Overflow Jobs: What to Do With the Calls You Cannot Run Yourself

Overflow Jobs: What to Do With the Calls You Cannot Run Yourself

Every locksmith shop turns work away, and most owners have never counted how much. The call comes in, it is forty minutes the wrong direction, or it is a safe and you do not do safes, or both vans are committed until Thursday, and the answer is some version of "sorry, we can't help you with that today." Then the call ends and the job is gone — not just this job, but the customer record, the phone number, the vehicle, and the next three years of that person's key work.

As of August 2026 the tooling to do better exists and is cheap, but the reason most shops still lose overflow is not tooling. It is that nobody has ever written down what happens to a job the shop cannot run. Without a rule, the answer defaults to whoever answered the phone, and their answer is almost always the fastest one, which is "no."

This piece is about the alternatives. What overflow actually is, the three honest things you can do with it, how a reciprocal referral network gets built and kept honest, what has to be captured on the call for any of this to function, and the liability and quality-control questions that decide whether subcontracting makes you money or costs you a review.

Overflow is four different problems wearing one label

"We're too busy" is the excuse for all of them, and only one of them is actually about being busy. Separating them matters because each has a different correct answer.

Geographic overflow. The job is real, the work is in your wheelhouse, and it is on the wrong side of the metro. The drive kills the margin and blows up the rest of the day's schedule. This is the single most common category in any metro with real sprawl, and it is the easiest one to solve with a partner, because the exact job that is unprofitable for you is a fifteen-minute drive for somebody else — and vice versa, in the other direction, which is what makes reciprocity possible.

Specialty overflow. The job is close but you do not do it. High-security commercial cylinders and restricted keyways. Safes — opening, combination changes, relocking devices. ECU and module programming beyond what your equipment covers. Antique or classic vehicle work. Access control and electronic hardware with a network component. Turning this away is not laziness; it is correct. Attempting a safe with no safe experience is how a shop acquires a very expensive lesson and a very public review.

Capacity overflow. The work is yours, in your area, and there is nobody free until Thursday. This is the one that feels like a good problem and quietly is not, because the customer with a car that will not start is not waiting until Thursday. They are calling the next result.

Trough overflow — the one nobody talks about. The mirror image: it is a slow Tuesday, both vans are idle, and the shop across the metro is buried. If you have a reciprocal relationship, you are running their overflow at a margin you agreed to in advance. If you do not, you are washing the van. Most owners set up referral relationships thinking about the busy weeks and then discover the slow weeks are where the arrangement actually pays them.

The three honest options

For any job you cannot run yourself, there are exactly three outcomes. Everything else is a variation on one of them.

1. Refer it away. You give the caller a name and a number and the relationship ends there. It is fast, it is free, it carries almost no liability, and it converts a customer into someone else's customer. Done reciprocally it is a genuine trade. Done one-directionally it is a donation.

2. Subcontract it. You keep the job. A partner runs it. You invoice the customer, you pay the partner, and the difference is your margin. The customer stays yours — your name on the invoice, your record, your review, your follow-up when they need a spare key next spring. You also inherit the partner's work quality as your own, which is the real cost of this option and the reason it needs rules.

3. Schedule it later. You keep it entirely and give the customer a time. This works for genuinely non-urgent work — a spare key, a rekey, a commercial job with a lead time. It does not work for a lockout or a no-start, where the caller's actual constraint is now and any answer involving Thursday is a decline in a polite voice.

Here is the comparison an owner should keep in view when writing the shop's own rule:

The practical read: subcontract geographic and capacity overflow, refer true specialty work, schedule only what is genuinely schedulable. Most shops do the reverse by accident — they refer away the easy geographic jobs because it is the fastest sentence to say, and try to schedule emergencies because they do not want to admit they cannot cover them.

Building a reciprocal network that survives contact with a busy week

A referral network is not a Facebook group. It is a small number of specific relationships with specific terms, and it works or fails on whether both sides believe the flow is roughly even.

