Starting a Programme From Nothing
The first ninety days, in order. Almost none of it requires a purchase, and the sequence determines whether the programme is trusted or resisted.
An operation with no formal quality programme can build something useful in a quarter without buying anything.
Weeks one to two: find out what you have
Check the audio. Stereo or mono, per queue. This determines what is possible and it is frequently unknown.
Check the recording notice against what you actually intend to do. Analytics may exceed the stated purpose.
Check retention. What is kept, where, for how long, including transcripts and any existing analytics.
Establish which jurisdictions your agents and customers are in, and whether employee consultation is required before monitoring changes.
None of this costs anything and all of it constrains what follows.
Weeks three to four: define quality
Write down the outcome you are trying to produce. Resolution, retention, compliance, sales.
Take fifty calls that produced that outcome and fifty that did not. Listen to them blind and note what differs.
Draft a scorecard from what you found, not from a template. Ten items maximum, split into compliance and behavioural sections.
Weeks five to eight: start evaluating
Two evaluations per agent per week, delivered within days.
Fifteen-minute coaching conversations, one or two points, with a recorded action.
A calibration session at the end of week five and week eight, with agreement measured.
Expect the first calibration to be a shock. That is the finding.
Record everything that is not an agent problem in a separate list.
Weeks nine to twelve: close the loop
Verify. Did the actions from week five produce a change.
Route the process findings to whoever owns systems, knowledge and policy, quantified.
Run the outcome correlation on the calls you have scored, crudely. Which items separate the good and bad outcomes.
Revise the scorecard based on that.
Report: compliance failures, the outcome trend, the process findings raised and fixed, and the calibration agreement.
After the quarter
Then consider analytics, with a scorecard that has been tested, a calibrated evaluator pool, a defined workflow and a known audio position.
A platform bought before this arrives into an operation with no definition of quality, no calibration and no route for findings, and it produces dashboards.
The three things that matter most
Fix the audio. Stereo capture is the highest-return change available and it precedes every software decision.
Tell agents what you are doing, before you do it, including the limits.
Route the process findings first. The programme's reputation is set by whether its first outputs helped agents or blamed them.
What to buy first, when you do
Not a full platform. Transcription and search, so you can find things.
Then categorisation, built by you on your own operation.
Then compliance automation, once accuracy is measured.
Automated behavioural scoring last, if at all, and only with a dispute route and a measured accuracy gap by speaker group.
The measure of whether it worked
After a year, can someone name three process problems that were found and fixed, and show the effect?
If yes, the programme is producing value beyond monitoring. If the only output is a monthly score, it is a number, and it will be cut in the first budget round that needs one.
The three-month checklist
Condensed, for an operation starting from nothing.
Month one: confirm stereo or mono per queue; review the recording notice against intended use; check retention and where recordings live; establish jurisdictional obligations; listen to fifty calls with a good outcome and fifty without, and note what differs.
Month two: draft a ten-item scorecard split into compliance and behavioural sections; evaluate two calls per agent per week; deliver feedback within days; run a calibration session and measure agreement; keep a separate list of everything that is not an agent problem.
Month three: run a second calibration and compare; verify whether coached actions produced changes; route the top three process findings with frequency and cost; run a crude outcome correlation on the scored calls; revise the scorecard.
End of quarter: you have a tested scorecard, a measured agreement figure, a working delivery loop, a route for process findings, and a known audio position. That is the foundation a platform needs and rarely finds.