Google Forms is the default answer whenever a survey is needed by Friday, and for good reason. It costs nothing, everybody already has an account that opens it, and the first version of a five question survey takes about four minutes to build. Nothing in this article argues against that.
What follows is the part that gets skipped: the handful of settings that decide whether the answers are usable, and an honest account of what happens after the responses land. The building is easy. The deciding is where surveys go wrong.
Starting the form, and what a form actually is
Go to forms.google.com and click Blank form, then name the untitled form. That name is what respondents see at the top and what the file is called in Drive, so it is worth writing as a sentence a respondent would understand rather than an internal label.
It helps to be precise about what is being built. A Google Form is a collection tool with a reporting view attached. It gathers answers, stores them, and shows them back as counts and charts. Everything it does well sits inside that description, and everything teams struggle with sits just outside it.
Building the questions
To the right of the question title is the question type selector. Change it first, before writing the question, because the type determines what the answers will look like in a spreadsheet later. Free text produces sentences that have to be read one by one. Multiple choice produces something countable. A survey that is mostly free text is not a survey, it is a pile of reading, and the decision was made at the point the question type was chosen.
Each question can carry a description, an image, or a YouTube video. Descriptions earn their place when a question contains a term that means different things to different people. Define it there rather than hoping everyone guesses the same way.
Questions can be dragged into a different order, and Required can be turned on to prevent people submitting without answering. Required is worth using carefully. Every required question raises the chance that somebody abandons the form part way, and an abandoned form produces nothing at all. Make required the questions without which the response is worthless, and leave the rest optional.
Sections are the structural tool, and each section starts on a new page. Two things follow from that. Long surveys feel dramatically shorter when split into sections of three or four questions, because the respondent sees a short screen rather than a wall. And sections are how a survey asks different things of different people, by routing a respondent onward based on what they answered, so nobody is asked to rate a service they just said they had not used.
Plan the sections before building them. Screening questions first, branches in the middle, anything everybody answers at the end.
When the answers have to be scored
A form can be turned into a quiz, which changes what the responses are for. Instead of a distribution of opinions, each submission produces a score against an answer key. This is the mode for tests, compliance checks, onboarding knowledge checks, and any screening step where a pass mark exists.
Two settings decide how it behaves. Email addresses need collecting, which is done from the Responses section by turning on Collect email addresses, because a score with no name attached is not a result anybody can act on. And under Release grades there is a choice between Immediately after each submission and Later, after manual review.
That second choice is not a detail. Immediate release suits a self check where the point is for the respondent to learn the answer. Manual review suits anything where the score has a consequence, because it gives a person the chance to look at the borderline cases, at the free text answer that was right but worded oddly, and at the respondent who clearly misread the question. Released immediately, those cases become complaints. Held for review, they become a five minute correction nobody ever hears about.
Making it look like it came from the organisation
Themes control the colour and the header, and the font can be set separately for headers, questions, and body text. A custom photo can be used as the header image.
This matters more for response rates than it looks. A survey link arriving by email from an unfamiliar address, opening on an unbranded page, looks exactly like the phishing attempts staff are trained to report. A header with the organisation's name and colours answers the question "is this real" in the first second, which is the second where most people decide whether to continue.
Keep the type sizes readable rather than stylish. Most survey links are opened on a phone, and a header font chosen on a laptop can be unreadable there.
Who can answer, and whether you know who they are
Two settings do the heavy lifting.
Responder access can be restricted to the organisation or opened to people outside it. Pick the tightest setting that lets the intended audience in. An internal survey left open to anyone can be forwarded outside without anybody noticing, and the results afterwards give no sign it happened.
Email addresses can be collected in two different ways, and the difference is important. Verified means the address comes from the respondent's signed in Google Account. Responder input means the respondent types it themselves. Verified addresses are reliable and require a Google Account. Typed addresses work for anybody but arrive with typos, and a typo in an email address on a form that needs a reply means the reply never arrives. For anything where a follow up is planned, verified addresses are worth the constraint they impose.
One more thing belongs in this decision rather than in the analysis afterwards: whether the survey is anonymous. Collecting addresses makes a response traceable to a person, which is what allows a reply and also what makes respondents more careful about what they write. Leaving them off produces franker answers to uncomfortable questions and forfeits any possibility of following up. Both are legitimate, and the mistake is not choosing. Whichever way it goes, say so plainly in the form description, because a respondent who cannot tell whether the answers are anonymous will assume the less comfortable of the two and answer accordingly.
Then use Preview before publishing. Answer the form yourself, on a phone, following each route through the sections. Every combination of answers is effectively a different survey, and the only way to know they all reach the end is to walk them.
Reading the results
The Responses tab offers three views, and each suits a different question.
The summary view aggregates everything into counts and charts. This is the view for "what did people say overall", and for many surveys it is the entire job, finished in the browser with no export.
