Welcome to The Ops Digest!

Each week, we drop no-BS insights and one AI prompt to cut wasted costs, tighten workflows, and eliminate manual grunt work.

Today: Part 2 of 2. According to Dixon's research, about two-thirds of how customers perceive effort is emotional, not objective. In Part 1 we measured the objective side of the Customer Effort Score. Today, we're adding the emotional dimension - then combining both to identify where churn risk is most concentrated in your customer journey.

Most of Customer Effort Is Emotional. Here's How to Find It.

In Part 1, we scored Customer Effort from email threads using mostly objective signals. Channel switches. Repeat contacts. Time elapsed from first message to resolution. The model returned a number for each interaction and a journey type, and you ended up with a baseline percentage. Probably somewhere in the 60s.

That number is a starting point, not a verdict. The reason: it's missing most of what actually matters to the customer.

Matt Dixon's team at the Corporate Executive Board (now Gartner) measured what drives a customer's perception of effort. The objective stuff (number of contacts, hold times, channel switches, all the things you can count) explained about 34.6% of how customers actually felt. The other 65.4% was emotional and perceptual: whether the rep showed empathy, whether the resolution felt fair, whether the customer felt heard, whether the tone of the interaction matched the customer's state of mind.

Two-thirds emotional, one-third objective. A Customer Effort Score built mostly from countable signals, like the one in Part 1, captures the third you can count. The rest lives in the language customers use, the tone they take, and what they don't say. Layering sentiment in is the fix. Clustering by journey afterward is what makes the data actionable.

Your Team Deserves an Easier Workflow Too

Manual order entry creates more friction than most companies realize.

The Y Meadows Starter Web App lets you test a simpler way using your own orders, on your own time.

Go Ahead. Upload a messy, complex order. The blurry PDF. The spreadsheet with six tabs.

👉 Try it free

Friction Doesn't Spread Evenly

Here's the second thing the average is hiding from you. Customer friction isn't uniform across your business. It concentrates.

Imagine an overall Customer Effort Score of 68%, a fine first measurement. The journey-level view often tells a different story. Reorder workflows running at 89%. Quote follow-ups at 71%. RMAs at 41%, with most of the negative sentiment also concentrated there. The average says you have a moderate problem everywhere. The journey breakdown says you have a serious problem in one place, and a fine business almost everywhere else.

That difference matters because the action is completely different. You don't fix moderate-problem-everywhere with one project. You can fix serious-problem-in-one-place with one project. The journey-level ranking is what tells you which kind of problem you actually have.

Build It: Add Sentiment, Then Rank Your Journeys

You already have the CES Scoring Desk project from Part 1, with the prompt that scores effort and assigns a journey type. Two updates and a single Cowork instruction get you the rest.

One AI concept this week: extending a single scoring prompt to capture multiple dimensions in one pass, then aggregating the results into something that points at action.

Step 1: Add sentiment fields to your project's prompt

Open the CES Scoring Desk project in Claude.ai and edit the custom instructions. In the section that defines what the model outputs for each interaction, add two new fields after JOURNEY TYPE:

5. SENTIMENT: classify as POSITIVE, NEUTRAL, or 
   NEGATIVE.
   - POSITIVE: customer expressed appreciation, 
     satisfaction, or warmth.
   - NEUTRAL: businesslike tone throughout.
   - NEGATIVE: customer expressed frustration, 
     anger, or displeasure, even if politely 
     worded.

6. TONE NOTES: one sentence on what signals you 
   used to assess the sentiment (specific phrases, 
   tone shifts, escalations in language).

When scoring sentiment, weight what the customer 
says about the experience and how they say it, 
not just what objectively happened. A polite "I 
suppose this will have to do" is NEGATIVE. A 
neutral "Thanks, got it" is NEUTRAL.

Save the project. The model now scores effort and sentiment in the same pass.

