The Annual Customer Review SOP

Purpose

To re-derive, once per year, the facts about the customer book that every other lifecycle activity depends on: how often each customer actually calls, which records are still reachable, which customers carry the forward value, where the book is dangerously concentrated, and how much of next year's demand the existing book will supply on its own.

This is the strategic review, not the working one. The quarterly cycle sorts customers by recency and produces a call list for the next ninety days; it consumes the numbers this review produces and does not question them. If nobody rebuilds those numbers annually, the quarterly cycle keeps running against intervals derived from data that has aged out, and it will drift into calling people who are not late and ignoring people who are.

Scope

In scope: every customer record with at least one completed paid job, active or not. Interval derivation, cohort survival, value tiering, record retirement, concentration measurement, and next-year demand sizing.

Out of scope: running any outreach. This review produces lists and numbers; the quarterly review and the reactivation cycle do the calling. Do not let the annual review turn into a campaign, because the analysis stops the moment the phones start.

Also out of scope: pricing decisions, staffing decisions, and marketing budget. The review is an input to all three and should be finished before any of them are set.

Roles and responsibilities

Timing and inputs

Run it in your slowest month, after the fiscal year is closed enough that job completion and payment status are settled. It takes one focused day plus a half day of follow-up for a book under about a thousand records.

Inputs: completed and paid job history for all years available, current contact details, equipment records, technician assignment per job, and last year's version of this review if one exists.

Procedure

1. Freeze the data and write the cutoff on the document

Pick a cutoff date and use it for every number in the review. Every rate, count, and tier carries that date. A review whose numbers were pulled across three different weeks cannot be compared against next year's, and a year-over-year comparison is most of the value here.

2. Recompute the service interval for every customer and every service type

For each customer with three or more completed jobs, recompute the median gap between consecutive jobs using the full history including the year just ended. For each service type, recompute the shop-wide median gap across all customers, which becomes the default applied to records with too little history of their own.

Compare against last year's figures and investigate any service-type default that moved by more than about a month. A default that shortens usually means outbound reminders are generating jobs that are being counted as naturally-timed calls; a default that lengthens usually means the customer base aged, equipment was replaced, or a segment quietly left.

3. Rebuild the cohort survival table

Take a first-job cohort old enough to be readable (at least two typical intervals past its end, three preferred) and count how many reached a second, third, and fourth job, dividing each survival rate by the population of the previous stage. Keep last year's table beside it. A single year's table tells you where the losses are; two years of tables tell you whether anything you did last year worked.

4. Re-tier the book on forward contribution

Three tiers, decided on evidence in the record, not on affection:

• Tier A: short interval, consistent, no payment issues, work you are set up to do well.
• Tier B: genuine repeat customers on a longer or less consistent interval.
• Tier C: single-job records, unreachable records still in the count, and repeat customers whose work is a poor fit for the shop's capability or margin.

Tier is a forward statement, so a customer with a large history of work you no longer do is not Tier A. Write the tier to the record. It drives call order, notice sequencing on a rate change, and who gets a person versus a message.

5. Measure concentration in three directions

• Customer concentration. What share of last year's completed job hours came from the top ten and top twenty customers.
• Technician concentration. For each tech, what share of Tier A accounts have that tech on all three of their last three visits. This is your exposure if that person resigns.
• Source concentration. What share of new customers came from a single referral source, builder, or property manager.

Each of these is a standing risk that is invisible until it fires. The review's job is to state the number, not to fix it that day.

6. Retire records permanently and correct the reachable set

Working from the office lead's list, mark records as retired where there is real evidence: mail returned, phone disconnected, property sold, customer deceased, or a documented request not to be contacted. Retired means excluded from every list and every denominator going forward, and it means preserved, not deleted. The history stays; the record stops counting as a customer you could call.

Two rules that keep this honest. A record is never retired for silence alone, only for evidence. And a do-not-contact request is permanent and survives every future list build, including one built by somebody who was not here this year.

7. Size next year's demand from the book

Assign each tier a rough probability of initiating paid work in the next twelve months, taken from your own survival table rather than optimism, then multiply by tier headcount and sum. Compare the total against last year's completed job count. The difference is the volume that has to come from new customers, one-time work, and anything you do to lift survival.

