A referral is the result of three things: a relationship that earned the right to ask, a system that makes the ask easy, and a thank-you that makes the referrer glad they did it. Most agencies have the first and neither of the other two, which is why referrals arrive as luck rather than as a channel. Ask specifically rather than generally - "is there anyone in your network with the problem you had last spring?" produces a name, and "know anyone who needs us?" produces nothing, because the second asks the person to run a search on your behalf.
Referrals are the largest source of new business for most agencies and the least managed.
The pattern is familiar. Work goes well, a client mentions you to someone, an enquiry arrives, and everyone agrees referrals are their best channel. Then nothing is done to produce more of them, because it feels like something that either happens or does not.
It is not. Referrals respond to process about as reliably as anything else in an agency, and the process is short.
This guide covers why the usual ask fails, when to ask, exactly what to say, how to thank people, whether to pay for referrals, and how to build a network of referrers who are not clients.
Why the usual ask fails
"Let us know if you know anyone who could use our help."
Well-meant and almost entirely useless, for three reasons.
It asks them to run a search. The person has to scan their entire network against a vague criterion. That is cognitively expensive, so it does not happen.
It has no trigger. Nothing tells them when to think of you. Without a prompt tied to a situation, the memory never surfaces at the moment it would be useful.
It arrives at the wrong time. Usually mentioned in passing during a project or months after it ended, rather than at the point of peak goodwill.
The fix for all three is specificity. A narrow question is easier to answer than a broad one, which is counterintuitive but reliably true - "who do you know in B2B SaaS whose website is embarrassing their sales team?" makes the person picture one or two individuals immediately.
When to ask
Timing does most of the work.
After the testimonial, not before. A client who has just written a positive quote has, in their own words, made the argument for you. Asking then is a very small step. Our guide to testimonials and case studies covers collecting those at two to four weeks after delivery.
After a visible win. Results landing, a launch going well, a compliment arriving unprompted. These are the moments when goodwill is highest and specific.
At the 90-day check-in. Part of client offboarding, and it doubles as a natural, non-salesy contact point.
Never during a difficulty. Obvious, and it happens - an agency behind on a deadline asking for referrals reads as tone-deaf.
What to actually say
Four rules, and a script.
Be specific about who. Name the profile, the situation, or the problem.
Make it low-commitment. An introduction, not a recommendation. People are cautious about vouching and comfortable about connecting.
Give them something to forward. A one-line description they can paste, or a case study. Removing the drafting burden is the same principle that makes a pre-written testimonial work.
Make it easy to decline. Explicit permission to say no makes yes more likely, because it removes the social pressure that causes people to simply not reply.
Hi Sarah,
Really glad the launch went well - and thank you again for the quote.
One ask, and please ignore it if nothing comes to mind. We are looking to work with a couple more B2B software companies with the same problem you had: good product, site that the sales team apologises for.
Is there anyone in your network in that position? Happy with just an introduction - no need to recommend us, I would rather have a conversation and let them judge.
If it helps, here is a line you could forward:
"These are the team who rebuilt our site - demo requests went up about a third. Worth a conversation if you are thinking about yours."
Specific profile, low commitment, forwardable text, easy exit. AgencyAnalytics' guide to getting more referrals and Rock's breakdown of agency referral strategies both converge on the same mechanics: the ask has to be easy to act on, not merely polite.
Thanking properly
The step that determines whether someone refers twice.
Acknowledge immediately, before you know whether anything comes of it. The referrer took a small social risk on your behalf and the response should not be contingent on the outcome.
Close the loop. Tell them what happened, even when it went nowhere. "We spoke to James - not the right fit for either of us, but a genuinely useful conversation. Thank you for thinking of us." This is the single most-skipped step and the one that most affects whether a second referral ever arrives. People who refer into silence stop referring.
Thank in proportion. A note for an introduction. Something more considered for a referral that becomes a client - not a formal reward, something personal that shows you noticed.
Never make it transactional with a client. Which leads to the awkward question.
Should you pay for referrals?
Depends entirely on who is referring.
From clients: usually not. A commission changes the nature of the relationship, and it can create genuine awkwardness - the referrer now has a financial interest they may need to disclose, and in some sectors a payment would breach their own policies.
If you want to recognise a client referral, prefer something that is not cash: a donation to a cause they care about, a piece of work done free, a genuinely thoughtful gift. And ask yourself whether the relationship needs a payment to produce referrals, because if it does, that is telling you something.
From partners and other agencies: yes, and be explicit. A designer who regularly passes development work, a consultancy that does not build, a complementary agency - these are business relationships and a stated referral fee is normal, clean and expected. 10% of first-year revenue is a common structure. Write it down, including how long it applies and what happens on renewal.
