Performance
Your numbers, and the calls behind them.
Ask Maddox how you are doing and it reads your scorecard, your metric history and your call record. Managers can ask the same questions across the team they are responsible for, within the scope their permissions already define.
How it works
Numbers, then behavior, then evidence
A metric on its own rarely tells you what to change. These three layers do.
The number
Your scorecard, your pacing to target, and the history of any tracked metric over time: connect rate, decision-maker reach, appointments set, calls made, win rate, revenue and attainment.
The behavior
What you actually did on calls. Whether you got past the gatekeeper, whether you asked for the meeting, whether you asked again after a no, your share of the talking, and the playbook items graded done, partial or missed.
The evidence
The calls themselves, newest first, with who was called, the outcome and the assessment where you are permitted to see it.
In the product
A number you can trace to a call.
Worked example
Who is doing better, and why
A normal CRM hands a manager a leaderboard and leaves the “why” to a guess. Here is what happens when a sales manager asks MADDOX the question directly.
The question
A manager types: “Who on the team is doing better this month than last, and why?” Nothing about the wording is special. Maddox resolves “the team” to the reps that manager is already authorised to see, and “this month” to a window it can quote back.
First, the numbers over the same window
It reads every rep’s scorecard, and for anyone who stands out it reads that rep’s panorama, which returns this month and the month before side by side: activity, scorecard, coaching record and commitments, each as its own ledger with its own date. Metric history answers better-or-worse directly — connect rate, decision-maker reach, appointments set, win rate, attainment — with the sample size each figure rests on.
The answer says which reps moved and by how much, quoting the figures as the scorecard holds them. Where attainment rests on too small a sample, or a rep is still ramping, that flag is repeated beside the number rather than left in a footnote.
Then, the behaviour that differs
A number moving is the what. The why is usually on the phone, so Maddox reads the call behaviours of the reps in question: how often they got past the gatekeeper, asked for the meeting, asked again after a no, and how much of the call they did the talking. Every rate comes with its denominator — a re-ask rate is counted over calls where somebody actually refused, not over every dial — and both windows are read by the same rules, so this month is compared with last month rather than with everything since.
For account executives the same question reaches different records: win rate, the playbook steps each rep most often misses, and how often each one took a close the customer offered on a recorded meeting. A rep with fewer than three of those moments gets no rate at all, because two chances is not a track record.
Then, the calls as evidence
A rate on its own invites an argument, so the answer cites the calls behind it: the actual exchanges where a rep asked again after a no, with who said what, the timestamp and a link that plays the moment. Ask for more and it opens one call down to the words, with its QA criteria and any objection graded on it. An assessment you are not cleared to see is marked withheld, never quietly dropped.
What it will not tell you
The honest part of an answer is what it is missing, and the answer carries that too.
And it will not hand down a verdict. Maddox reports what the scorecard computed and what the calls recorded; it does not decide that one rep is better than another as a person, or offer an assessment of anyone. Where a manager has stored an evaluation, it is quoted as written.
- Calls that could not be measured are counted and grouped by reason, not silently left out of the rate
- Calls scored by an older version of the classifier are reported as awaiting a recompute rather than mixed in
- A metric with nothing to measure comes back with the reason it is empty, never as a zero that would rank someone below a rep genuinely struggling
- Figures withheld from the person asking are named as withheld
- The CRM call log and the softphone’s weekly rollup are reported separately and never combined
Where people take it next
The follow-up questions are where an answer like this earns its keep, because each one is a lookup rather than a fresh guess.
- “Show me the calls where [rep] asked again after a no.”
- “Play me the moment [rep] got past the gatekeeper at [company].”
- “What is on [rep]’s development plan, and has it been confirmed?”
- “What did [rep] commit to at our last 1 on 1, and is it done?”
- “Which reps have a recorded finding on discovery?”
The mechanism
Two ledgers that do not reconcile, and a field that says it is withheld
Performance is the area where the difference between a missing number and a hidden one changes what someone does next, so most of the engineering here is about not blurring it.
The assistant calls the scorecard page’s own producers
Asked how you are doing, Maddox does not compute a scorecard. It calls the same builders the scorecard screen calls, with the same defaults an unfiltered visit uses, so the answer and the page agree by construction. Recomputing any of it would produce a second set of numbers that disagreed with the screen you are looking at, which is worse than not answering at all.
The CRM ledger and the softphone ledger are never blended
They count different things over different windows and they do not reconcile — that is stated in the code, not discovered by accident. The CRM window runs back from the date of the answer by a configured lookback, floored to the start of a day; it is not a calendar week. The softphone band is the softphone’s own weekly rollup. Both come back under their own heading with their own window stated, and the assistant is told not to average across them. Two ledgers reported separately is an honest answer. One number made from both is a fabrication with a decimal point on it.
Withheld, unmeasurable and zero are three different facts
Where a judgement metric is not yours to see, it is removed rather than nulled, and the list of what was removed travels with the row. That distinction is load-bearing: on this dashboard a null already means “the softphone did not report it”, which is a fact about the source rather than about the viewer. Collapsing the two would make an unscored call and a call you are not cleared to see look identical.
So a manager reading a team scorecard is told which figures are withheld from them by name, and never shown a zero standing in for one. Where there is genuinely too little evidence, the value comes back null with a reason attached rather than as a number, because a fabricated zero would rank someone with no data below someone genuinely performing badly.
Fairness stamps travel with the figure
Attainment against target carries flags the assistant must repeat beside the number: one saying the figure rests on too small a sample to lean on, another saying the rep is still ramping. A percentage quoted without them invites a conversation the data does not support. Forecast rows are read rather than recomputed, for the same reason — re-deriving a forecast at the moment somebody asks would quietly replace what a rep actually committed to with what the pipeline would say today, and comparing those two is the entire point of a forecast review.
Who it is for
For reps first, managers second
The self-scoped questions need no management permission at all, so a rep can interrogate their own numbers without asking anyone.
- Reps checking their own pacing without waiting for a 1 on 1
- Managers asking where the team is slipping, within their own scope
- Anyone who has been handed a number and wants to know what is behind it
Related
Questions
The things people actually ask.
Can my manager see everything I can see?
Their scope is defined by the team they are responsible for, and by the same permissions the rest of the product uses. Where a specific metric is withheld from them it is marked as withheld, not omitted, so they can tell the difference between a number they may not see and a number that does not exist.
What happens when there is not enough data to compute a metric?
It comes back as not measurable, not as zero. A fabricated zero would rank someone with no data below someone genuinely performing poorly, which is the opposite of useful.
Can I ask what someone said on a call?
Yes, where you are permitted to see that call and its transcript. Assessments you are not permitted to see are reported as withheld rather than quietly dropped.
Can a manager see everything about one rep in one answer?
Yes, within their scope. The rep panorama reads activity, scorecard, coaching record, commitments and meeting grades for one person in a single pass, with every withheld figure named as withheld. It is a reading of what was recorded, not an assessment of the person.
Does Maddox judge my performance?
No. It reports what was already recorded by a person or computed by the scorecard, and it is explicitly instructed not to assess anyone itself or infer anything from their numbers.
Ask why a number moved.
The interesting answer is rarely the metric. It is the behavior underneath it, and that is on the record too.