OnClarity Insights
Know, Fix, Prove:
a framework for customer problems that stay fixed
Three clauses, five stages and three numbers, each with an owner and an exit test
10 min read
Know, Fix, Prove is a customer service framework in which a problem counts as handled only when the company knew first, fixed it once at the source, and proved on its own data that the fix held. It runs as a five-stage loop and reports three numbers: time to know, repeat cost removed and High Value Loop rate.
Your AI agent reports a resolution rate. Your QA tool reports a score. Your survey reports CSAT. None of them tells you whether a problem stopped. This framework gives each stage of the work an owner, an exit test and a number you can compute from your own data. It works with any tools.
The framework on one page
Know, Fix, Prove: five stages, each with an owner, and one number per clause.
The three clauses
Know first. The last to know pays the most. Problems show up in customer conversations days before they reach a dashboard, the CEO or the press. Knowing first means turning the first signals into one issue, with a cause and a named owner. Its number is time to know.
Fix once. Resolved isn’t fixed. A closed ticket is one customer helped. A problem is fixed when customers stop contacting you about it. Fixing once takes two actions: handle the customer’s request now, and change the cause where it starts. Its number is repeat cost removed.
Prove it. Closed means it worked. A loop counts as closed only when the outcome registered at its opening happens, tested on the company’s own data after the test window. Whoever did the work does not grade it. Its number is High Value Loop rate.
The five stages
| Stage | Definition | Input | Output | Owner | Exit test | Number it moves |
|---|---|---|---|---|---|---|
| 1. Listen (Know first) | Every customer signal captured with a timestamp: conversations, surveys, public feedback. | Contacts in every channel, survey comments, reviews and social posts | A searchable record of signals, each with a timestamp and a customer ID where known | Head of CX insights or VoC lead | Every channel in scope feeds the record, and the first mention of any issue can be found by date | Time to know (starts the clock) |
| 2. Reason (Know first) | Signals grouped into one issue with a cause, a value and a candidate owner. | The signal record | An issue record: what customers say, the likely cause, contacts at stake, value in money, candidate owner | CX operations analyst | The issue is raised with a named owner outside service who accepts it; for a high-value issue, the loop opens with its desired outcome, test and window written down | Time to know (stops the clock) |
| 3. Act (Fix once) | The customer’s request handled now, and the cause changed at the source: the article, policy, product or process. | The issue record and the open loop | Customers answered; a dated change at the source | Frontline team lead for customers; the issue owner for the cause | The change at the source is live and dated, and customers get the right answer while it ships | Repeat cost removed |
| 4. Learn (Fix once) | The change recorded and reused, approved by a person, with who, what and when. | The shipped change | A change record: who approved it, what changed (before and after), when; knowledge, scripts and training updated | Knowledge manager or QA lead | A named person approved the change on record, and the next contact on the issue is answered from the updated source | Repeat cost removed (the fix is not redone) |
| 5. Verify (Prove it) | The outcome registered when the loop opened, tested on the company’s own data after the test window. | Desired outcome, baseline, change date, contacts after the fix | Closed on outcome, or still open with a reason; repeat cost removed calculated | A named person who did not do the work, such as a finance partner or analytics lead | The window has ended, the test ran on company data, and the result is signed | High Value Loop rate |
The three numbers
| Number | Formula | Unit | Owner | Cadence | Start with a ticket export |
|---|---|---|---|---|---|
| Time to know | From the first customer signal to the issue raised with an owner. | Hours; median and 90th percentile for high-value issues | Head of CX insights | Monthly | For your last three big issues, search contact text for the first mention before the tag existed; take the time the owner accepted it; subtract. |
| Repeat cost removed | Sum over loops closed on outcome of (repeat contacts at baseline minus repeat contacts after the fix) x cost per contact. Every channel; 8-week baseline adjusted for seasonality. | $ and % of cost to serve | CX leader; finance signs the baseline | Per loop when its window ends; quarterly total | Export 8 weeks of contacts with customer ID, channel, issue tag and timestamp. Count repeats on one issue. Multiply by finance’s cost per contact. |
| High Value Loop rate | Value of high-value loops closed on outcome / value of all high-value loops whose test window has ended. | % | CX leader; graded by someone who did not do the work | Quarterly | List last quarter’s high-value issues. Any issue with no outcome written when work began cannot count as closed. That list is your first finding. |
Five terms, one sentence each
Loop: the work opened on one customer problem, with an owner, a desired outcome, a test and a test window.
Desired outcome: the result the loop must produce, written down when the loop opens and never chosen after the fact, for example “contacts on this issue down 40% against baseline, held for 30 days”.
Test window: the fixed period after the fix in which the outcome is tested, set when the loop opens.
Baseline: the 8 weeks of contacts on the issue before the fix, in every channel, adjusted for seasonality.
High-value loop: a loop whose value (contacts at stake x cost per contact, plus revenue at risk where known, plus a weight for complaint or regulator exposure) is above a threshold set in advance; every loop above it counts.
A repeat is a contact about a problem already raised: the same customer back on the same issue, or a known issue reaching new customers after its loop opened.
How to use it in five steps
Map the stages to owners. Write a role, then a person, against each of the five stages. The stage with no name is where your loop breaks today.
Pick one high-value issue. Value last quarter’s top contact drivers (contacts x cost per contact) and choose one above your threshold whose cause sits outside service.
Register the outcome and window first. Before anything changes, write the desired outcome, the test, the window and the 8-week baseline. Finance signs the baseline.
Run the stages. Confirm the cause and the owner, handle customers and change the source, record who changed what and when, then wait out the window without moving the target.