Pick partners by complement, not by friendliness. The right partner is one whose weaknesses are your strengths. If you are automotive-heavy and there is a shop two suburbs over that is commercial-heavy, that is a real trade — you send them the master-key system, they send you the all-keys-lost. If you are both automotive and both in the same three ZIP codes, you are competitors with a pleasant relationship, and the first busy month will show it.

Geography first. The cleanest arrangements are with shops whose service area barely touches yours. You cover the north half of the metro, they cover the south, and the overlap is a strip where either of you can go. There is no serious competitive tension, and every referral is genuinely useful to the receiver rather than a job they would have gotten anyway.

Three to five partners is the right number. One partner is a single point of failure — when they are also buried, you are back to "sorry." Fifteen partners is a directory you will not maintain and cannot vouch for. Three to five means you know each of them by name, you know what they are actually good at, and you know whether they show up.

Write down what each partner is for. Not in your head. A one-page sheet, or a note in whatever system you already use, listing: partner name, phone, what they do that you do not, which ZIPs or which side of the metro, their hours and whether they run nights, whether they carry the insurance you need for subcontracted work, and their agreed rate structure. Whoever answers your phone at 9 PM cannot use knowledge that lives only in the owner's head.

The reciprocity ledger

The failure mode of every referral network is one-sided flow, and it always goes unnoticed until somebody is quietly resentful. The fix is boring: count.

Keep a simple running tally per partner — jobs sent, jobs received, month by month. It does not need to be software. A note or a spreadsheet with two columns per partner is enough, and the act of maintaining it changes behavior more than the numbers themselves.

What the ledger gives you:

• An early warning. Twelve sent, one received, three months running is a conversation to have while it is still a conversation and not a grievance.
• Something to say. "I've sent you nine this quarter and had two back — is my north-side stuff not working for you?" is a specific, non-hostile question. "I feel like this isn't fair" is not.
• A basis for changing terms. If the flow is structurally uneven — say you are in the denser half of the metro and simply generate more overflow — that is fine, but then the arrangement should be a paid one rather than a trade. Uneven flow is not a problem. Uneven flow pretending to be an even trade is.

Review it quarterly. Two shops that both look at the same tally twice a year almost never fall out.

Referral fees and splits, described as structures

There is no published market rate for locksmith referral fees, and anyone who quotes you one is guessing. What there are is four common structures, each with a predictable set of arguments attached. Pick one deliberately per partner and put it in writing.

Straight reciprocity — no money changes hands. Each shop keeps whatever it collects on jobs the other sends. Simplest, zero accounting, and it only works when the flow is genuinely two-directional and roughly comparable in job value. Best between complementary shops of similar size.

Flat fee per referred job. The receiving shop pays a fixed amount for each job that turns into work. Easy to compute and easy to argue about, because a flat fee is a large share of a small job and a rounding error on a large one. Works best where job values are consistent.

Percentage of the job. The receiving shop pays an agreed percentage of what they collect. Scales correctly across job sizes, but requires trust in the reported number — you are relying on the partner's invoice being what they say it is. Best with partners you have a track record with.

Subcontract pricing — the cleanest one. Not a referral at all. You quote the customer your price, the partner quotes you their wholesale price for running it, and your margin is the difference. Nobody is reporting anything to anybody, because you are the customer's vendor and the partner is yours. Everything is on two invoices that both parties can see. This is why subcontracting, not referring, is the structure most shops end up preferring for high-volume overflow.

Whichever you choose: agree it before the first job, not after. The single most common cause of a partnership blowing up is two shops discovering after a $700 job that they had different assumptions about who owed whom what.

What has to be true operationally, or none of this works

The structures above are the easy part. The reason most overflow arrangements collapse is operational, and it is nearly always one of these five things.