This is the number that turns the review into a plan, and it is the one owners skip.

8. Set two or three targets, not ten

Pick from the review's own findings: a survival rate to lift at a named transition, a concentration figure to bring down, a reachable-record percentage to improve. Each target names its base and its measurement date so next year's review can score it without argument.

9. Publish one page and hand off the outputs

One page: the cutoff date, the interval defaults, the survival table with last year's beside it, tier counts, the three concentration numbers, the demand estimate, and the targets. Hand the tier assignments and the interval defaults to the quarterly review, and the retired-record list to whoever maintains the contact data.

Worked example: one shop's annual review

The shop starts with 612 customer records carrying at least one completed job.

Step 6 retires 100 of them: returned mail, disconnected numbers, sold properties, and four do-not-contact requests. That leaves 512 reachable records, so 84% of the raw book is contactable. The 100 retirements matter beyond tidiness, because every rate computed against 612 would have been understated by about a sixth for no reason other than dead records sitting in the denominator.

Step 2 finds one service type's shop-wide default interval has moved from 14 months to 12. Investigation shows the shop began sending due reminders midway through the prior year and those reminder-generated jobs were counted as naturally-timed calls. The correct action is to exclude reminder-generated jobs from the baseline and keep the default at 14 months, which prevents next year's quarterly cycle from flagging customers as late two months before they are.

Step 4 tiers the 512 reachable records as 61 Tier A, 158 Tier B, and 293 Tier C, which sums back to 512.

Step 5 finds the top twenty customers supplied 24% of last year's completed job hours, and that one tech is the sole tech on 19 of the 61 Tier A accounts, which is 31% of Tier A. Neither number is a crisis on its own; both are written down, and the tech figure becomes a routing target rather than a conversation.

Step 3 produced the survival table these probabilities come from. Of 512 accounts whose first job fell in the review window, 331 returned for a second (65%), 216 of those came back for a third (65% again), and 178 of those for a fourth (82%). Read stage over immediately preceding stage, not over the original 512, or the numbers collapse and every tier looks like churn. The flattening after the third job is what sets the tier probabilities.

Step 7 sizes demand. Using probabilities read off that table (Tier A at 0.9, Tier B at 0.55, Tier C at 0.2), the arithmetic runs 61 times 0.9 is 54.9, plus 158 times 0.55 is 86.9, plus 293 times 0.2 is 58.6. The total is about 200 expected repeat calls in the next twelve months.

Last year the shop completed 340 jobs. So the existing book is expected to supply roughly 200 of them, which is 59% of last year's volume, and about 140 jobs have to come from somewhere else. That single sentence is what changes behavior: an owner who believed the book carried the year now knows that two-fifths of next year's work is not in it, and can decide between chasing new customers, lifting survival, or accepting a smaller year, before the slow months arrive rather than during them.

Step 8 sets two targets off those findings: lift the second-to-third job survival rate from 65% to 72% by the next cutoff date, and bring the single-tech share of Tier A accounts from 31% to under 20% through routing changes.

Verification

• Every number in the one-page summary carries the same cutoff date, and every percentage names its base in the same line.
• The tier counts sum to the reachable record count, and the reachable count plus retirements sums to the raw book.
• No record was retired without a stated evidence reason, and every prior do-not-contact request from earlier years still appears in the current exclusion set.
• Interval defaults were checked against reminder-generated jobs before being adopted. If nobody looked, assume the defaults are contaminated and short.
• Last year's targets were scored, not quietly replaced. A review that sets new targets without grading the old ones is a planning exercise, not a review, and it will do this again next year.
• The quarterly review's next cycle actually consumed this year's tiers and intervals. If the quarterly list still runs on last year's numbers, the annual review changed nothing.

References

• See related: The Quarterly Customer List Review SOP (the working cycle that consumes these outputs)
• See related: The Lifecycle Stage Where Most Shops Lose People (the cohort survival method used in step 3)
• See related: How to Use Service History to Predict the Next Call
• See related: The Dormant List Audit SOP (record retirement and reachability)