Speed to Lead
How to Measure Your Own Speed to Lead
The number on your dashboard is probably flattering, and the reason is that the clock starts in the wrong place.
Spent 8 years running operations for a startup studio that launched more than a dozen companies. Leads CRM, automation, AI assistants, websites, and delivery infrastructure at Agency Logics.
Published
The clock usually starts in the wrong place
Before trusting any response-time figure, establish what its start timestamp actually records. The difference is not cosmetic. It is frequently the entire quantity you are trying to measure.
| Clock starts at | What it measures | What it hides |
|---|---|---|
| Lead arrival at your system | The real wait, from the customer's point of view | Nothing. This is the one you want |
| Record creation by an integration | Close to real, if the sync is immediate | Any sync delay, which can be hours on batch integrations |
| First time a user opened the record | How fast someone acted once they looked | The entire period the lead sat unopened. Usually most of it |
| Assignment to a salesperson | Speed after routing | Routing delay, which is often the bottleneck itself |
The third row is the common default and the reason dashboards report flattering numbers. A lead that sat untouched from 9pm to 9am and was then called at 9.04am is recorded as a four-minute response.
Verify it with a test lead
Submit an enquiry through your own public form, note the exact time you pressed submit, then look at what the CRM recorded. If those two times differ, you have found a measurement problem before you have found a speed problem, and the measurement problem has to be fixed first.
The method
Export 90 days of leads with two timestamps
You need the moment the lead arrived and the moment of the first genuine outbound contact attempt. Ninety days is enough to cover weekly and seasonal rhythm without the export becoming unwieldy.
Check what your arrival timestamp actually records
This is where most measurements go wrong. Some systems stamp the record when it is created by an integration, others when it is synced, others when a user first opens it. If the arrival stamp is really an open stamp, every number downstream is fiction. Verify it by submitting a test lead and comparing.
Define first contact as an attempt, not a connection
An outbound call that rang out still counts as a first attempt. A logged note does not. An automated acknowledgement is worth measuring but should be tracked as its own separate metric, because it tells you something different from human response.
Compute the median, not the mean
Response times have a long tail: a handful of leads answered nine days later will drag a mean into meaninglessness. The median tells you what a typical enquiry actually experienced. Report the 90th percentile alongside it to show the tail.
Split by arrival window
Business hours, evenings, and weekends behave completely differently, and a single blended median hides the overnight gap entirely. This split is usually the moment the real problem becomes visible.
Count attempts per lead as well
For the same 90 days, count total contact attempts per lead and group them: one, two, three, four or more. This distribution is a separate diagnosis from response time and is frequently the more expensive of the two.
Re-measure quarterly, the same way
Speed decays as volume grows and staff change. A number measured once is a fact about last quarter, not a control. Keep the method identical so the comparison means something.
What to report, and to whom
Four figures, on one line, refreshed quarterly. Anything more elaborate stops being read.
- Median time to first human attempt, all leads, after-hours included. This is the headline.
- 90th percentile, which shows how bad the tail is. A good median with a terrible 90th percentile means a subset of leads is being systematically dropped.
- The business-hours and after-hours medians separately, so the overnight gap is visible rather than averaged away.
- The share of leads with fewer than four contact attempts. The single most useful number in the set, and the one nobody currently reports.
Do not exclude the inconvenient leads
The temptation, once someone is accountable for the number, is to exclude after-hours enquiries, spam, or leads from a channel deemed low quality. Each exclusion improves the metric and degrades its usefulness. If spam is genuinely distorting the figure, fix the form, do not filter the report.
What the answers mean
- Bad median, bad after-hours split. A coverage problem. Start here, because it needs no behaviour change.
- Good median, bad 90th percentile. A routing problem. Some subset of leads is falling into a gap, usually a specific source, form or assignment rule.
- Good median, most leads under four attempts. The most common result we see, and the most expensive. The follow-up cliff.
- Everything looks fine. Then the leak is elsewhere, and this exercise has saved you from buying a solution to a problem you do not have. That is a real result.
Whatever the diagnosis, the leads already abandoned before you started measuring are a separate and immediately addressable pool. Sizing that is what your database is worth.
Common questions
- Why is my CRM dashboard number so much better than reality?
- Almost always because the clock starts at the wrong moment. Many systems record response time from when a user first opened the record rather than from when the lead arrived, which excludes the entire period the lead sat unopened. That is precisely the period you are trying to measure, so the dashboard reports the one interval that does not matter.
- Should I measure median or average response time?
- Median. Response-time distributions have a long tail, and a small number of leads answered days later will inflate an average until it describes nothing. Report the median as the headline and the 90th percentile alongside it, which shows how bad the tail actually gets.
- Does an automated text count as a response?
- Measure it, but as its own metric. Time to automated acknowledgement and time to first human contact answer different questions, and reporting them as one number is how vendors produce impressive figures. Both are worth knowing, and the gap between them is worth knowing too.
- What is a realistic target?
- Under five minutes to acknowledgement and under an hour to a human attempt, measured as a median with the after-hours window included rather than excluded, is a demanding but achievable standard for most service businesses. Excluding after-hours to make the number look better defeats the purpose of measuring.
- What should I measure alongside response time?
- Attempts per lead. Response time tells you whether the conversation started. Attempts per lead tells you whether anyone kept trying, and for most businesses the second distribution reveals a larger and more recoverable loss than the first.
Sources
- 1.Oldroyd, McElheran and Elkington, The Short Life of Online Sales Leads, Harvard Business Review, March 2011 Context for why measurement matters: 23% of 2,241 audited companies never responded, and the average among responders was 42 hours.
- 2.The Lead Response Management Study The response-decay curve underlying the targets discussed here.
Two numbers, one export
Your real median time to first contact, and the share of leads touched fewer than four times. Most businesses have never seen either, and the second one tends to decide what gets fixed first.
Apply For a Growth Partnership