The distinction is between people referring you as a favour and people referring you as part of how their business works. Payment corrupts the first and clarifies the second.
Referrers who are not clients
The half of the referral channel most agencies ignore entirely.
Complementary agencies. The strongest source available. A design studio that does not build, a development shop that does not design, a PR firm that does not do web. These businesses talk to your exact buyer, regularly, about adjacent needs.
Cultivating them is straightforward and slow: refer to them first, repeatedly, with nothing expected. Reciprocity in this space is strong and it compounds over years.
Former colleagues and employees. People who have worked with you know exactly what you are good at. An employee who leaves on good terms is a referral source for a decade, which is one more argument for handling departures well.
Suppliers and adjacent professionals. Accountants, lawyers, recruiters and consultants who serve your buyer. They are asked for recommendations constantly and usually have nobody specific to name.
The people you hand off to. Covered in our guide to project handoff: a clean handover to a competing agency is one of the more reliable ways to generate agency-to-agency referrals, because the incoming team forms a view of you within a day.
For all of these the mechanism is the same and it is not a programme. It is being specific about what you want ("we are looking for B2B SaaS companies who need a site rebuild") and being useful first, consistently, without keeping score too precisely.
Making it a system
Five things, none of them elaborate.
A trigger in the process. The referral ask should exist as a scheduled task after the testimonial request, with an owner. Everything about referrals is easy and all of it falls outside the window where anyone is naturally paying attention.
A written profile of who you want. One sentence, shared internally, so everyone asks for the same thing. Vague asks produce vague results.
Forwardable material. A one-liner and a case study, ready to send.
A record of who referred whom. So you can close the loop, thank properly, and notice who your genuine advocates are - it is usually a smaller group than expected and worth investing in specifically.
A quarterly review. How many referrals came in, from whom, what converted. Referrals only become a channel when they are measured like one.
Why specialisation multiplies this
Worth connecting explicitly.
A generalist's referrals are unpredictable, because "they do marketing stuff" could mean anything and the referrer cannot judge fit. A specialist's referrals arrive pre-qualified, because the referrer knows precisely what you do and only mentions you when it matches.
That is one of the strongest practical arguments for niching, and it operates independently of anything you do to your referral process. Narrowing what you are known for improves referral quality without any additional effort.
Why most referral programmes fail
Formal programmes - the kind with a name, a landing page and a reward structure - underperform in agency contexts more often than they succeed, and the reasons are consistent.
They add friction to something that works informally. A client who would happily mention you in conversation is less likely to fill in a form. Any step between the impulse and the introduction loses most of the volume.
They make it transactional. The moment a referral carries a stated reward, the referrer's motivation changes from goodwill to compensation, and goodwill produces better referrals.
They are built and then not operated. The programme exists; nobody asks. Which returns you to the original problem with a landing page attached.
What works instead is unglamorous: a specific ask, at a scheduled moment, from a named person, followed by closing the loop. That is a habit rather than a programme, and habits are what survive a busy quarter.
Reciprocity, and how to give referrals well
The fastest way to receive referrals is to give them, and most agencies do this badly or not at all.
Refer without keeping score. Reciprocity in professional networks works over years and it stops working the moment it becomes transactional. Someone who can feel you counting will not refer to you.
Refer specifically, the way you would want to be referred to. "You should talk to Priya at X - she does exactly this kind of migration and I've seen her work" is a referral. "You could try a few agencies" is not.
Make the introduction, do not just give the name. A double opt-in introduction - checking both sides are happy first - takes two extra messages and converts far better than passing on an email address.
Tell the person you referred what you said. So they can pick up the conversation with context rather than fielding a cold enquiry they cannot place.
Refer work you could have taken. This is the one that builds real reciprocity. Passing on something outside your niche is easy; passing on something you could have done but that someone else would do better is what people remember, and it is also a positioning statement about what you actually specialise in.
When referrals are your only channel
A caution worth stating, because referral-dominant agencies frequently discover this the hard way.
Referrals are the highest-converting channel and they have one structural weakness: you do not control the volume. A quiet quarter among your clients produces a quiet quarter for you, with no lever to pull. Agencies that have relied entirely on referrals for years often have no other functioning channel when the flow slows, and building one takes six to twelve months - which is exactly the time you do not have when the pipeline is already thin.
The mitigation is not to deprioritise referrals. It is to run one deliberate second channel alongside them, worked consistently even when referrals are strong. Our guide to how agencies get clients covers choosing one that suits your actual strengths.
The related structural point: referral flow is roughly proportional to how many clients you have finished work with recently. An agency with long retainers and few new engagements has fewer referral moments, however good the relationships are - which is worth knowing before concluding that the ask is not working.
