Frequently Asked Questions

The questions people actually ask in their first weeks with Functionize Studio, grouped by when they tend to come up. Each answer links to the article that covers it properly.

Open in Studio
The screens these answers refer to. Links open in a new tab.

Getting started

Q What is Studio, in one sentence?
A An independent testing agent for your web application: you describe what should be true in plain language, and it builds the test, runs it in a real browser, and keeps it working as your application changes. See What Is Functionize Studio?
Q Do I need to know how to code?
A No. Everything in our exercise series is done by describing outcomes in ordinary English. Code is optional, and only for unusual cases — see Extensions: Running Your Own Code Inside a Test.
Q Do I have to install anything?
A No. Tests run on our infrastructure in a real browser. There is nothing to install or maintain.
Q What should I do first?
A Work through Your First 30 Minutes with Functionize Studio. It deliberately starts on our demo store rather than your application, so you learn what "working" looks like before you meet your own environment's quirks.
Q How long before I get value?
A One passing test in about ten minutes. A team genuinely onboarded in about two weeks — see Your First Two Weeks: An Onboarding Plan.
Q Is there a practice application I can use?
A Yes, and no sign-in is needed: https://ecom.functionizeapp.com/testsites/anaqa/index.php?route=common/home. The whole exercise series uses it.

Setup and access

Q Studio cannot reach my application. What now?
A This is the most common first-week problem and it is almost always access rather than testing. Work through Before You Point Studio at Your Own Application. If your app is not on the public internet, stop and ask us rather than losing a week.
Q My app is behind a login prompt from the browser itself.
A That is HTTP basic auth: Project Settings → Access & Auth → Basic authentication.
Q We use self-signed certificates on staging.
A Certificate handling → Allow invalid TLS certificates.
Q A cookie banner blocks every test.
A Do not add steps to dismiss it. Set the consent cookie once at project level under Access & Auth → Cookies.
Q It cannot find elements that are clearly on the screen.
A If your front end uses web components, turn on Advanced → Browser behavior → Open Shadow DOM. Otherwise raise the Missing element timeout. Both are in Timeouts, Access and Browser Behavior: Getting Studio Into Your Application.
Q Should I test against production?
A Generally no — you will create real orders and send real emails. Use a stable QA or staging environment. An unstable environment produces failures that teach your team to distrust the tool.

Writing tests

Q What makes a good prompt?
A A starting URL, a description of what a user does rather than how the page is built, and an explicit statement of what "passed" means. See Exercise 1 - Create Your First Test From a Prompt.
Q Why should I not describe CSS selectors?
A Because Studio finds elements visually and contextually. Describing intent is what lets the test survive a redesign; pinning it to a selector throws that away.
Q My test passes but I am not sure it proves anything.
A It probably does not. A test with no assertion cannot fail for the reason you care about. See Exercise 2 - Write Assertions That Prove Something.
Q How long should a test be?
A One test, one question. If a red result would leave you unsure which part failed, it is too long. See Choosing What to Automate First.
Q The agent asked me a question instead of building. Is something wrong?
A No — that is the product working. It flags contradictions rather than building something destined to fail. Read it; it is usually right.
Q Can I describe a flow with a screenshot instead of words?
A Yes. The + button on the prompt box attaches files, and for a fiddly screen a picture is often clearer than a paragraph. See Starting a Session: The Studio Home Screen.
Q How do I change a test after it is built?
A Ask, in plain language. See Exercise 4 - Change a Test by Asking.

Running tests and reading results

Q Where do tests actually run?
A On a clean runtime host in our cloud, in a real browser. You can close the tab; the run continues.
Q What do the numbers on the steps mean?
A Whole numbers are the instructions you wrote; decimals are the actions Studio generated to carry them out. See Understanding a Test Run and Its Results in Functionize Studio.
Q How do I run several tests together?
A Group them into an orchestration. See Exercise 6 - Group Tests Into an Orchestration.
Q Should I put my suite on a schedule straight away?
A No. Decide who reads the result and what they do when it is red, first. A scheduled suite nobody reads manufactures confidence.
Q What is the difference between a test, an orchestration and a workflow?
A A test proves something and passes or fails. An orchestration is a group of tests run as one job. A workflow is a scheduled job written in plain language that produces a result rather than a verdict. See Workflows: Scheduled Jobs Written in Plain Language.
Q Can I see who changed a test, and when?
A Yes — the history icon in the test toolbar shows every run and every change with its author. See History: Every Change and Every Run, Recorded.