Report the numbers. Time to know for the issue; repeat cost removed if the loop closed on outcome; High Value Loop rate once several windows have ended. Report the loops that missed as well.
Worked example
A marketplace changes its refund policy on 12 January: refunds go to store credit unless the customer asks for the card. The confirmation email still says “refund processed”.
Listen. 12 Jan, 15:20: the first chat asks why a processed refund is not on the card.
Reason. 2 Feb: the monthly report shows refund contacts rising; an analyst groups them into one issue. 4 Feb, 10:20: the payments product manager accepts it. Time to know: 547 hours. Outcome registered: repeat contacts on the issue down at least 40% against baseline, held for 30 days.
Act. Agents explain store credit and offer the card. 9 Mar: the email copy changes at the source.
Learn. 10 Mar: the change is recorded with approver, before and after; the agent macro is updated.
Verify. Baseline: 3,920 repeat contacts in 8 weeks, 70 a day; seasonally adjusted, 1,890 expected in 30 days. Actual: 340, down 82%, and at least 40% down in every week. 8 Apr: the finance partner signs. Closed on outcome.
Repeat cost removed: 1,550 fewer contacts x $5.20 (finance’s cost per contact for this channel mix) = $8,060, 0.4% of the $1.9M cost to serve in those 30 days. This loop (about $9,800 at stake) and two others, worth $29,000 of the quarter’s $41,000 in matured high-value loops, closed on outcome. High Value Loop rate: 71%.
Five common mistakes
Counting a survey follow-up as a closed loop. Calling an unhappy customer back is good service. It closes nothing until the cause changes and the outcome is tested.
Fixing the answer, not the cause. A better macro or bot reply makes the repeat cheaper, not rarer.
Choosing the outcome after the fact. If the target is written after the results, every loop passes.
Treating closed tickets as proof. A closed ticket measures your activity. The customer’s next contact measures the result.
Letting the vendor grade its own work. Whoever made the change can report it. Someone else signs the result, on your data.
Credits
Closing the loop comes from voice of the customer practice; Bain’s Net Promoter System runs it from frontline follow-up to policy change (Markey and Reichheld, 2012). John Seddon named failure demand (Vanguard) and warned against making it a target. Bill Price and David Jaffe’s The Best Service Is No Service (2008) treats each contact as a sign of dysfunction. DORA’s defined metrics became an industry standard (DORA history, 2026); Learn borrows from the blameless postmortem (Google SRE, 2017). Gartner says AI will “predict service issues before they occur” (June 2025). New here: the outcome registered first, and repeat cost in money.
How the other frameworks fit
This is the umbrella. It is not a maturity model or a map of tool categories.
Know first: The Four Clocks of Time to Know splits the delay into timed intervals so you can see which one to shorten. Surveys vs conversations vs public feedback compares the signal sources the Listen stage draws on.
Fix once: The Repeat Cost Model defines a repeat, the windows and the baseline in full.
Fix once: Fix, Automate, Keep sorts each contact driver before anyone automates it.
Prove it: The Fix Maturity Model places your team in one of four tiers: answering, resolving, fixing, proving.
FAQ
What is closed-loop feedback?
In voice of the customer practice, closing the loop means following up with the customer who gave feedback and fixing the recurring causes behind it. Know, Fix, Prove keeps both halves and adds a test: a loop is closed only when the outcome written at its opening happens within its test window, measured on your own contact data rather than reported by a tool.
What is the difference between an open loop and a closed loop?
An open loop has a signal and maybe an action, but no tested result. A closed loop has all five stages done: the issue raised with an owner, the customer handled, the cause changed at the source, the change recorded, and the desired outcome met after the window. A loop that missed its outcome stays open, and the team goes back to the cause.
Which CX metrics show that a problem was fixed?
Resolution rate, CSAT and QA scores describe single contacts, not whether a problem stopped. This framework uses three numbers: time to know, in hours, for how fast you learn; repeat cost removed, in money, for what the fix saved; and High Value Loop rate, in percent, for how many high-value loops met the outcome set before the work began.
What is a customer experience framework?
A customer experience framework is a shared model of how a company listens to customers, acts on what it hears and measures the result. Many cover strategy, journeys and culture. Know, Fix, Prove is narrower: it covers one customer problem from the first signal to a proven fix, with an owner, an exit test and a number at each stage.
Glossary
Loop. A loop is the work opened on one customer problem, with a named owner, a desired outcome, a test and a test window. It opens when the issue is raised with an owner and closes only when the desired outcome happens, tested on the company’s own data. Loops above a value threshold are high-value loops.
Desired outcome. The desired outcome is the result a loop must produce, written down when the loop opens, before any fix ships. It names the measure, the target and the window, for example contacts on the issue down 40% against baseline for 30 days. It is never chosen after the results and never graded by the vendor.
Time to know. Time to know is the number of hours from the first customer signal to the issue raised with an owner. The first signal is the earliest contact, survey comment or public post about the issue, found by date. Report the median and 90th percentile for high-value issues each month.
Repeat cost. Repeat cost is the cost of contacts about problems already raised, from the same customer or from new customers after the issue’s loop opened. Repeat cost removed = sum over loops closed on outcome of (repeat contacts at baseline minus repeat contacts after the fix) x cost per contact; every channel; 8-week baseline adjusted for seasonality.
High Value Loop rate. High Value Loop rate is the value of high-value loops closed on outcome divided by the value of all high-value loops whose test window has ended, as a percentage. Closed means it worked: a loop counts as closed only when the outcome registered at its opening happens, tested on the company’s own data.