What to do when a referral does not fit
It happens, and handling it well protects the channel.
Respond quickly and take the conversation anyway. Even when it is obviously wrong. The referrer took a social risk; going quiet on their introduction teaches them not to make another.
Decline clearly and helpfully. "This is not something we are the right people for, but here are two firms who would be" is a good outcome for everyone. The prospect gets help, the referrer looks good, and you have generated goodwill with whoever you recommended.
Tell the referrer why, briefly and without criticism of the prospect. "Not quite our area - I pointed them at someone who does more of that work" is enough. It also calibrates them for next time, which quietly improves the quality of future referrals.
Repeated poor-fit referrals are usually a positioning problem rather than a referrer problem. If people consistently send you the wrong thing, they do not know what you do - which is an argument for specialising and for being much more specific in the ask.
Timing across the client lifecycle
Referrals are usually treated as an end-of-project ask. In practice there are four distinct moments, and each has a different conversion rate.
Mid-project, after a visible win. Something landed well, the client is pleased, and the relationship is warm but not yet complete. Low conversion for a formal ask, and an excellent moment for a soft one: "if this is the kind of thing your peers struggle with, do mention us." Plants the idea without making a request.
Two to four weeks after delivery, alongside the testimonial. The highest-conversion moment by a distance, because the outcome is visible and the experience is fresh. This is where the specific ask belongs.
At ninety days, during the check-in. Second-best, and it has an advantage the earlier ask does not: by now there are results to point at, so the referrer has something concrete to say.
On renewal or repeat engagement. A client who has chosen you twice is a considerably stronger advocate than one who has chosen you once, and the second engagement is a natural moment to ask again without it feeling repetitive.
The mistake is asking once and concluding the client is not a referrer. Asking at two of these four moments, six months apart, roughly doubles the yield from the same relationship.
Measuring whether it is working
Four numbers, reviewed quarterly, turn referrals from an anecdote into a channel.
Referrals received per completed engagement. The core rate. With no process most agencies sit well below 0.3; with one, above 0.5 is achievable.
Conversion rate of referred leads. Almost always dramatically higher than any other channel - frequently three to five times - which is the argument for spending effort here rather than on outbound.
Time from ask to referral. If it is consistently long, the ask is probably too broad and the person is having to run a search rather than recall a name.
Who refers. Usually a small group produces most of them. Identifying that group and investing in those relationships specifically is more productive than a general programme, and it is invisible unless you record the source of every enquiry - which almost no agency does.
Our guide to how agencies get clients covers why measuring revenue rather than enquiry volume matters, and referrals are the channel where the distinction is starkest: they produce fewer leads and far more clients.
Making referrals easier to give
Two small assets do a disproportionate amount of work.
A one-paragraph description of what you do and who for, written to be forwarded. Most people cannot articulate what an agency does, even one they rate highly. Giving them the words removes the hardest part of the task.
One case study that travels well. Recognisable problem, clear outcome, short. This is the thing a referrer sends when someone asks "who did your site?"
Both live in your own inbox as much as on your site, because most referrals happen in a message rather than on a web page.
The system, assembled
Everything above, as a process you could implement this month.
At the wrap-up call. Mention that you will follow up about a testimonial in a couple of weeks. Signals the later ask so it arrives expected.
At two to four weeks. Testimonial request with a pre-drafted quote, sent by the person they worked with.
Immediately after the testimonial arrives. The referral ask - specific profile, low commitment, forwardable line, easy exit.
Within 48 hours of any referral. Acknowledge, whatever happens next.
Within two weeks of any referral. Close the loop. Tell them the outcome even when it went nowhere. This is the step that determines whether there is a second referral.
At ninety days. The check-in, which doubles as a second natural moment to ask.
Quarterly. Review referrals received, conversion, and who is actually referring. Invest in that small group specifically.
Six touchpoints, four of which are emails of under a hundred words. The whole system is perhaps two hours to set up as templates plus a scheduled trigger in whatever runs your projects.
What makes it work is not sophistication. It is that each ask is specific, each arrives at a moment of genuine goodwill, and every referrer finds out what happened - which is the part almost everybody skips and the part that turns a single referral into a habit.
The summary
Referrals stop being luck when three things exist: a specific ask, at the right moment, with a thank-you that closes the loop.
Ask after the testimonial. Name the profile rather than asking generally. Give them something to forward. Tell them what happened either way.
And spend as much attention on non-client referrers - complementary agencies, former colleagues, adjacent professionals - because they are a permanent source rather than a by-product of individual projects, and almost nobody is competing for their attention.