When tests fail

Q A test went red. Where do I start?
A Which step is red, and did the ones before it pass? Then read the Execution Error — it names the comparison, the expected value and the value found. See Diagnosing a Failed Test in Functionize Studio.
Q How do I tell a broken test from a broken application?
A The Data tab. If Current differs from Previous Successful Run, your application changed. If they match and it still failed, the expectation moved.
Q I clicked Fix Test and it refused to fix it. Is that a bug?
A No, and it is important. Studio repairs drift — a test looking in the wrong place. It will not silently rewrite an assertion to match a page that genuinely disagrees with it, because that would delete the failure you wanted to catch. It stops and asks instead.
Q My tests fail intermittently.
A Usually timing or environment rather than the product. Raise the Missing element timeout, or move the Timing model toward Conservative. See Self-Healing and Timing: Controlling How Studio Recovers. Do not tolerate flake — it is how suites stop being read.
Q What should I include when I raise a ticket?
A The test ID, the browser and version from the run header, and the time of the run. Those three let us find the exact execution instead of asking you to reproduce it.

Test data and secrets

Q Where do I put passwords?
A Never in a prompt or a step. Put them in an execution preset variable with Encrypt value switched on. See Test Data: Execution Presets, Variables and Secrets.
Q How do I run the same tests against different environments?
A Execution presets. Define a variable set per environment and choose one at run time, rather than duplicating the suite.
Q Can a later step use a value an earlier step captured?
A Yes — an order number or reference. Check it resolved in the Variables tab.
Q What if my test needs a generated value, like a one-time passcode?
A That is what extensions are for — your own code in Node, Python, Go or Java. See Extensions: Running Your Own Code Inside a Test.

Credits and plans

Q What uses credits?
A Agent work: creating, running, and diagnosing or fixing tests. Manual edits, dashboards, reports, exports, integrations and admin are free. The rule: if an agent is doing the work it uses credits; if you are, it does not. See How Credits Work.
Q Why did our trial burn credits so fast?
A Because building is the heavy part and a trial is mostly building. One new test costs roughly what fifty runs cost. Consumption is front-loaded and settles once your suite exists.
Q What happens if we run out?
A You are never locked out. Your tests, results, data, manual editing, dashboards, exports and admin keep working. Only new agent work pauses until credits are added, you upgrade, or the next cycle begins.
Q Do unused credits roll over?
A On self-serve monthly plans they reset each cycle. On Enterprise they roll over for the term.
Q How do we reduce consumption?
A Fix rather than rebuild, be specific up front so there are fewer regenerations, make small edits by hand (free), reuse flows as components, and prune stale tests.
Q Will costs rise as we scale?
A Functionize runs its own purpose-built models, so costs are not tied to third-party token pricing. Users are unlimited on Enterprise, so adding people does not add cost.

Skills and integrations

Q What is a skill and do I need one?
A A short document teaching the agent about your application. You do not need one to start, and it is the biggest single lever on quality once you are going. See Why Skills Matter.
Q Do you have one for Salesforce / SAP / Workday / ServiceNow?
A Yes, and for Microsoft 365 and marketing sites. See Functionize Skills: What Is Available and How to Install Them.
Q Our application is bespoke. Can we still have one?
A Yes. Write it yourself with How to Build a Skill in Claude for Your System Under Test, or ask us to build it — that is complimentary.
Q Can I drive Studio from my editor?
A Yes, over MCP, from Claude Code, Claude Desktop, Cursor, VS Code with Copilot, Gemini CLI, Codex CLI and others. Start with Getting Started with the Functionize MCP Server.
Q We already have a Selenium or Playwright suite.
A Use the migration skills to bring it across in reviewable batches — and treat it as a chance to delete the tests that were never earning their keep.

