Activities Per Lead: Effort Against Outcome

Why this matters

Activities per lead is the cleanest example in a shop of a number that measures effort and tells you nothing about whether the effort worked. On its own it can go up because the office is chasing harder, or because it is chasing the wrong leads harder, and both readings look identical. It only becomes a diagnosis when it is cut by what happened to the lead, and before you can cut it you have to fix two things about how it was assembled, because as normally reported the top and the bottom of the fraction are not describing the same leads.

What the average is made of

• The numerator is logged activities in the window: calls, texts, emails, site visits, whatever your records treat as a touch.
• The denominator is leads created in the window.
• Those are two different sets. An activity logged in March against a lead that arrived in January is in the numerator, and that January lead is not in the denominator. Work carried over from older leads inflates the figure; leads created in the last days of the window deflate it.
• Anything nobody logged does not exist. A hallway conversation, a text from a personal phone, a call returned from a truck. The average is a measurement of logging discipline before it is a measurement of effort.

One more decision belongs here and it changes the number by a lot: does an unanswered attempt count as an activity? Either answer is defensible, but pick one and write it down, and if a ring-out counts, report attempts and contacts as two figures. Otherwise a shop can raise its activity average substantially without speaking to a single customer.

Two corrections before the average means anything

A quarter, 13 weeks. Logged activities in the window: 784. Leads created in the window: 180. The figure as published is 784 divided by 180, 4.4 activities per lead.

Correction one, put the numerator and denominator on the same leads. Of the 784 activities, 172 were logged against leads created before the quarter started. That leaves 784 less 172, so 612 activities against the 180 leads created in the window, 3.4 each. A full activity per lead sat in the gap, and it was not effort anybody applied to this quarter's leads.

Correction two, take the calendar out of it. Of the 180 leads, 41 were created in the last 3 weeks of the 13 and carry 44 activities between them, about 1.1 each. Nobody neglected them. They are days old. Re-cut on the 139 leads created in the first 10 weeks: 612 less 44 is 568 activities over 139 leads, 4.1 each.

So the same quarter is 4.4, 3.4 or 4.1 depending on which set you take, and the three are not interchangeable. Use the cohort figure of 3.4 for anything about this quarter's demand, and the mature-cohort figure of 4.1 for anything you intend to compare against a prior period. If you compare 4.1 against last quarter, last quarter gets both corrections too. Comparing a corrected figure against an uncorrected one manufactures a trend out of arithmetic.

Crossing it with outcome is where it becomes a diagnosis

Split the 612 cohort activities by what happened to the lead:

Won leads carry 6.2 activities each against 2.75 on lost ones, 2.25 times as many, and both of those are per-lead averages inside the same cohort. The blended 3.4 sits between them and describes neither. That is the whole case against reporting the single number: it is an average of two populations that behave nothing alike, and it will move whenever the mix between them moves, with no change in anyone's behaviour.

The still-open row is a reminder rather than a finding. Those 37 leads have no outcome yet, so the booked rate here is 47 of the 143 resolved leads, 32.9 percent, or 47 of all 180 created, 26.1 percent. The gap between those two is cohort truncation and a sibling card owns it.

The distribution inside the losses

The lost average of 2.75 is itself a blend, and this is the cut that produces something to do on Monday:

Thirty-eight of the 96 lost leads, 39.6 percent, were closed after exactly one touch. That is a number you can act on directly and it is entirely hidden inside the 2.75, which reads as a shop making a reasonable if unspectacular effort.

At the other end, 12 lost leads absorbed 8 touches each, 96 activities, which is a third of what the 47 won leads consumed in total. Whether that is a problem depends on a threshold you set, and the rest of this article uses one.

Three patterns, and the tell that separates them

These are patterns in a distribution, not three types of shop. A real shop usually shows more than one at a time, which is why each needs its own tell.

Run the shop above through all three.

Abandoned early: yes. 39.6 percent of lost leads at 1 touch or fewer, against a gate of about 25 percent. This is the finding, and the fix is a standing rule rather than a campaign: no lead is closed as lost before three attempts across at least two channels and two days, unless the customer said no or it is out of scope.

Chasing the unreal: no. 12 of 96 lost leads at 6 or more touches is 12.5 percent, under the 15 percent gate. There is a tail and it is not yet worth a rule. Watch it, because a stop rule introduced to fix a tail this size costs more in argument than it saves in hours.

Pre-sold: yes, on one source. 14 of the 47 wins came from a single referral source and closed on 32 activities, 2.3 each, against 7.8 each across the other 33 wins. That source produced 33 leads in the quarter of which 14 booked, 42.4 percent, against those 33 wins from the other 147 leads created, 22.4 percent, the shop's other sources with this one taken out so the two do not overlap. Thirty-three leads clears the 30-lead floor the source-ranking card sets, so this is a finding rather than a candidate, and the response is budget and attention, not a sales process change.

What the number cannot tell you about cause

Nothing above establishes that touches cause bookings, and the arrow mostly runs the other way. A customer who answers the phone, asks for a revised scope and requests a second visit generates activity because they are engaged. A customer who has already hired somebody else generates one unanswered call. So the 6.2 on won leads is substantially an effect of winning rather than a recipe for it, and a shop that reads it as a recipe hands its office a quota of touches and gets the quota met against leads that were never going to move.

That is why the actionable cut is the floor of the lost distribution rather than the won average. A lead closed at one touch never got the chance to engage either way, so it carries no information about itself, which is the one case where more effort is unambiguously worth trying. And it is why the pre-sold row above ends in a budget decision rather than a process one.

Two guards keep this honest. Watch attempts and contacts separately, so a rule about attempts cannot be met by dialling. And read the lost reasons behind the single-touch pile before writing the rule, because if most of them say the customer said no on the first call, the pile is correct behaviour and the rule would be waste.

Two ways to move the same average in opposite directions

Say the shop wants the cohort figure of 3.4 to move. There are two honest routes and the number cannot tell them apart.

Add the touches. Bring the 38 single-touch losses up to 3 touches each. That is 76 more activities, so 612 plus 76 is 688 over the same 180 leads: 3.8. Cost is office hours. Upside is whatever share of those 38 leads was reachable, which nobody knows yet because nobody called them.

Remove the demand. Twenty-four of those 38 single-touch losses came from one source that produced 31 leads, 36 activities and 1 booking in the quarter. Drop it: 612 less 36 is 576, over 180 less 31, so 149 leads: 3.9. Cost is 1 booking a quarter. Upside is the hours back, and 31 leads clears the same 30-lead floor, so that row can be ranked, and 1 of 31 is 3.2 percent against 46 of the other 149, 30.9 percent. That is not close.

Both land the average at about the same place from 3.4, and they are opposite acts: one adds work to the demand you have, the other removes the demand. A shop reporting only the average sees an identical improvement either way.

There is one difference the average does hide and it is worth having. Cutting that source also takes its 1 booking out of 47 and its 31 leads out of 180, so the blended booking rate moves from 26.1 percent to the 30.9 percent the rest of the book was already running at, without anybody booking an extra job. That is not performance. It is the mix, and it is the reason to check what a source did to your rate before congratulating anybody for the rate moving.

References

• See related: Lead Source Performance: Rank by Conversion, Not Volume, which sets the 30-lead floor used above
• See related: Estimate Conversion Rate and the Cohort Problem, which owns the truncation in the still-open row
• See related: Lost Reasons Are Only as Good as the Field Being Filled, for reading the single-touch pile before you rule on it
• See related: Overdue Follow-Ups Cannot Count a Lead With No Date Set, for the companion measure of leads nobody has touched at all