The First 90 Days of a New Customer Relationship
Why this matters
A shop usually learns whether a new customer became a real customer about eighteen months after the fact, when the second job either arrives or does not. By then the visit that decided it is unrecoverable, the tech has forgotten the property, and the only available response is a generic win-back to a stranger. The first 90 days after a first job carry a signal that predicts the outcome well ahead of the outcome, and it is measurable from records a shop already keeps. This card is about reading that window as an early-warning instrument, not about the onboarding tasks inside it.
What the window is actually deciding
Nothing about the first job decides whether a customer returns, as long as the work was competent. Competent work is the entry fee, not the differentiator, and a shop that does good work and still loses two thirds of its first-time customers is the normal case, not a broken one.
What the 90 days decides is whether the shop becomes a name the customer reaches for by default. That is a memory question, and memory of a service call decays on a curve that is steep early. A customer who has had no contact with you 90 days after a job has, functionally, filed you under "somebody I used once." They will not consciously reject you next time. They will search, because searching is what you do when no name comes to mind.
This is why the leading indicator is contact rather than revenue. A second job inside 90 days is a great sign and it is rare, because most service intervals are longer than 90 days. A second contact of any kind is common, achievable, and predicts the same thing.
The milestone map
These are the moments inside the window where the relationship is either reinforced or left to decay. Treat the timings as defaults to tune to your service intervals, not as fixed rules.
| Point | What happens or should | What it tells you |
|---|---|---|
| Day 0 | Job completed, next need named and logged | Whether there is anything to build on at all |
| Day 1 to 2 | Confirm the fix held | Catches a callback before it becomes a complaint |
| Day 7 to 14 | Deferred item decision, if one exists | The customer's real appetite for more work |
| Day 30 | First unprompted contact from the customer, or none | The strongest single signal in the window |
| Day 45 to 60 | Shop-initiated touch tied to the named next need | The last cheap intervention before the trail is cold |
| Day 90 | Second contact has occurred, or has not | The window closes and the customer's default is set |
The day 30 row is the one to read most carefully, and it is passive. You are not doing anything at day 30. You are checking whether the customer did.
The two kinds of second contact, and why they are not equal
A shop-initiated touch and a customer-initiated one both count as contact, and both correlate with the customer coming back, but they mean different things and the difference matters when you are deciding where to spend hours.
Customer-initiated contact is a preference already formed. They had a question, a new problem, a related need, and your name came up. This customer is close to a repeat customer already, and the work is not persuasion, it is being easy to reach and fast to respond.
Shop-initiated contact is you building the memory that would otherwise decay. It works, but it works less reliably, and it works only when it carries a real reason. It is also the only one you control, which is why the whole retention program lives here.
The practical read: a low count of customer-initiated contacts in the window means the first visit is not producing the impression you think it is, and no amount of outbound touching will fully fix that. A low count of shop-initiated contacts means you have a process gap, which is much easier to close.
The worked example
A shop pulled every first-time customer from one quarter: 74 records. For each, it counted whether any contact of any kind occurred within 90 days of the first job, then checked whether that customer booked a second job within the following 12 months.
Of the 74, 29 had at least one contact inside the 90-day window and 45 had none. Those sum back to 74, which is worth confirming, because a contact log that does not reconcile against the customer count usually means calls are being logged against jobs rather than against customers, and the customers with no job have no place to hang a note.
Of the 29 with contact in the window, 22 booked a second job within 12 months, a rate of 76%. Of the 45 with no contact in the window, 11 booked, a rate of 24%. Overall, 33 of 74 booked again, or 45%.
Splitting the 29 by who initiated: 12 were customer-initiated and 17 were shop-initiated. Of the 12 customer-initiated, 11 later booked, a rate of 92% of that subgroup. Of the 17 shop-initiated, 11 later booked, a rate of 65% of that subgroup. Those two elevens sum to the 22 above.