Strategy and planning

Q Do we need a written test strategy, or can we just start building tests?
A Start building — one passing test teaches you more than a week of planning. But write the strategy before you have forty tests, because that is the point at which nobody can remember why any particular test exists. A strategy that fits on a few pages is enough.
Q Do you have templates we can use?
A Yes, and they are free to download, edit and put your own name on. Four documents in both Word and PDF: Master Test Strategy Template, Integration Test Strategy Template, Regression Test Strategy Template and User Acceptance Test Strategy Template. Start at Test Strategy Templates You Can Use Today.
Q Which template should I fill in first?
A The Master Test Strategy Template, and only sections 2, 4 and 5 — what you test today, your critical journeys, and what you automate. That is an afternoon, and it is most of the value. The rest can follow.
Q What is the single most important section in any of them?
A Section 2 of the master template: an honest baseline of what you test manually today, with the date on it. Six months later it is the only evidence you have of where you started, and it is the section teams are most tempted to skip.
Q How do I decide what to automate?
A Score each candidate on six questions: how often it runs, whether the outcome is deterministic, what it costs if it breaks silently, whether the behavior is stable, whether you control the data, and whether it runs independently. See What Makes a Good Automation Candidate, which has the scoring bands.
Q How many critical journeys should we have?
A Usually ten to twenty. If your list has sixty, you have listed features rather than journeys. The test is: if this broke at 2am on a Friday and nobody noticed until Monday, what would it have cost?
Q Should we automate everything?
A No, and saying so explicitly is a mark of a mature program. Exploratory testing, subjective judgment, anything needing a physical action, one-time migration checks, and behavior still being designed all stay manual on purpose.
Q We are migrating from SAP ECC to S/4HANA. Do you have anything for that?
A Yes — a separate set covering TUT, FUT, SIT and UAT, written for how those programs are actually organized and for the systems-integrator split. Start at Testing an SAP S/4HANA Transformation. It is also honest about which SAP screens a browser can drive: Fiori and SAP GUI for HTML, but not SAP GUI for Windows.
Q How do we roll this out across several teams?
A Sequence them rather than launching them all at once, and agree five conventions everyone follows: project boundaries, test naming, tier tags, the promotion rule, and who reads red. See Scaling From One Team to Many.

Best practices

Q What is the biggest single thing we could get right?
A Never let the same AI session that wrote the code also write and judge the tests for it. Models rate their own output more highly than equivalent output from anyone else, and a test written from the implementation inherits the implementation's misunderstandings exactly. See The Independent Tester Principle.
Q How do we actually set that up?
A Two sessions, ideally on different models, with only the requirement crossing between them. The testing session connects to Functionize over MCP and has no repository access. See Setting Up an Adversarial Test Loop With Claude and the Functionize MCP Server.
Q What makes a prompt produce a test that lasts?
A Describe what a user does and what must then be true — never how the page is built. Name the starting URL, state the specific value that must hold, and add what must not happen. See Writing Prompts That Produce Durable Tests.
Q Do we need to review tests the agent generates?
A Yes, and it is about ninety seconds each. The one question that catches most weak tests: if the thing this test protects were broken, would this test go red? See Reviewing AI-Generated Tests.
Q How should we organize the suite?
A Three tiers — smoke, critical, regression — scheduled differently, with a named reader for each. Nothing enters a tier that gates a release until it has passed three times in a row. See Keeping a Suite Green.
Q A test fails intermittently. Can we just rerun it?
A No. Flake is a defect, not weather, and there are exactly three options: fix the cause, demote it out of any tier that gates something with a date, or delete it. Quarantine with no date is how flake becomes permanent.
Q Who should own a red result?
A A named person per scheduled suite, with a timescale and a first action. This single decision predicts whether the suite is still being read in six months better than anything else.

