Operated function · referrals
When it goes well they look good to a person they know. When it goes badly they look careless to that same person, and no incentive you offer compensates for that — which is why a referral programme measured only on volume eventually stops producing any.
The willingness exists. Ask any customer success team and they can name customers who would happily introduce somebody, and most of those customers have never been asked — because asking is awkward, it is nobody’s scheduled work, and the person best placed to ask is the person most worried about the relationship.
When asking does happen, it is at the wrong time. Renewal season, or a quarterly push, or immediately after a purchase when nothing has been delivered yet — rather than after the thing worked, which is the only moment somebody genuinely wants to tell a colleague.
The ask is also usually too big. "Do you know anyone who might be interested" requires the customer to do the thinking, remember somebody, judge fit, and compose an introduction. Most people, wanting to help, say yes and then do nothing, because the task has no shape.
What follows an introduction is where the damage happens. The referred person is contacted late, or by somebody with no context, or dropped into a standard sequence that treats them as a cold lead — and the referrer finds out, because the referred person tells them.
And the incentive is usually the whole design. A payment per referral turns a relationship into a transaction, changes who participates, and attracts exactly the referrals that are worth least: introductions made for the payment rather than for the fit.
A referral requires a specific person, at a specific moment, to make a specific introduction — and most programmes ask everybody, at a scheduled moment, for an unspecified one, then handle the result like a lead.
Timing first. The moment to ask is immediately after something demonstrably worked for that customer, which is observable in your own systems, and which almost never coincides with a marketing calendar.
Then shape. Instead of asking who they know, a good ask names a specific kind of person, offers a written introduction the customer can send or amend in one action, and requires nothing else from them. That turns a vague willingness into a completed act, which is where nearly all the loss currently is.
Then the follow-through, which matters more than either. A referred person must be contacted quickly, by somebody who knows who introduced them and why, and never dropped into a sequence designed for a stranger. If that fails, the referrer learns that referring you costs them, and they stop — and so, eventually, does everyone they would have told.
And closing the loop is not optional. Telling the referrer what happened, even when nothing happened, is what makes a second referral possible. It is skipped almost universally because it is unglamorous and nobody owns it.
What is never delegated is the commercial relationship. The conversation with a referred prospect belongs to your salespeople, and any compensation is your decision under your own policy.
Customers actually asked — measured by asks made to eligible advocates, against a baseline where willingness was known and asking was not scheduled.
Asks that convert into a completed introduction — measured by introductions made as a share of asks, which is where a shapeless ask fails and a specific one does not.
Time from introduction to first contact — measured by elapsed hours, and the share contacted with the referrer context attached rather than as a cold lead.
Referrers told what happened — measured by loops closed as a share of introductions, including the ones that went nowhere.
Whether referrers refer again — measured by second and third introductions from the same person, which is the only real measure of whether the programme is spending or building.
Advocates protected from an ill-timed ask — measured by asks suppressed by your protection rules — an outcome, even though it produces no referral.
that referral volume can be scaled indefinitely. It cannot: the supply is bounded by real relationships and pushing past that consumes them. Nothing here contacts a referred person before your salesperson does, decides compensation, or treats an introduction as a lead to be worked by a sequence.
The trigger reads your own product and delivery data, because asking after something worked requires knowing that something worked. Where that is not observable, the trigger falls back to a human flag and the limitation is stated.
Referred people enter your CRM with the referral relationship attached, so nobody can accidentally work them as a cold lead. That attachment is the mechanism that protects the referrer, and it is more important than any part of the ask.
Protection rules read the account state — escalations, renewals, recent asks — so an advocate is not approached during a bad week by an operation that could not see it.
Protection rules come before conversion rules. An account in escalation, in a difficult renewal, or asked recently is excluded, and those exclusions are set by your customer leadership rather than tuned for volume.
The referred person is a person, not a lead record. They are contacted by a named human with the context of who introduced them, and they are never entered into an automated nurture sequence — which is the specific failure that costs the referrer.
On compensation, the position is deliberate and it reflects how this company operates: money stops at the person you actually met. Multi-level structures that pay on introductions made by people who introduced somebody else are not built here, because they change what a referral is and they attract the people who are least useful to you.
Every ask and every closed loop carries a receipt, so a referrer who says they were never told what happened is answered from a record.
Operational access is not permission to train. Your customers’ relationship graph — who knows whom — is among the most commercially sensitive things you hold and does not become material improving anything serving another organisation.
Customer leadership owns the protection rules, and they must be set before the conversion rules exist. A programme optimised for referral volume will eventually ask somebody in an escalation, and one such ask costs more than a quarter of referrals returns.
Legal will want the compensation position and any disclosure that attaches to it, which differs by jurisdiction and by sector — in some, an incentivised introduction must be disclosed to the person being introduced.
Account owners should agree the ask language, because it goes to their customers and they carry the consequence of a mistimed one.
One trailing period — read-only, nobody contacted — identifying customers whose observable results would have justified an ask, and how many were asked at all.
The first phase contacts nobody. It produces the list of moments where an ask would have been natural and was not made, which is usually most of them, and it establishes how many of your existing referrals arrived unprompted.
The second number is the useful one. A business receiving referrals with no programme has an advocacy base that a well-run operation can extend and a badly-run one can destroy, and knowing the current unprompted rate is the only way to tell afterwards which happened.
If you continue, the first delegation is the specific, well-timed ask on newly-observed results only — with protection rules live, the introduction prepared for the customer to send in one action, and the follow-through owned by a named salesperson before the first ask goes out.
Most do, and the three usual reasons are all mechanical: the ask was mistimed against a calendar rather than a result, it was open-ended enough that willing people did nothing, and the follow-through treated the referred person as a lead. The first phase shows which applied to yours by looking at what your own data says about when asks were made relative to when things worked.
Some will not, and the protection rules exist so those are excluded by rule rather than by somebody remembering. The more common finding is that customers who would gladly introduce somebody have never been asked, because the person best placed to ask is the person most protective of the relationship. If your first phase shows a genuinely low advocacy base, the honest conclusion is that referrals are not your channel.
It changes what you get. Payment attracts introductions made for the payment, which are usually the worst-fitting ones, and it converts a relationship into a transaction that some of your best advocates will decline to participate in. Whether to pay is your decision with your legal team — what is not built here is a multi-level structure, because money that keeps flowing past the person who actually made the introduction changes the programme into something else.
Correct that they cannot be scaled indefinitely — the supply is bounded by real relationships and the page says so. The argument is that most businesses are far below their bound rather than at it, because nobody is asking at the moment when asking would work. If your first phase shows you are already asking at the right moments and closing the loop, you are near the bound and this is not your constraint.