Read those three rates against each other rather than in isolation. Customer-initiated at 92% is nearly a guarantee, but it applied to only 12 of 74 customers, or 16% of the cohort, and the shop did not cause it. Shop-initiated at 65% applied to 17 customers, and the shop did cause it. No contact at 24% applied to 45 customers, which is 61% of the cohort, and is the single largest block.
That last figure is where the money is. The shop was leaving 61% of its new customers untouched for a full quarter after paying to acquire them. Moving even half of those 45 into the shop-initiated bucket at that bucket's observed 65% rate, against the 24% they were getting, would be roughly 22 or 23 customers shifted at a gap of about 41 percentage points, which is on the order of 9 additional second jobs from one quarter's intake.
Two honest limits on that projection. First, it assumes the 45 untouched customers would respond like the 17 touched ones, and they will not, because the 17 were probably touched precisely because they had a logged next need to touch them about. Expect the real lift to be lower. Second, the 24% baseline is not zero, so the gain is the difference, not the whole 65%. A shop that models this as "65% of 22 customers," which is about 15 jobs, instead of "the 41-point gap applied to 22 customers," which is about 9, will forecast roughly 1.6 times the actual result and then conclude the program failed.
What changes the answer
A very long service interval. If your typical customer genuinely needs you once every two or three years, a 90-day window is measuring something further from the buying decision, and the contact you make inside it has to be lighter and more informational or it reads as pressure. The window still predicts; the intervention changes.
Emergency-driven trades. Where most first calls are urgent failures, the customer's memory of you is stronger and decays more slowly, because the event was significant to them. The 90-day contact matters less and the quality of the emergency handling matters more. Check this at your own shop before assuming it: many owners believe it and their records do not support it.
Rental properties and property managers. The occupant and the decision maker differ, and contact with the occupant does not build the relationship that produces the second job. Count contact with the payer only, or the metric will look healthy while the account quietly lapses.
A first job that was a warranty or callback visit. Contact in the window is guaranteed and means nothing about preference. Exclude these from the cohort or they will inflate the contact rate and flatten the signal.
A shop small enough that the numbers are noise. Below about 30 first-time customers in a quarter, a 90-day contact rate swings on a handful of records. Use two or three quarters combined, and be slow to react to a single quarter's movement.
How to read this as an instrument, not a target
The window's value is that it fails early and loudly. Watch two numbers per quarter and nothing else.
Contact coverage: the share of new customers with any contact inside 90 days. This is a process measure. It should be stable, it is fully within your control, and a drop in it precedes a drop in repeat rate by roughly a year.
Customer-initiated share: the share of new customers who contacted you unprompted inside 90 days. This is an outcome measure and a proxy for how the first visit landed. It moves slowly. A sustained decline here is a signal about the work or the crew, not about the outreach.
The failure mode is turning coverage into a quota, which is what happened at the shop above the following year. Coverage rose because the office started logging courtesy texts that carried no reason. The repeat rate did not move. Contact with no substance is measured as contact and is not contact, and the only reliable way to catch it is to read a sample of the actual touches rather than the count.
How to verify you got this right
Reconcile the buckets to the cohort count. Contacted plus not-contacted must equal the total first-time customers. If it does not, contact is being logged somewhere that does not tie to a customer record.
Exclude callbacks and warranty returns from the contact count. These are the single largest source of a falsely healthy coverage number.
Read ten actual touches, not just the count. For each, ask what reason the touch carried. Touches with no reason are noise being counted as signal.
Check the repeat-rate definition once and reuse it. A second job counted as "any subsequent invoice" and a second job counted as "a new job, excluding return visits on the same fault" will differ by a large margin at most shops. Pick one, write it down, and apply it to every bucket in every quarter.
References
- U.S. Small Business Administration, small-business customer-retention guidance
- Trade-standard practice for service-history and customer-contact logging
- See related: The New Customer Intake SOP
- See related: How to Earn the Second Job During the First Visit
- See related: The Signals That a Customer Relationship Is Deepening