Measuring quality in an AI-DLC

Q What is an AI-DLC, and is it a standard?
A The AI-driven development lifecycle: an AI agent drives the work across the lifecycle while people validate and decide, rather than authoring everything themselves. It is a useful frame and it is not a standard — it is a vendor coinage from 2025, its own sources describe either three phases or five, and analysts use a different label for the same shift (AI-native software engineering). Adopt it deliberately rather than comply with it. See AI Testing and AI-DLC: A Glossary.
Q Why does measurement matter more in an AI-DLC than it did before?
A Because the thing that got faster and the thing that got riskier are not the same thing. Three independent results say so. DORA found AI adoption has a positive relationship with throughput and a negative one with delivery stability. A randomized trial found experienced developers took 19% longer with AI tools while believing they had been 24% faster. Analysis of 211 million changed lines found copy-pasted code rising and refactoring falling. If you measure only speed, none of that is visible.
Q Which of the five metrics moves most once agents are generating the code?
A Maintenance share and escaped defects. Writing tests is cheap now, so test count stops carrying information almost immediately. What tells you whether the program is compounding is whether the share of effort spent repairing tests is falling, and whether fewer defects reach production. Both need a dated baseline to mean anything — the definitions and targets are in What to Measure: QE Metrics and KPIs with Studio, which also has a printable tear sheet.
Q We already track DORA metrics. Does anything change?
A Two things. Add a stability measure if you only track throughput, because that is precisely where the AI-adoption penalty shows up. And check your definitions: DORA has retired MTTR, which expanded three different ways in three different standards. The current pair is failed deployment recovery time and a rework rate.
Q Self-reported productivity gains — are those worth reporting?
A Not on their own. The METR trial is the cautionary case: the developers who were 19% slower believed they were 24% faster, and they were not being careless. Feeling faster and being faster came apart. Pair any survey number with a measured one.
Q Can we measure the AI itself, or only the code it produces?
A Both, with different instruments. Code an agent wrote is tested the ordinary way — it is just code, and it needs the same independent checks. If your own product contains a model, you need evals, which is a different discipline: a test case is a distribution rather than a pass or a fail, and pass rate has to become a rate across repeated runs. See Evaluating and Testing AI: Terms.
Q Our coding agent runs the tests and reports green. Is that enough?
A No, and this is the sharpest risk in the whole model. An agent that runs your suite treats green as its stopping condition, so a weak test now actively licenses bad code rather than merely failing to catch it. Keep the thing that verifies independent of the thing that builds — see The Independent Tester Principle.
Q A vendor says they cut test maintenance by 70%. How do we check that?
A Ask for the definition, the denominator and the baseline, in writing. There is no standard definition of maintenance share, and ISTQB does not define "flaky", "self-healing", "autonomous testing" or "code coverage" either — four of the words such claims lean on hardest. The glossary flags 56 contested terms for exactly this conversation.
Q Where do we start if we have measured nothing so far?
A Half an hour, this week, before building anything: escaped defects last quarter, manual regression hours per release, and which critical journeys are covered today — with the date written on it. Section 2 of the Master Test Strategy Template is that page, already laid out. It is the section teams skip and the only evidence you will later have of where you started.

Team, scale and proving value

Q What should we measure?
A Escaped defects, critical journey coverage, time to feedback, maintenance share and flake rate. Capture a baseline in week one or you will improve things and be unable to prove it. See What to Measure: QE Metrics and KPIs with Studio.
Q How do we justify the spend month over month?
A Hours displaced, escaped defects prevented, and credit consumption with its front-loaded shape explained. Month one looks build-heavy; show the trend rather than the snapshot.
Q How many tests should we have?
A Wrong question. Ask instead: if every test passes, what do we now know is true? If that does not include "customers can do the thing we charge them for", the suite is not finished regardless of size.
Q When should we bring the rest of the team in?
A Around day six, once the environment works — so the session teaches rather than debugs. Give everyone the exercise series and its printable lab handouts.
Q Something is not covered here.
A Start at Start Here, or raise a ticket. If you are stuck on access, a skill, or a migration, ask in week one rather than week six.