4 lessons · about 9 min · written by the PTS AI team from live deployments
Prove it, then repeat it
You cannot measure it later
2 min
The baseline has to be captured before you switch anything on. Afterwards it does not exist and cannot be reconstructed from memory or from a report built for a different purpose.
Capture before launch
Response time to a new lead, from records and not from policy.
Contact rate, manual hours, appointments set and kept, conversion at each stage.
Revenue for the same period last year, for seasonality.
Pick the one number the project will be judged on, say it out loud, and get the person who controls the budget to agree it is the right one. Before, not after.
This takes an afternoon and it is the difference between a result and an argument.
Twenty thousand calls is not a result
2 min
Activity metrics feel like progress and prove nothing. There is a ladder, and only the top of it survives a conversation with someone who controls money.
Calls made is an input. It measures how hard the machine worked.
Conversations means someone actually talked.
Qualified opportunities means something was worth a human's time.
Attributable revenue is the number.
Where the win is capacity rather than revenue, the equivalent top of the ladder is hours returned to selling, valued at what those hours produce. Same discipline, different unit.
If someone asks you for calls made, they are asking the wrong question, and it is worth saying so.
What else changed?
2 min
Your before-and-after table looks great. Now find everything else that changed in the same period, because a skeptical CFO will, and it is better if you found it first.
Seasonality. Compare to the same period last year, not to last month.
A simultaneous marketing change. Somebody always launches something.
A new rep, or one who left. Team composition moves numbers more than most software.
A price or product change.
A holdout — two locations that do not get the automation for a quarter — costs a little upside and buys you a result nobody can argue with. In a multi-location business it is nearly free.
Name the confounds yourself, in the report, before anyone asks. It reads as confidence, and it is the version that survives the second meeting.
Prove one. Then ten.
3 min
The temptation after a working first campaign is to launch three more. It is also the most reliable way to end up with four campaigns nobody can explain.
Proven means all four
Stable inputs: the list refreshes without someone remembering.
Stable outputs: consistent for four weeks, not one good week.
Known cost per outcome, all in.
A named owner who is not you. This is the one people skip.
Then improve before you expand
The list, the timing, the script opening, the bar. One change at a time. Most operations leave twenty to forty percent on the table in the first tuning pass, which beats any second play you could launch instead.
Ten locations without a weekly review is ten independent systems that share a logo. Fifteen minutes, same time, comparing every location against the median.
That is the whole method: find the leak, pick the play, wire it, keep the humans where they matter, prove it moved money, then do it again somewhere else.
Rather not build it yourself?
We already run this across 28 finance branches.
Same system, installed and operated for you — calling, qualifying and handing leads to your team while you focus on closing.