1. Capture the full job even when you are not running it

This is the discipline that separates a shop with a working overflow process from a shop with good intentions. The intake does not get shorter because the job is going somewhere else. If anything it gets longer, because a partner arriving cold needs more written down than your own tech who knows your area.

For an overflow job the minimum capture is: name, callback number confirmed rather than dictated, exact location including cross-street or business name, vehicle year/make/model or the door and hardware type, the actual symptom in the customer's words, whether keys exist, whether proof of ownership will be available, urgency, and — critically — what you told them. If your phone said "someone will call you within fifteen minutes," that promise is now an obligation and the partner needs to know it was made.

A referral without a record is just a phone number given away. You cannot follow up on it, cannot count it in the ledger, cannot check whether the partner ever called, and cannot rescue it when they did not.

2. Warm transfer or callback — pick one per partner and per situation

There are two ways to hand a caller to a partner and they fail differently. We have written about the mechanics of both in warm versus blind call transfers, and the short version applies directly here.

Warm transfer — you stay on, introduce the job, hand it over — is better for the customer and impossible to guarantee. If the partner does not pick up, and you have already told the customer you are transferring them, you have manufactured a worse experience than a plain decline. A blind transfer into an unanswered line is the worst outcome available: the customer's problem is now unsolved and they think you dropped them.

Structured callback — you take everything, tell the customer exactly who will call and in what window, and send the partner the record — is less impressive and far more reliable. It also survives the partner being on a roof or under a dash, which they usually are.

The rule that works: warm transfer only when the partner is a live human who has agreed to be interruptible, and only after they have actually answered. Everything else is a callback with a specific promise attached and a specific person responsible for it.

3. Decide who invoices the customer — before the job

This is the question that determines everything else, and it has exactly two clean answers.

You invoice. It is a subcontract. Your price, your invoice, your payment collection, your customer record, and you pay the partner separately at your agreed wholesale rate. The customer never sees a second company's name on a bill. This keeps the relationship and gives you margin. It also means you eat the collection risk and you own the outcome.

They invoice. It is a referral. Their price, their invoice, their customer. You may or may not get a fee. You do not own the follow-up.

What does not work is any hybrid — the partner collecting on-site and you invoicing later, or the customer receiving two bills, or nobody being sure. If a customer ever gets asked to pay twice, or gets an invoice from a company they have never heard of, you will spend more on that phone call than the job was worth.

4. Liability, insurance and licensing on subcontracted work

Treat this as the seriousness gate on whether you subcontract at all.

When you subcontract, the customer's contract is with you. You quoted, you scheduled, you invoiced. If the partner damages a door, a dash, a steering column, or an ignition, the customer's complaint is directed at your company — and in many cases so is the legal exposure.

The minimum you should require of any shop you subcontract to:

• Proof of general liability insurance, with the certificate on file and a diary note for its expiration date. Not "yeah, we're covered." The actual certificate.
• Whatever licensing your state requires. Locksmith licensing is state law and varies widely — some states run a mandatory program, others have none. The Associated Locksmiths of America maintains industry-facing resources, and your state's licensing authority is the actual source of truth. Verify your partner's standing yourself rather than taking it on trust.
• A written understanding of who is liable for damage, including whether the partner will handle a claim directly and how a dispute gets resolved.
• Clarity on worker classification. A regularly-used subcontractor is a business relationship with real tax and employment implications. Occupational and employment context for the trade is published by the Bureau of Labor Statistics, but the classification question itself is one for your own accountant rather than a competitor's advice.

If a prospective partner is vague about insurance, that is the whole answer. Refer to them if you like; do not subcontract to them.

5. Quality control, because their bad job becomes your review

This is the part owners underestimate. On a subcontracted job, the review lands on your Google profile. The customer does not know or care that a different van showed up.

Three controls, in order of cost:

Start narrow. Give a new partner the jobs where a bad outcome is survivable — a straightforward rekey, a spare key, a non-urgent commercial call — before you give them a stressed customer at midnight with a damaged door.

