Datasheet · LUFT Host Analysis report

How to read Host analysis

Every number on this screen is counted live from the booking system's own records. This page shows exactly what gets counted, how, and how we know it's right.

The one rule

Everything is a share of Scheduled

Scheduled means every call booked into the placement calendar for that month — whether it later completed, was cancelled, was a no-show, or hasn't happened yet. Every percentage on the screen divides by that same number. Learn that once and every table reads the same way.

118Scheduled

Scheduled

Calls booked for the month. Mkaya had 118 in June — that's the bottom of every one of her June percentages.

63Completed

Completed

Calls that actually happened. 63 of Mkaya's 118 did — 53%.

49Customers

Customers

Completed calls where the person went on to become a paying customer. 49 for Mkaya — 42% of everything she had booked.

Where numbers come from

The journey of one cell

Take Mkaya's June "Scheduled → Completed: 53%". Here is its whole life:

1 · Booked
A call is booked
The booking system writes a record: who, which host, what time
2 · Happens
Outcome recorded
Completed, cancelled or no-show — stamped on the same record
3 · Copied
Warehouse copy
The DataHub copies the record within five minutes, and keeps it forever
4 · Counted
63 ÷ 118 = 53%
The screen counts records and divides. That's all it does.
No spreadsheet anywhere in that chain. The screen counts the booking records themselves, and re-counts them live every time you open the page — so it can never drift out of date.
The six rates

Three about winning, three about losing

The funnel tables show how booked calls turn into customers. The drop-off tables show where the rest went. Same denominator throughout.

53%Sch → Completed

Scheduled → Completed

Of everything booked, how much actually happened.

78%Comp → Customer

Completed → Customer

Of the calls that happened, how many were converted. The purest measure of the host's selling.

42%Sch → Customer

Scheduled → Customer

The two above multiplied — bookings all the way through to money.

27%No-shows

No-shows

Booked calls where the person never turned up.

7%Host cancels — after

Cancelled by host AFTER start

The host cancelled at or after the moment the call should have started. The worst kind.

0%Host cancels — before

Cancelled by host BEFORE start

The host cancelled ahead of time. Better, still lost.

Host cancellations only. When the client cancels (666 times so far, vs 474 host cancellations), that's not a host-performance signal — those live on the LUFT cancellations report instead. And "before vs after" is decided by the moment the booking system logged the cancellation, not by when the record was last edited — an edit can happen days later and lie about the timing.
Grand totals

Totals pool the calls — they never average the percentages

A team rate is all the calls added together, then divided once. Averaging the hosts' percentages instead would let a tiny month shout as loudly as a big one.

1 / 1
Host A: one call booked, it completed — 100%
+
10 / 100
Host B: a hundred booked, ten completed — 10%
=
11%
the team's real rate (11 of 101) — not the 55% an average would claim

The same rule runs sideways: a host's Grand Total column pools all their months. That's why a Grand Total is never just the middle of the numbers beside it.

Cost efficiency

Two costs, shared out per call

Winning a booked call takes ad money; running it takes a host's time. The cost tables put both against each host's calls, then divide by calls and by customers won.

Advertising: there's one ad bill a month, and a toggle for how to split it. The screen offers two views, because they answer two different questions:

"All booked calls" (the default) splits the bill equally across every call booked that month — including calls that came in free through Google search. June:

£10,879
Google Ads spend in June
÷
722
every call booked in June
=
£15.07
per booked call — same for every host
×
118
Mkaya's June calls
=
£1,778
Mkaya's June advertising cost

Why include the free calls? Because hosts don't pick their calls — bookings just land on the rota. Charging every call the same share means a host's ad cost is purely about how many calls they took, never about which calls luck happened to hand them. It's the fair way to compare hosts.

"Ad-traced calls" splits the same bill across only the calls we can trace back to an ad. We check each booking against our click records: if that person arrived through a paid ad before booking, their call is ad-traced. June had 637 of those, so the same bill works out to £17.08 per traced call, and each host is charged for their own traced calls. This view shows the ad money sitting on the calls it actually bought — use it when the question is about the ads rather than the hosts.

Either way, the whole bill always gets allocated — the toggle only changes how it's spread.

Host wage is what the host's time cost. Right now it shows what each host is owed for their rota shifts, because host invoices are only just starting to come in. Every wage figure carries a dotted underline — hover it and it tells you which it is. As each invoice is agreed on the invoices screen, the real figure takes over automatically. For the two team leaders, the wage is their call shifts only — their support and training time funds the whole team, so it's never charged against their own calls.

CPA — the headline number — is simply both costs added up, divided by customers won: what one new customer cost through that host.

Why trust it

How we know the numbers are right

"Right" here means: recount the booking records yourself and you get the same answer. Four checks stand behind the screen.

Counted from the system of record

Every cell is a live count over the booking system's own rows — 4,460 booked calls to date: 2,301 completed, 1,140 cancelled, 974 no-shows. There is no manual step anywhere in the chain.

The warehouse copy is audited

The DataHub's copy of the booking tables is compared against the live site table-by-table, row-count for row-count. When the live site falls behind (it happens — a sync race was caught and healed this week), the audit finds it.

Cancellation times can't be dragged

We don't guess when a call was cancelled from when its booking record was last edited — a later edit drags that time forward, which can make a host look like they cancelled after the call when they cancelled well before it. We use the moment the system itself wrote "cancelled" in its log, which never changes. That's what decides the before/after split.

Every cell is traceable

Hover any cell and it tells you how many booked calls sit underneath it. Ask the warehouse assistant on the page and it will list the actual calls. There is no number on the screen that can't be walked back to individual bookings.

Limits

Three things to hold while reading

The current month always reads low

A call booked for next week is already in Scheduled but can't have completed yet — so this month's rates sag until the month closes. Judge hosts on finished months.

Wages are shift-owed money until invoices arrive

The dotted figures are what the rota says the host is owed, not what they billed. They swap to agreed invoices automatically — the cell itself tells you which one you're reading.

Pick the ad view that matches your question

"All booked calls" is for judging hosts — everyone carries the same per-call share, so only volume separates them. "Ad-traced calls" is for judging the ad money — it concentrates the bill on the calls the ads actually bought. Comparing one host's number from one view against another host's from the other view means nothing.