The leading indicators,
and the ones that lie.

When a measure becomes a target, it stops being a good measure. Put dials on a leaderboard and dials go up — the number rises, the pipeline does not, and within a quarter you are managing a metric that has been fully decoupled from the outcome it was chosen to predict.

This is not a reason to stop measuring activity. It is a reason to be deliberate about which activity you publish. Measure effort privately to spot someone who has stopped; measure progress publicly to run the team. The two lists are different, and most CRMs ship with dashboards built entirely from the first one.

Open tracking works by embedding an invisible image and counting the fetch. Apple Mail Privacy Protection pre-fetches that image on Apple’s servers whether or not the recipient ever looks at the message, and other providers cache images similarly.

The consequence is specific: a meaningful share of your opens are machines, they are not evenly distributed across your list, and they skew toward exactly the segments most likely to be on Apple devices. Any lead score with open rate as an input is inheriting that bias silently.

Clicks and replies still work, because both require a deliberate act. The useful signal is not that a link was clicked but which link, and by how many people at the same account — three people at one company opening the pricing page in a week is a buying committee assembling, and it is the one email signal worth alerting on.

Feed clicks and replies into lead scoring as a rate against contacted accounts, never as a raw count. Raw counts reward whoever emailed the most people.

Manually logged activity is the least reliable data in the CRM. It is entered at the end of the week, from memory, by someone who knows it will be used to judge them. Every one of those three facts damages it.

Take activity from the systems that already record it. The telephony platform knows the call happened and how long it lasted. The calendar knows the meeting was accepted. The mail server knows the reply arrived. None of these require a human to remember anything, and none of them can be inflated by someone having a slow week.

That leaves exactly one thing worth asking a rep to enter: the next committed step and its date. It cannot be captured automatically, it is the strongest predictor on the list, and it is a reasonable ask precisely because it is the only one.

A high-volume inbound team was logging several hundred calls a week as a duration and a dropdown disposition. Managers sampled a handful; the rest was invisible. We transcribed the calls and classified them against a taxonomy the sales team agreed to — outcome, objection, competitor named, next step committed — written back onto the CRM record. 100% of calls classified, six objection categories tracked, same-day. Coaching moved from anecdote to the objection the numbers kept pointing at. Case study.

An inbound-led services business had enquiries landing in a shared inbox, picked up by whoever noticed. We put an SLA clock on every record with an escalation when it expired, which made per-rep response performance visible for the first time. Median first response fell from around two hours to four minutes, with no leads left unassigned overnight. Case study.

Both worked for the same reason: the measure was taken from a system that recorded it automatically, and it was attached to an action rather than a leaderboard.