give you the test. It can't give you your numbers.
- Don't manufacture a positive result by crediting support for every retention or product event that happens near a support interaction. If a customer would have reordered anyway, that's not care-attributed revenue. Sloppy attribution here defeats the entire point of running the test honestly.
What Are the Levers for Turning Support Into a Profit Center?
Eleven specific moves, split two ways.
There are eleven concrete levers for shifting a support team from cost center to profit center, split into two routes: increasing the value support generates, or lowering what it costs to run. Picking one lever from each route, with traceable evidence, beats trying to run all eleven at once.
Route 1: Increase the value support generates.
- Reduce churn. Keep customers and help them stay subscribed or renew.
- Increase loyalty. Turn a difficult interaction into stickiness. Service quality as a differentiator, not just damage control.
- Increase revenue. Help customers use parts of your product or catalog they haven't discovered, or earn a genuine referral once trust exists.
- Feed product with real customer evidence. Document the friction customers describe in tickets and route it into product or catalog decisions. Support sees problems before anyone else does.
- Find early adopters through support relationships. The customers who message you the most are often your best beta testers.
- Ask for small trust-based favors. Once real trust exists in a conversation, a testimonial or referral ask lands differently than a cold email request.
Route 2: Reduce what support costs.
- Use AI agents and automation carefully. Augment the existing team first. Segment which issues are safe to hand off before you hand off broadly.
- Automate repeated fixes. If the same issue keeps generating tickets, that's a signal to build a fix, not hire another agent.
- Build more resilient systems. Testing, QA, and process fixes upstream prevent the support contact from ever happening.
Both routes at once.
- Test usability with the care team before customers see it. Agents who talk to customers all day catch friction a design review misses.
- Use support as a career path. Agents who understand the product and the customer are natural candidates for product, sales-engineering, or account roles, which reduces the cost of hiring those roles externally.
Pick one lever from each list. Assign an owner, a customer segment, and a way to measure the result. Running all eleven with no owner and no measurement is how a good framework turns into a wall poster nobody executes.
What Is Contextual Plus One and How Does It Generate Revenue From Support?
The offer only works if the fix comes first.
Contextual Plus One is a support-to-revenue mechanism where an agent resolves the customer's stated issue, then identifies the larger goal behind it and offers one relevant piece of content, feature, or upgrade tied to that goal. The offer is tracked so the resulting click or purchase can be attributed back to the conversation. It is not a generic upsell script.
The chain looks like this:
Customer issue arrives
-> agent resolves the issue
-> agent (or AI) identifies the larger goal behind the issue
-> system surfaces relevant approved material or an opportunity
-> agent checks the fit
-> agent offers the added value
-> a tracked link or action records use and purchase
Say a customer messages about a sizing issue with a jacket. The agent resolves the exchange. The larger goal underneath the ticket: this customer is building a cold-weather wardrobe. The agent mentions the matching base layer that pairs with that jacket, with a tracked link. If the customer clicks and buys, that's a traceable Contextual Plus One conversion. If they don't, nothing was lost, because the issue was already resolved before the offer happened.
Two rules make this work instead of turning support into a sales channel customers resent.
Rule 1: Resolution first, always. If the original issue isn't resolved, you don't get to the offer. An unresolved problem is never the opening for a pitch. A customer who's still frustrated about a broken tracking link doesn't want a cross-sell. They want their package found.
Rule 2: Count the action, not the offer. Making the offer is not revenue. Only the click, the purchase, or the plan upgrade that follows counts as attributed revenue. If you start counting offers as wins, your two-number test from earlier becomes fiction.
We're not going to hand you a made-up attach rate for this mechanism. We don't have real client data on Contextual Plus One specifically yet, and pretending otherwise would be exactly the sloppy attribution problem this whole framework warns against. If you're running it, track your own click-through and purchase rate from day one. That's your real number, not a borrowed one.
What Numbers Actually Prove This Approach Works?
One brand, one interview, worth naming precisely.
Eli Weiss, Director of CX at Olipop, described sustaining a 90%+ top-box customer effort score through the Q4 2020 carrier-chaos period on DTC Pod episode 103 (around 2021). These are one operator's stated figures from one interview, not an independently audited industry benchmark. Treat them as a proof point, not a target you're guaranteed to hit.
Listen to the source interview: DTC Pod episode 103 with Eli Weiss.
The context matters as much as the number. Q4 2020 was the period when carrier delays, driven by pandemic shipping volume, were breaking delivery promises across the entire DTC industry. Sustaining a 90%+ top-box score through that window means the CES question was correctly isolating "was your issue resolved smoothly" from "was the carrier late," a distinction a lot of support teams fail to make in their own surveys.
A second, related figure from the same interview: Eli Weiss reported that 70%+ of customers who have been on subscription for a few months have swapped or skipped at least once, treating that as a leading indicator of longer-term retention rather than a red flag. Two separate figures, kept separate because they come from different parts of the source material. Subscription volume grew roughly 10x in the window Olipop's SMS swap/skip/cancel mechanic launched. A distinct claim, from a different artifact (Eli Weiss's March DTC-era Twitter post referenced in the same interview): Olipop doubled its subscriber base in 100 days.
None of these are benchmarks you should hold your own store to. They're a single well-run team's stated results, worth citing by name because attribution matters more than the number itself. If you find a version of this metric quoted anywhere without naming who said it and when, treat it with real skepticism.
Frequently Asked Questions
What is a good customer effort score?
There's no single "good" CES because the two scoring methods aren't comparable. On the conventional 1-7 average scale from Dixon, Freeman, and Toman's original research, scores above 5 are generally considered strong. On the top-box (percentage-strongly-agree) method, a well-run DTC support team can sustain figures in the 80-90%+ range, but that's a reported result from one team, not an industry-wide standard.
How is customer effort score different from customer satisfaction score?
Customer effort score measures how easy it was to get a specific issue resolved. Customer satisfaction score (CSAT) measures general satisfaction with an outcome, which is easier to inflate with a discount or refund. CES is the more support-specific metric because it isolates seamlessness of the fix rather than overall happiness with the result.
How often should I survey customers for CES?
Send the CES survey immediately after each support ticket closes, while the interaction is still fresh in the customer's memory. Review the aggregate score monthly rather than weekly, because a Shopify store doing a few hundred tickets a month needs the larger sample size to distinguish a real trend from a single noisy week.
Can customer support really generate revenue, not just cost money?
Yes, when revenue from support-driven purchases, upgrades, or referrals is tracked and compared against the team's operating cost. Support becomes a documented profit center when tracked care-attributed revenue exceeds the cost of running the team. An unclear result means the tracking system needs improvement, not that support has no value.
What should a CES survey look like for a WhatsApp-based support team?
A CES survey on WhatsApp should be a single follow-up message in the same chat thread right after the agent marks the issue resolved, asking the customer to reply 1 to 5 on agreement with "it was super easy to get your issue resolved today." Keeping the survey inside the existing chat, with no separate link or login, gets meaningfully higher response rates than a follow-up email in Malaysia and Singapore, where WhatsApp is the dominant support channel.
Keep Reading
Free Download
We audited 10 Shopify stores doing $1M–$20M. Here are the 15 leaks every one had.
Real screenshots (names blurred). Most operators have 8 of the 15 and don't know it.
From 10 audits across SG, MY, AE — fashion, beauty, electronics, food.
PDF in your inbox in 30 seconds. 1 email. No follow-up unless you ask.
