Feedback JourneyFeedback Journey

What are the best practices for collecting feedback from business clients?

Collect feedback from business clients by asking everyone who actually saw the work, at fixed points in every engagement, through a short link they answer alone, in writing, with the scale explained as they score. Everything else is detail. But the detail is where most feedback programmes quietly fail. A badly timed request that reads like a plea for praise gets you a 9 out of politeness, and nothing you can use.

Collecting feedback from a business client is harder than collecting it from a consumer, and for a structural reason. Your client is not one person. It is a sponsor who signed the contract, a project leader who ran the schedule, and a handful of people who sat next to you for six months. They saw different things. They have different incentives for telling you the truth. And one of them controls whether you get invited back.

This guide covers the collection step specifically: who to ask, when, how to word the request, what response rate to expect, and what to do the moment the answers arrive. If you are still deciding what to measure, start with how to measure customer satisfaction for consulting services. If you already collect feedback and want it to hold up later, read how to document client feedback for service improvement.

What are the best practices, in short?

Eight practices carry almost all of the result. They are unglamorous, and firms skip them in roughly this order:

  • Ask everyone who saw the work, not only the person who signed the contract.
  • Ask at least twice — at about one-third of the way through, and again at the close. On a long engagement, at least once a year.
  • Ask for constructive feedback, in those words. A request that reads as a request for praise gets praise.
  • Keep the request short. Brief invitations outperform detailed ones by a wide margin.
  • Keep the questions fixed. The same four questions every time, or nothing compares to anything.
  • Ask people by name, and let them answer alone, in writing, in their own time.
  • Explain the scale at the moment someone picks a number.
  • Close the loop. Tell the client what you changed. It is the only thing that makes the second request credible.

Who inside a client organisation should you ask?

Everyone who worked with you, and all of them with the same four questions. The sponsor saw the outcome, the project leader saw the process, and the team saw you every day. The average is only worth something if it includes all of them. Asking every co-worker also changes the conversation with the sponsor: instead of one leader's impression of you, you can show them what their own people think.

Consulting-practice advisor David A. Fields recommends surveying the project sponsor as the primary respondent, with the client's project leader and other important contributors added as appropriate (see his guidance on how and when a consulting firm should solicit feedback (opens in a new tab)). Treat that as the floor. For most engagements the invitation list is short and obvious:

  • The project sponsor — the person who bought the work and decides whether to buy it again.
  • The client-side project leader — the person who saw the plan, the promises and the status reports.
  • The people you worked beside — developers, product owners, analysts, whoever sat in the same meetings and the same code.
  • Every attendee, if the engagement was a workshop or a course.

Do not tailor the questions to any of them. A sponsor who cannot judge your collaboration inside the team answers “I don't know” on that category, and the score is left out of the average instead of guessed. That one option is what lets you send the same four questions to everyone and still get an honest picture from each person.

The question set follows the shape of the engagement, not the client and not the responder's role. A six-month embedded assignment and a two-hour workshop need different questions, and the same client may give you both in a year. Three fixed sets cover nearly all professional work: one for long assignments inside a client's team, one for workshops and courses, one for service delivered at a distance. Those are the client satisfaction surveys for consulting work, and each scores four categories on the same 0–10 scale.

How many people should you invite?

Invite everyone with first-hand exposure, and err on the side of more. On an embedded assignment that is usually between three and ten people; for a workshop it is everyone in the room. Fewer than three and one grumpy Tuesday dominates your average. The only people to leave out are those who never saw you work at all.

There is a second reason to care about the number, and it is the one buyers care about most. A score is only as credible as the count behind it. An 8.4 from eight responders is a track record; the same 8.4 from one person is a compliment. Every extra person you invite makes the result harder to dismiss, which is why the responder count belongs next to the score whenever you show it.

Give people a genuine way out, too. A category scored “I don't know” is a better data point than a guess, because it tells you the responder was not positioned to judge. Without that option, people invent a 7 and your average quietly absorbs a fiction.

When during an engagement should you ask?

Ask about one-third of the way through, again within two weeks of the close, and at least once a year on engagements that run longer. Fields recommends three touchpoints. Informal conversations with the sponsor run throughout, roughly monthly on a long project and weekly on a short intense one. On top of that sit two formal surveys: one at a third of the way in, one at the conclusion.

The mid-project round is the one almost nobody runs, and it is the one that pays. Feedback at the end tells you what you should have done. Feedback at one-third tells you what to do, while you are still there and while a correction still reads as service rather than an apology.

On an engagement that runs for years, do not let the close be the next time you ask. Add a formal round at least once a year, and every six months if the work changes shape. A year is long enough for a small irritation to harden into a reputation, and a two-minute survey is a far cheaper way to find it than a non-renewal.

When not to ask

Timing can invalidate an otherwise good survey. Three windows to avoid:

  • The week you invoice. A request for praise that arrives with a bill reads as a request for praise that arrives with a bill.
  • While an extension or renewal is being negotiated. Both sides are performing, and the scores will show it.
  • Long after the close. Two months on, people are recalling a mood rather than the work. Ask within two weeks or accept softer data.