Follow up yourself. A single call or text the next day — "did the tech get you sorted?" — is the entire quality control system for most small shops, and it is the difference between finding out from you and finding out from a one-star review three days later. It also reinforces to the customer that they are your customer.

Watch the review pattern, not the incident. One bad job is noise. Two complaints about the same partner is a pattern, and the correct response is to stop sending them work, not to have a difficult conversation about it a fourth time.

The phone layer decides all of this

Here is the thing that makes the entire discussion above conditional: you cannot triage, refer, subcontract, or schedule a call that nobody answered. Every option in this article requires that a human or a system picked up, understood what the job was, and made a decision.

That is not a small caveat. Overflow calls arrive at exactly the worst moments by definition — when you are already busy, which is when the phone is most likely to ring out, and after hours, when the wrong-specialty and wrong-geography calls tend to concentrate. Concurrency is the specific structural failure: a person answers one call at a time, so the third call in the same minute gets a busy signal regardless of how good your referral network is. The arithmetic is spelled out in the busy-signal problem.

The other precondition is triage. Deciding whether a job is yours, a partner's, or a Thursday requires knowing the service, the location, and the urgency — which is the same qualification work described in emergency versus scheduled call triage. If nobody is asking those questions, every call is either accepted blind or declined blind.

And if you run more than one service area — or you are effectively running two, yours and a partner's — the routing question is the same one covered in multi-location call routing: rules that decide, per call, where it goes, rather than a person guessing at 11 PM.

Where an AI receptionist actually fits

Narrowly, and usefully. It is not the referral network. It is the layer that makes the referral network reachable.

It answers everything, including the concurrent calls. The fourth call in a minute gets the same intake as the first. That alone converts a category of jobs — the ones that currently ring out during a busy stretch — from lost to at least decided.

It runs your rules instead of a tired judgment call. Your rules, configured once: this ZIP plus this service equals a partner job; this service is one we do not do at all; this ZIP inside our area with a non-urgent job goes on the schedule. The rules are yours; the consistency is the software's contribution. It applies the same logic at 3 AM that you would apply at 10 AM, which is more than most humans manage.

It creates a record either way. This is the underrated part. Whether the job goes to a partner, onto your calendar, or gets declined outright, you end up with a structured record: who called, from what number, for what, where, and what happened. That record is what feeds the reciprocity ledger, the follow-up call, and the honest count of how much work you are actually turning away — a number almost no shop can produce today.

It does not run the job, cut the key, or vouch for the partner. It is call-answering software. What it changes is that the overflow decision gets made, on every call, instead of being decided by whether anyone was free to pick up.

On cost, plainly: KeyBot Lite is $149 a month — message-taking only, no quoting and no booking, 100 calls included and 50 cents a minute after, with the first 5 answered calls free on a 7-day trial. For a shop whose overflow problem is simply "the call rang out and I never knew," that is often the whole fix. If you want the bot to quote from your confirmed price sheet and book on the call, that is Core at $500 a month for 500 AI minutes, Pro at $750 for 1,000, Elite at $1,200 for 2,500, each with a 14-day free trial. All tiers are on pricing, the locksmith-specific setup is on the locksmith receptionist page, and if you want to hear it before reading another word, have it call you as your own shop — about 30 seconds, first demo free.

One adjacent category worth naming: motor club and roadside dispatch work is structured overflow arriving from the other direction, and it has its own economics and its own call-handling shape. That is covered separately in roadside and motor club dispatch calls.

The five rules to write down before Monday

Not a project. Five sentences on one page, given to whoever answers your phone.

1. These services we do not do at all — list them — and they go to [named partner] by structured callback within 15 minutes.
2. Outside these ZIPs, a job goes to [named partner] as a subcontract. We invoice, we collect, we pay them [agreed rate].
3. Inside our area and non-urgent, we schedule it ourselves — always. We do not refer away work we can do.
4. Every overflow job gets a full intake anyway, including what we promised the caller and when.
5. We follow up on every subcontracted job the next day. Every one.

