Every contractor agrees this is a good idea. The electrician should send work to the plumber, the plumber should send work back, and both should do better than they would alone.
Almost none of these arrangements survive six months. Not because the idea is wrong, but because "let's send each other work" is an intention rather than an agreement, and intentions evaporate the moment everyone gets busy.
This is a different thing from asking your customers to refer you. That is a customer relationship. This is a business relationship with someone who is standing in your customer's house for their own reasons, which makes it far more valuable and considerably easier to get wrong.
Why they work when they work
The electrician is already in the basement. They can see the water heater is original and failing. They have the homeowner's trust because they are mid-job and being useful.
A recommendation in that moment is worth more than any advertising you could buy. It is warm, it is specific, it arrives with someone else's credibility attached, and it costs nothing. The customer is pre-qualified in a way no lead marketplace can match, because a tradesperson who has seen the actual problem is making the call.
That is the whole appeal, and it is real. The difficulty is entirely in the structure.
The reciprocity trap
Here is what quietly kills most of these, and it is worth naming before you start rather than discovering it in month four.
Flow is almost never equal. A roofer might see a plumbing problem twice a year. A plumber sees roof problems, or at least roof symptoms, considerably more often. One side sends five referrals, the other sends one, and nobody says anything while both keep a private tally.
Then the imbalance turns into resentment, the referrals quietly stop, and neither party ever has the conversation.
The fix is to say it out loud at the start. Some trades naturally generate more referrals than they receive, and that is fine as long as both people know it going in. If the flow will obviously be lopsided, either accept it explicitly or put money on the side that gives more than it gets.
What to actually agree
Five things, and the conversation takes twenty minutes.
Who the person is. Not a company, a person, with a mobile number. Arrangements between businesses fail. Arrangements between two people who know each other survive.
How the handoff happens. A warm introduction beats handing over a card by an enormous margin. Best is a text to both parties while still standing in the house. Worst is "you should call someone about that."
What happens when a job goes badly. Agree in advance that whoever received the referral tells the other immediately. Finding out from the customer that your recommendation went wrong is how these end.
Whether money changes hands. Covered below. Both answers work; ambiguity does not.
When you talk next. A standing check-in, even quarterly, even fifteen minutes. This is the single most predictive detail. Arrangements with a recurring conversation survive. Arrangements without one evaporate quietly and neither party notices for a year.
The money question
Two workable models.
No fee. Reciprocal goodwill, and the arrangement runs on the relationship. Simplest, no paperwork, no tax questions, no disclosure obligation. It is also the one most likely to fade, because nothing structural holds it together.
A referral fee. A flat amount per referral that converts, or a percentage of the job. This creates real obligation and tends to produce more consistent behaviour. It also introduces bookkeeping, and it introduces disclosure.
Neither is better. What matters is that it is decided and stated, because the failure mode is one party assuming a fee is coming and the other assuming it was a favour.
If money changes hands, say so
The principle here is not complicated. When someone recommends a service and has a financial interest in that recommendation, the person receiving the advice should know.
The Federal Trade Commission's endorsement guidance requires that any material connection between an endorser and a business be disclosed clearly, and treats payments, discounts, and free goods alike. There is no minimum value that exempts you. The test is whether knowing about the connection would change how the audience weighs the recommendation, and with a paid referral it plainly would.
Whether an agency would ever pursue a two-truck plumbing operation over an undisclosed referral fee is a different question from whether the obligation exists. But treat the reputational risk as the real one. A homeowner who later discovers their electrician was paid to recommend you has learned something about both of you, and that story travels in exactly the local network you were trying to build.
Some licensed trades and some states have their own rules on referral compensation, so this is worth a question to your own attorney rather than an assumption. Nothing here is legal advice.
The practical version costs nothing: "I work with Dave regularly and he takes care of me for the referral, but I'd send you to him either way." That sentence is honest, it discloses, and in practice it usually increases trust rather than reducing it.
Vet them like a customer would
You are lending your credibility. If the person you send work to does poor work, misses appointments, or is uninsured, that becomes your problem with your customer, and you will pay for it in a relationship you spent years building.
Before you send anyone a single job: confirm they are licensed and insured, look at their recent reviews, and ask how they handle a complaint. That last question tells you more than the first two.
Start with one
The instinct is to build a network of six trades at once. Do not. Start with the single adjacent trade whose work most overlaps with yours, get the arrangement working properly for a quarter, and only then add a second.
One partnership that actually functions is worth more than six that exist as a vague understanding, and it will teach you what to agree the next time.