Who should send the feedback request?

You should, through a link the client answers alone, in writing, in their own time. Fields observes that a client will rarely tell the consultant who ran the work, to their face, that the work was mediocre, and he suggests an outside interviewer for that reason. The problem is real. The fix is the wrong shape for most consultants, because what softens the criticism is the conversation, not the sender.

A written form removes the conversation without removing you. The client opens a link, scores four fixed questions with the scale explained beside each one, and writes what they want to say with nobody watching. No interviewer, no meeting, no third party on the payroll. The request should come from you, in your own words, because a personal note from the person who did the work is the one that gets answered. Save the meeting for afterwards, when there are results to go through together.

What should the invitation say?

Keep it short, personal, and specific about what happens next. This is the one part of the process where there is genuinely good experimental evidence, and it says something counterintuitive: less text works better.

A Cochrane systematic review pooled randomised trials on questionnaire response. For electronic questionnaires, a brief covering letter roughly tripled the odds of a reply against a detailed one (odds ratio 3.26). Shorter questionnaires raised response by about half (OR 1.51). Personalisation helped (OR 1.24), and so did stating a deadline (OR 1.18). Using the word “survey” in the subject line actually reduced response (OR 0.81) — see Edwards et al., Cochrane Database of Systematic Reviews, 2023 (opens in a new tab). Those trials studied research questionnaires rather than client feedback, so treat the direction as reliable and the exact sizes as indicative.

A request that follows the evidence has six parts and fits on a phone screen:

  1. A subject line that names the work, not the instrument. “Your view on the Nordic rollout” beats “Client satisfaction survey”.
  2. Their name and the engagement in the first line, so it is visibly not a mass mailing.
  3. Why you are asking — one sentence that says you want honest, constructive feedback to improve the work, not a testimonial.
  4. How long it takes. Four scored questions and an optional comment is roughly two minutes. Say so.
  5. A deadline. An actual date. “When you get a chance” is not a deadline.
  6. What happens to the answers — who sees them, that you will read every one, and what you will do with them.

Resist the urge to explain your feedback philosophy in the email. Every extra paragraph is a reason to close the tab and deal with it later, and later is where response rates go to die.

What response rate should you expect from business clients?

Expect well above half, and treat anything under that as a problem with the ask rather than the relationship. Your list is warm, small, and made of people who know your name. That is a very different situation from a cold research survey.

For a reference point, the largest analysis of survey response in organizational research examined 1,607 studies published in 2000 and 2005 across 17 refereed journals. It found an average response rate of 52.7% when data came from individuals and 35.7% when it came from organizations (Baruch & Holtom, Human Relations, 2008 (opens in a new tab)). Those are cold-contact numbers from academic research. If you invited five people you worked beside for six months and three of them ignore you, the invitation or the timing is wrong.

Record the denominator either way. An 8.4 from twelve responders out of fourteen invited is evidence. An 8.4 with no responder count is a claim, and a buyer is right to read it as one.

How do you raise a low response rate?

Change the wording, the length and the timing before you change the tool. In roughly the order worth trying:

  1. Warn people it is coming. One sentence in a status meeting or a closing call. An expected request is answered; an unexpected one is triaged.
  2. Cut the invitation in half. Then cut it again. This is the single best-evidenced lever available.
  3. Send one reminder, and only one. A week after the first, to the people who have not answered. Two reminders is pressure, and pressure buys you politeness rather than participation.
  4. Move the send. Away from invoice week, away from the renewal conversation, into the fortnight after delivery.
  5. Write it in your own words. A personal note from the person who did the work is answered. A template that reads like a system mail is triaged.
  6. Shorten the form itself. Four scored questions plus one optional comment. If your form takes ten minutes, the problem is not the email.

One thing not to do: never offer an incentive for client feedback. Incentives do raise response, and in a commercial relationship they also make every resulting score unusable. You cannot pay for an opinion and then present it as independent.

Should feedback from business clients be anonymous?

Not by default. Ask people by name, and make honesty safe instead. A named answer is one you can act on: you can thank the person, ask what a 6 meant, agree what will change, and note the follow-up against their response. An anonymous 6 is a number you can only wonder about, and wondering is not improvement.

Anonymity is often sold as the price of honesty. In a small, warm group it buys less than it promises. With three to eight responders, a written comment usually gives its author away in any case, and the honesty you actually need comes from elsewhere:

  • A private, written answer. Nobody softens a score they type alone the way they soften one they say across a table.
  • A scale explained at the moment of scoring. When a 7 is described as a good result, a 7 becomes a comfortable thing to give.
  • A request that asks for constructive feedback in so many words. People give what they are asked for. Ask for praise and you get praise.
  • A visible follow-up. The client who saw their last comment turn into a change will write the next one.

Anonymous answering has one clear home: workshops and courses, where you share one link with the room, collect scores before people leave, and were never going to follow up with thirty attendees one by one. For an engagement with a client you work beside, invite people by name, say who will see the answers, and then honour that. Keep responder contact details only as long as the engagement needs them; ours are handled under the privacy policy.