If those five lines exist and are actually followed, you have a working overflow process. If they do not, you have a shop that says no on a Tuesday afternoon and never counts the cost.

The bottom line

Overflow is not a sign of a healthy business or a failing one — it is a permanent structural fact of running a service shop in a metro. What varies is whether it is handled or improvised. The three options are refer, subcontract, and schedule; subcontracting is usually right for geographic and capacity overflow because it keeps the customer record and the margin, referring is right for genuine specialty work you should not touch, and scheduling only works when the caller's need is not immediate. Build the network from three to five complementary partners, keep an honest reciprocity ledger, agree the money structure before the first job, demand a real insurance certificate before you subcontract anything, and follow up on every subcontracted job yourself because the review is yours regardless of whose van showed up. And recognize the precondition under all of it: none of these decisions can be made about a call that nobody answered. Fix the answering layer first — that is the one part of this that takes about ten minutes and can be tested by having the AI call you as your own shop before you commit to anything.

Frequently asked questions

Should I refer an overflow job or subcontract it?

Subcontract it when the work is inside your competence and only the geography or your schedule is the obstacle, because subcontracting keeps the customer record, the invoice, and the margin with you. Refer it when the work is a genuine specialty you do not perform — safes, restricted commercial keyways, module work beyond your equipment — since taking responsibility for work you cannot evaluate is how a shop acquires a damage claim. The deciding question is simple: can you judge whether the job was done correctly? If yes, subcontract. If no, refer.

How many referral partners does a locksmith shop actually need?

Three to five is the working range for most single-metro shops. One partner is a single point of failure, because the week you are buried is often the same week they are, and fifteen partners is a list you cannot maintain or vouch for. Choose partners whose strengths complement yours rather than duplicate them — a commercial-heavy shop for your automotive-heavy one, and a shop covering the opposite side of the metro — so that every job you send is genuinely useful to them and the flow can run both directions.

What is a fair referral fee between locksmith shops?

There is no published market rate, and any specific number quoted to you is a guess rather than a benchmark. What exists are four structures: straight reciprocity with no money changing hands, a flat fee per referred job, a percentage of the collected job value, or subcontract pricing where you quote the customer and the partner quotes you a wholesale price. Subcontract pricing is usually cleanest because nobody has to report a number to anybody — the margin is simply the difference between two invoices that both parties can see.

Who is liable if a subcontracted locksmith damages a customer vehicle?

The customer contracted with you, so the complaint and often the exposure land on your company regardless of whose van arrived. That is why you should hold a current general liability insurance certificate on file for every shop you subcontract to, with a diary note for the expiration date, and confirm whatever licensing your state requires rather than accepting a verbal assurance. Put the damage-responsibility understanding in writing before the first job, and start any new partner on low-stakes work before trusting them with a stressed customer at midnight.

How much does an AI receptionist for a locksmith shop cost?

KeyBot Lite is $149 per month with 100 calls included and 50 cents per minute after, and the first 5 answered calls are free on a 7-day trial — it takes structured messages only and does not quote, book, or dispatch. The full platform is Core at $500 per month for 500 AI minutes, Pro at $750 per month for 1,000 minutes, and Elite at $1,200 per month for 2,500 minutes, each with a 14-day free trial and no per-seat fees. All tiers are listed at https://www.thekeybot.com/pricing.

Can an AI receptionist decide which jobs to send to a partner shop?

It can apply the rules you configure, which is the practical version of that question. You define the conditions — this ZIP plus this service routes to a partner, this service we never perform, this area with a non-urgent job goes on our own schedule — and the software applies them identically on every call including the fourth one arriving in the same minute. What it will not do is exercise judgment you have not written down, and it does not perform locksmith work or vouch for a partner's quality; those remain entirely yours.