Havish M Consulting
Customer Case Study · Customer Experience
Retail × Havish M Consulting

SLA Guidelines & Customer NPS

GenAI mapped every customer comment to a service SLA, and showed that being on time was not the same as being good enough.

ClientRetail FunctionCustomer Experience ThemeSLA & NPS
Delivery SLA met
0/ 15 comments
▲ on time within promised window
Perception gaps
0/ 15 comments
▼ met SLA, still felt slow
SLA dimensions breached
0of 6
▼ communication, support, quality, app
Comments mapped to an SLA
0%
▲ every verbatim, with evidence

Based on the 15-comment analysis shown below. NPS figures are indicative; validated metrics available on request.

The Challenge

"Delivered on time" but the customer still wasn't happy

Customer feedback arrived as free text. The SLA guidelines lived in a separate document. Nobody could say which promises were being kept and which were hurting NPS.

Feedback as free text

Verbatims mixed delivery, packaging, app and pricing in one paragraph. Manual tagging was slow and inconsistent.

SLA met, NPS missed

Deliveries landed inside 2–5 days, yet customers compared the retailer with faster platforms and felt let down.

Issues across six teams

Logistics, app, catalogue, support, returns and packaging each owned part of the experience, and each saw only its own slice.

Breaches hidden in averages

Zero-tolerance SLAs, such as wrong items or tampered packaging, were breached without anyone tracking them against the guideline.

HMC Approach

Make every comment answer to an SLA

Two inputs, customer comments and SLA guidelines, turned into one decision-ready report.

Select a step to see what happened.

Discover · Gather the two sources of truth

The customer's voice and the company's promise, side by side.

  • Collected customer feedback comments from post-delivery surveys.
  • Consolidated SLA guidelines into six dimensions: delivery, order communication, customer support, returns and refunds, product quality, and experience.
  • Agreed what counts as "met", "perception gap", "partial breach" and "breach".

Deliverable: Comment set and a single SLA guideline sheet.

Design · A mapping frame for GenAI

Each comment gets the same structured treatment.

  • For every comment: SLA dimension, status, customer observation and SLA implication.
  • Per SLA dimension: guideline, status, evidence and the comment numbers behind it.
  • Recommendations split into immediate fixes, short-term gaps and strategic moves.

Deliverable: Prompt frame and output template.

Build · From verbatims to a four-tab report

Claude read every comment against the guidelines and produced the report.

  • Executive summary with six headline counts and a key takeaway.
  • Comment-by-comment SLA mapping, with evidence linked to the guideline.
  • SLA scorecard across six dimensions, plus prioritised recommendations.

Deliverable: Feedback vs SLA HTML report: Summary, Comment Analysis, SLA Scorecard, Recommendations.

Adopt · One review, six owners

The report became the agenda for a cross-functional CX review.

  • Each breached SLA was assigned to an owning team with a fix date.
  • Perception gaps went to marketing and product for messaging and options.
  • "Comment 9 moments" were defined as the experience standard to replicate.

Deliverable: Action log with owners and dates.

Scale · Run it on every month's verbatims

The same frame now runs on the full NPS comment stream.

  • Monthly run on all survey verbatims, not a sample.
  • Perception-gap share tracked alongside NPS and SLA attainment.
  • SLA guidelines updated when customer expectations move.

Deliverable: Monthly Feedback vs SLA pack.

The Solution

Every comment, every SLA, one view

Each row is a customer comment; each column an SLA dimension. Hover or click a cell to see the observation and the guideline it tests. Toggle the lens to see how much of the problem is a breach and how much is perception.

15 comments × 6 SLA dimensions
MetMet, perception gapPartial breachBreachNot raised

Screens from the actual analysis: two inputs and the four-tab GenAI report. Click an image to enlarge.

Impact Dashboard

From SLA findings to an NPS plan

Each recommendation is linked to the comments it would fix. Switch fixes on to see how sentiment, and an indicative NPS, would move.

NPS simulator

The 15 comments as promoters, passives and detractors

Indicative NPS
–

SLA scorecard

Comments per SLA dimension, by status

MetPerception gapPartialBreach

NPS classification of the 15 comments is indicative, inferred from comment sentiment. Validated metrics available on request.

Before vs After

How CX reads customer feedback

DimensionBeforeAfter
Reading feedbackSkimmed and manually taggedEvery comment mapped to an SLA dimension and status
Link to SLAsGuidelines in a separate documentEach finding cites the guideline it tests
"On time" deliveriesCounted as successSplit into met vs met-with-perception-gap
Zero-tolerance SLAsBreaches lost in averagesDamaged, open or wrong items flagged by comment number
OwnershipIssues bounced between teamsEach breached dimension has an owner and a fix
NPS storyA score without a causeScore linked to the SLA gaps behind it
Tools & Tech Used

Two inputs, one prompt frame

Claude comment-to-SLA mapping & report SLA guideline sheet six dimensions Survey verbatims post-delivery feedback HTML report four tabs, shareable Excel action log & owners
Key Takeaways

What transfers to any customer-facing business

01

An SLA met is not a customer won

13 of 15 deliveries were on time, yet 8 customers still felt short-changed. Competitors set the expectation, not the SLA document.

02

Fix upstream, not after delivery

Zero-tolerance SLAs need prevention: scan-verified dispatch and packaging checks at handoff, not better apologies.

03

Engineer the best moments

One fully satisfied customer showed the standard: fast, secure, as described. Design for more of those on purpose.