Step 2: Re-score your existing data

If you ran Part 1 with Cowork, ask it to update the existing baseline:

Re-process every interaction in 
CES_Baseline_[date].xlsx using the updated CES 
Scoring Desk project. For each row, add two new 
columns: Sentiment and Tone Notes.

Save the updated workbook as 
"CES_Baseline_with_Sentiment_[today's date].xlsx".

Cowork will re-process in batches. A 50-interaction baseline usually takes a few minutes.

Step 3: Build the journey-level ranking

Now ask Cowork to summarize the data by journey type:

Add a new sheet to the workbook called 
"Journey Friction Ranking" with one row per 
journey type and these columns:

- Journey Type
- Total Interactions
- Average CES Score
- High-Effort Count (interactions where CES 
  is 1, 2, 3, or 4)
- Negative Sentiment Count
- Friction Hotspot Count (interactions where 
  CES is 1-4 AND Sentiment is NEGATIVE)

Sort the sheet by Friction Hotspot Count, 
highest first. Save the workbook.

Step 4: Read the result and pick a journey

Open the workbook. The Friction Hotspot Count is the literal number of interactions in that journey where the customer was both working hard AND unhappy about it. The journey at the top of the ranking is your priority. A first run for a manufacturer or distributor usually looks something like:

• Reorders: 32 interactions, avg CES 5.8, 4 high-effort, 2 negative sentiment, 1 hotspot

• New orders: 18 interactions, avg CES 5.2, 6 high-effort, 4 negative, 3 hotspots

• Quote follow-ups: 14 interactions, avg CES 4.9, 7 high-effort, 6 negative, 5 hotspots

• Shipping issues: 8 interactions, avg CES 3.8, 5 high-effort, 5 negative, 4 hotspots

• Credit disputes: 12 interactions, avg CES 3.4, 10 high-effort, 9 negative, 8 hotspots

In this example, credit disputes is the priority. 8 customers in the period had high-effort, negative-sentiment credit dispute experiences. Those are 8 customers actively shopping you for the next PO. The reorder journey, by contrast, is fine. Don't put a project on it. Put the project on credit disputes.

What to Watch Out For

• Sentiment scoring is harder than effort scoring. LLMs are reasonably good at distinguishing positive from non-positive, but they sometimes flag polite frustration as neutral. Spot-check the negative-sentiment rows the same way you spot-checked Part 1's effort scores. Read the threads. Make sure the call matches your read.

• Small samples lie. A journey with 3 interactions and 3 hotspots is not a 100% problem journey. It's a 3-interaction sample. Set a minimum of 10 interactions per journey type before you act on the ranking. If a journey doesn't have 10, pull more data.

• The 65/35 split is from Dixon's research aggregate. The actual ratio in your business may be different. Treat the principle as directional. Emotional signals matter more than they appear to at first.

The Bottom Line

An average Customer Effort Score is a starting point. It tells you there's something worth digging into. The journey-level ranking, with sentiment layered in, tells you what to do.

Pick the journey at the top of your ranking. Bring the data, including the actual email threads, to the team that owns that workflow. Have a real conversation about why the friction is there. But most of the time, the answer isn't "we need to be nicer." It's "this process makes customers contact us four times for something that should take one," or "our system asks them for information we already have," or "the rep doesn't have the authority to fix this and the customer can tell." Fix that, and the friction drops, and so does the share of customers quietly directing their next order somewhere else.

That closes the Customer Effort Score series. The pattern underneath it (score multiple dimensions in one LLM pass, aggregate into a ranking, act on the top of the ranking) generalizes well beyond CES. We'll come back to it in future issues for sales call quality, support ticket triage, and a few other places where a journey-level view changes what the team does on Monday.

Customer Friction Usually Starts Behind the Scenes

The customer only sees the symptom. Your team feels the chaos underneath it.

Y Meadows automates the repetitive order management work that creates delays, errors, and constant follow-ups.

Book a free 20-minute Y Meadows Strategy Session to see how teams are eliminating that friction at the source.

👉 See it in action