Where every agent sits
Two questions, asked in order. The top block is about buying: when did they last book? The bottom block is about looking, and applies only to agents who have not booked in 90 days. Tier and value are merged into one grade taking the higher of the two, so a Silver agent on High value reads as High. Never-booked agents get their own column and still place on the search ladder.
| Behaviour | Never Booked | Silver or Low | Gold or Medium | Platinum or High | Row total |
|---|
Switch to Detailed above for the same five states split by loyalty tier and value band, with the booking rows further divided by whether the agent is still searching — plus lapsing risk, the rescue cut and the never-booked cross-tab.
The grid
Ever-booked agents. Blue cells show agents and best-year TTV by tier; the amber column shows how much of that value band's TTV sits on agents behind their own rhythm.
| Behaviour | Best TTV | Silver | Gold | Platinum | At risklapsing vs own rhythm |
|---|
The rescue cut
Risk totalled by value band across all three tiers — the audience a high-value rescue would actually target.
Never booked
Tenure against search recency. Tenure does not determine the search band, so this is a real cross-tab — long-tenure agents searching this month are the group worth finding.
| Days since last search | New · under 6mo | Old · 6mo+ | Total |
|---|
Rows are a strict waterfall, first match wins: booked ≤30d → last booking 31–90d ago → searched ≤30d → searched 31–90d ago → neither within 90d. Booking is tested before searching on purpose, which is why an agent who booked six weeks ago and searched last week reads as Bought 31–90 days ago rather than Searching now — the purchase is the stronger fact. The five rows partition the whole base, so the grand total equals rows read.
Grade is the higher of loyalty tier and best-year value band — Silver/Low = 1, Gold/Medium = 2, Platinum/High = 3. Where the two disagree it is almost always the value band that pulls an agent up — a Silver who once booked big reads as High, which is the behaviour the merge is there to protect.
The detailed rows are the same five states, with the two booking states split by whether the agent is also still searching. Seven rows roll up exactly into the five above: Bought in last 30 days (searched / no search), Bought 31–90 days ago (still searching / gone quiet), then Searching now, Searched, went quiet and Dormant one-for-one. The two views can no longer disagree.
"Not measured" appears where fewer than 60% of a band's agents have a computable MedianGapDays and none carry the lapsing flag. It means the rhythm calculation cannot reach those agents — not that they are safe. Those rows are governed by the fixed 90-day windows instead.
Best-year TTV is the better of trailing-12 and prior-12 revenue. At risk is exposure, not forecast loss.