The question view shows every answer to one question at a time. This is the right view for open text, because reading forty answers to the same question in sequence surfaces patterns that reading forty whole responses does not.
The individual view shows one respondent's complete set of answers. This is the view that matters when the survey is about people rather than about a total.
From there, Link to Sheets at the top right puts every response in a spreadsheet, and the More menu offers Download responses (.csv) for a static file.
The distinction is worth understanding. The linked spreadsheet keeps updating as responses arrive. The CSV is a snapshot of the moment the button was pressed. Teams that export CSVs repeatedly end up with several files that disagree, and a month later nobody can say which one a decision was based on. If the sheet is going to be the working surface, link it once and work in that one sheet.
There is a second trap in the linked sheet, and it catches almost everybody. Columns added by hand to the right of the answers, the ones recording who is dealing with a response and whether it has been handled, live in a spreadsheet that the form keeps appending rows to. That works while the rows only ever arrive in order. It stops working the moment somebody sorts the sheet, filters it and edits the filtered view, or deletes a row to tidy up, because the hand written columns and the form written columns can drift apart without any warning. The answers stay correct and the notes beside them quietly stop describing the right person.
Knowing a response arrived, and answering it
From the More options menu on the Responses tab, Get email notifications for new responses sends an alert as each submission comes in. For anything time sensitive, switch it on. Without it, a form is only checked when somebody remembers to check it, and a support request sitting unread for four days is worse than no form at all.
Beyond that, Google offers a Form notifications add-on for more notification options and customised follow up emails. That covers the acknowledgement, the message confirming the answers were received.
The real reply is a different matter. That gets written in Gmail, and nothing connects the sent message back to the row it answers. Collaborators invited to a form can edit any part of it, including the responses, which means everyone sees everything and anyone can change anything. What does not exist is a field on an individual response saying which person is handling it, or a status showing whether it has been answered.
For a survey that ends in a chart, none of this matters. Nobody needs to be assigned to a chart. It starts to matter the moment each row is a person waiting, which is why some form tools attach an owner and a status to every response rather than to the form as a whole.
Where a survey stops being a survey
Sort the forms in use into two piles.
In the first pile are the forms that end in a number. Satisfaction sweeps, event feedback, internal polls, pulse checks. These produce an aggregate and the aggregate is the deliverable. Google Forms is already the correct tool for them and there is nothing worth changing.
In the second pile are the forms that end in a conversation. Applications, enquiries, registrations needing confirmation, interest forms, support requests. These look like surveys and are not. Each row is a person expecting a reply, and the work is not the collecting, it is the tracking of who has been answered and by whom. Recruitment, course sign ups, enquiry handling and event registration all sit here, and the shape they take is fairly consistent from one organisation to the next.
The forms in the second pile are usually a minority and usually generate most of the manual work, because the spreadsheet that tracks the follow up has to be maintained by a person, and that person is doing unpaid admin that grows with volume.
What to change first
Turn on email notifications for any form where somebody is waiting for a reply, and switch email collection to verified rather than typed on anything that needs a follow up. Then look at the forms in the second pile and decide whether the tracking column in your spreadsheet is still holding, or whether those responses belong somewhere that keeps the submission and the reply on one screen.
Q1. Is Google Forms free, and does it limit the number of responses?
It comes with a Google Account at no additional cost, and there is no published per form response allowance to buy more of. The practical constraint is not the count of responses but what happens afterwards, since the tool gives you charts and a spreadsheet rather than any way to track which responses have been dealt with.
Q2. How do you ask different people different questions in the same form?
Use sections. Each section starts on a new page, and a respondent can be routed to a particular section depending on how they answered, so people only see the questions relevant to them. Build the section order first, on paper, so that every route runs forward and reaches the end.
Q3. What is the difference between verified and responder input email addresses?
Verified takes the address from the respondent's signed in Google Account, so it is accurate but requires an account. Responder input means the respondent types the address, which works for anyone but arrives with occasional typos. For any form that needs a reply sent back, verified is the safer choice.
Q4. Should responses be linked to Sheets or downloaded as a CSV?
Link to Sheets when the form is still collecting, because that spreadsheet keeps updating as responses arrive. Download the CSV when you need a fixed record of a particular moment, such as the numbers quoted in a report. Downloading repeatedly produces several files that disagree with each other.
Q5. Can two people share the work of replying to responses?
Both can see and change everything, since a collaborator on a form can edit any part of it including the responses. What is missing is a way to divide the responses between them, because an individual response has no owner field and no status. Teams usually add those as columns in the linked spreadsheet and maintain them by hand.
Q6. How do you find out that a new response has arrived?
On the Responses tab, open the More options menu and select Get email notifications for new responses. A Form notifications add-on is also available for more notification options and customised follow up emails, which covers the acknowledgement sent back to the respondent.