How do you keep the scores honest rather than polite?

Tell responders what each number means at the moment they choose it. Left to themselves, satisfied clients pick 9 or 10 out of courtesy, and every engagement you ever run looks identical. Guidance at the point of scoring fixes this better than any wording in the email.

Say that a 10 is world class and worth saving for a special occasion. Say that 7 to 8 is what a good engagement normally looks like. Say that 0 means “I don't know” rather than “terrible”. Say that a 5 or a 6 is a good moment to leave a note on what would have made it an 8. Scores then spread across a range you can manage against, and a 7 you can trust is worth more than a 10 you cannot. The mechanics of this are covered in more depth in the guide on measuring customer satisfaction.

What should you do once the answers are in?

Respond within a week, pick one thing to change, and tell the client what it is. Collection is only half the transaction. The client spent two minutes on your behalf, and what they get back determines whether they spend two minutes again next time.

  1. Thank everyone, quickly. Before you have analysed anything.
  2. Go through the results with the client. This is where a meeting belongs: after the answers are in, with the sponsor or the whole group, to understand a 6 and agree what will change rather than to defend the work. Fields recommends the same conversation with anyone who raised a concern. Note what was said and what you agreed against the response, so the next survey can be read against it.
  3. Read the categories, not the average. The average tells you how it went. The four numbers underneath tell you which part of your service to work on.
  4. Pick one category and set a goal with an end date. One, not four. That is what turning client feedback into an improvement plan is for.
  5. Tell the client what changed. Even a small change, named out loud. This is the step that makes the next request answerable.
  6. Ask the same questions next time. A repeated question is a before-and-after you did not have to construct.

Acting on the result is not optional politeness, it is what makes feedback work at all. The largest review of the field — a meta-analysis of 607 effect sizes across 23,663 observations — found that feedback raised performance on average, but that over a third of feedback interventions actually reduced it (Kluger & DeNisi, Psychological Bulletin, 1996 (opens in a new tab)). What separated the two was whether the feedback pointed at the task or at the person. Collect it so it points at work you can change.

Do you need consent to email a client contact for feedback?

In the EU, a satisfaction request to someone you are already working with normally rests on legitimate interests rather than consent. GDPR Article 6(1)(f) (opens in a new tab) allows processing necessary for a legitimate interest, balanced against the individual's rights. Asking a named project contact what they thought of an engagement they participated in is a fairly comfortable fit.

That is not a licence to build a mailing list. Keep it proportionate. Invite only people with a genuine connection to the work. Send one reminder at most. Say who you are and what the data is for, and honour any request to be left alone immediately. Delete the addresses once the engagement and its follow-up are closed. And if you want to reuse a recommendation in your marketing, ask for that separately, recording the person's name and title as they agreed to them.

What does good practice look like in a standard?

The international guidance says the same thing in colder language: plan the collection, do it regularly, and analyse what comes back. ISO 10004:2018, Quality management — Customer satisfaction — Guidelines for monitoring and measuring (opens in a new tab) sets out how to define and run a process for monitoring customer satisfaction. It covers planning what to gather, choosing the method, and reviewing what comes back.

For a certified firm this stops being advisory. ISO 9001:2015 (opens in a new tab) clause 9.1.2 requires the organisation to monitor how far customers feel their needs have been met. It also requires it to determine the methods for obtaining and reviewing that information. A dated, scored survey run the same way every time clears it. A folder of thank-you emails does not, because nothing in it was obtained by a method.

What are the most common mistakes?

  • Asking only the sponsor. One opinion, formed at a distance, about work they mostly did not watch.
  • Asking only at the end. You get a verdict when what you needed was a warning.
  • Asking in a meeting. Feedback given to your face is a measure of your client's manners. Feedback written alone is a measure of the work.
  • Rewriting the questions per client. Every edit resets the series, and comparability is the entire value.
  • A ten-minute form. Length is a response-rate problem long before it is a data-quality one.
  • Chasing. One reminder is a courtesy. Three is a campaign, and it distorts what you collect.
  • Silence afterwards. Feedback that visibly changes nothing trains people not to answer the next one.
  • Keeping only the good ones. A curated set of results is a marketing asset and a useless instrument, and buyers can usually tell.

How do you show a buyer what you collected?

Show a stated range of engagements with the responder count and the date attached. A buyer's three questions are always the same: how many people said this, how recently, and can I check it myself. Selection is what they are really probing, and you answer it by covering everything between a first and a last survey with nothing removed from the middle.

That is what a verifiable proof of client satisfaction does. It shows the average per category, the number of responders and the date of the most recent response. It also carries a QR code, so the person reading your proposal can check the numbers at source. Because it belongs to the person rather than the employer, client feedback for consultants stays checkable when they change jobs.

Frequently asked questions

Sources

Start collecting

Send your first client feedback request in a few minutes — fixed questions, a link your client answers in two minutes, free with no credit card.

Start free — no credit cardMore guides