Short answer: Both. A 20-lead test measures the source and how your office handled it, mixed together. To pull them apart, log for every lead when it arrived, when someone responded, how many tries were made, and whether it fit your service area and the jobs you want. Judge the source only on the leads that were handled properly. Twenty is a small sample, so treat it as a signal, not proof.
Why every lead test is two tests at once
Say you try a new source: a lead marketplace, Local Services Ads, Facebook lead forms, a new landing page. Twenty leads come in, three turn into jobs, and you decide whether to keep paying.
Those three jobs depend on two things. One is the source: did it send people who need what you do, where you do it? The other is your office: did someone reach those people quickly, ask the right questions and get them on the schedule? If leads sat for a day during a busy week, the source looks worse than it is. If your best estimator happened to work every lead, it might look better than it is.
Unless you record both, you can't tell which one you're grading.
What to log for each lead
Use a simple sheet. One row per lead, these columns:
- Source (and campaign or listing, if you know it)
- Date and time it arrived
- Time of first response from a person
- Number of follow-up attempts
- Fit: yes or no, with a short reason (area, job type, real buyer)
- Estimate sent: yes or no
- Booked: yes or no
- Job value if booked, once you know it
If you can't fill in the response time for a lead, write "unknown." That alone tells you something about your process.
Score fit separately from outcome
This is the step most tests skip. Decide fit from what the lead asked for and where they are, before you look at whether it booked. A homeowner in your area asking for a replacement estimate is a good-fit lead, even if nobody called them back and they hired someone else. That lead counts for the source and against the handling.
When you're done, you have four groups:
- Good fit, handled well
- Good fit, handled poorly
- Poor fit, handled well
- Poor fit, handled poorly
Judge the source mostly on the share of good fits. Judge the office on what happened to the good fits.
Why 20 is a small sample
Twenty leads can come in during one unusual stretch: a heat wave, a slow week, a week your office was short a person. Seasonality alone can swing results in HVAC, plumbing and roofing. Use 20 to spot an obvious problem, like most leads being outside your area, or most good fits never getting a second call. Don't use it to prove a source works or doesn't. If the call is close, keep testing.
Judge on cost per booked job, not cost per lead
Cost per lead is what the source charges you, divided by leads. Cost per booked job is what you spent on the source divided by jobs that actually got booked from it. A cheap lead that never books is expensive. A pricier lead that books more often may be the better deal. Track booked jobs, and job value where you can, so you're comparing what actually pays.
Keep in mind that cost per booked job also includes your office's handling. That's one more reason to log response times, so you know if a bad number came from the source or from the follow-up.
Run the test fairly: set the rules before it starts
Before the first lead comes in, agree on:
- How fast a new lead gets a first response during business hours, and what happens after hours
- Who owns each lead from the test source
- How many tries, over how many days, before a lead is closed
- What questions get asked every time
- Who logs each lead in the sheet
Then check, lead by lead, whether the rules were followed. If they weren't, you'll know which leads to set aside when you judge the source.
For agencies running paid tests
If you're running a test for a client, agree on the handling rules with them up front, in writing. Ask who will respond to leads and how fast, and who keeps the log. Then review the sheet together at the end. It keeps the conversation about what actually happened, not about whose fault it was. If you work with home-service companies and want to know how we work with partners, see our partner page.
For the broader version of this check, see Bad leads or slow follow-up? How home-service shops can tell the difference.
Where CallBack Engine fits
A source test is only fair if every lead gets handled the same way. CallBack Engine is a managed lead-response service. We follow up on the inquiries you already get, gather the job details, check them against the rules you approve, and move qualified leads to scheduling or a team handoff, with the outcome recorded for each one. That gives you a consistent process and a written record, so the source is closer to being the only thing you're testing.
We don't sell leads, and we can't promise bookings or results. If you want to talk through how your leads are handled today, review your lead flow with us.
FAQ
How many leads do I need to judge a lead source?
There's no exact number. Twenty is enough to spot obvious problems, like most leads being outside your area. For a closer call, keep going until you've seen the source through more than one kind of week.
What's the difference between cost per lead and cost per booked job?
Cost per lead is your spend divided by the number of leads. Cost per booked job is your spend divided by the jobs actually booked from those leads. The second one tells you what the source is really costing you.
Who should log the leads during a test?
One named person, usually whoever handles the office or the leads. If several people touch leads, they all log in the same sheet, and one person checks it every day.
What if the office is too busy to follow the rules during the test?
Then that's part of what the test found. Mark which leads didn't get handled by the rules and set them aside when you judge the source. If it happens a lot, fix the handling before you spend more on testing sources.
How long should a lead source test run?
Long enough to get a reasonable number of leads and to cover more than one kind of week. For seasonal trades, a test that only runs during a heat wave or a slow stretch can mislead you either way.