Use synthetic data in a product demo to show a real workflow without exposing customer information. The data can be invented, but the product behavior must be represented accurately. Tell buyers that the records and metrics are fictional, keep those records consistent across screens, and separate an illustrative outcome from a result achieved by a customer.
That distinction matters most when a demo looks convincing. A populated dashboard can help someone understand a SaaS product quickly, yet the same dashboard can imply that its figures are real, typical, or guaranteed. At DemoGo, we recommend treating every demo as three separate claims: what the software does, what the sample data depicts, and what a buyer might achieve in their own environment. Only the first can be demonstrated by clicking through a product tour.
The practical approach is to start with one buyer task, create a small fictional dataset that supports it, label the data at the point of use, and review the finished experience as a buyer would. This works for live presentations, shared screenshots, recorded videos, and interactive walkthroughs. It is particularly useful when prospects need to see a realistic scenario before they are ready to connect their own systems or create a trial account.
What synthetic data means in a product demo
Synthetic demo data is information created for a demonstration rather than taken from an actual customer account. It might consist of fictional organizations, users, transactions, projects, support requests, activity records, or analytics values. The purpose is to make the workflow understandable without publishing private information.
A fictional account name alone does not make a dataset useful. Imagine a customer success application displaying an account with declining usage. To explain the workflow, its usage chart, account details, support requests, assigned owner, and recommended follow-up must tell the same story. If the chart falls while every activity record shows increasing engagement, the viewer is left to guess which screen to trust.
Nor does synthetic automatically mean private. A file assembled from a customer export with only names changed can still contain recognizable text, unusual amounts, dates, or combinations of attributes. For a public demo, create records from a fictional scenario and review every field that appears on screen. Keep demo identities separate from real customer identities throughout the process.
The honesty test: What is this screen asking the buyer to believe?
A useful way to review any synthetic-data demo is to ask what inference a reasonable buyer will draw from each screen. The question is not just whether a label appears somewhere. It is whether the design, narration, and numbers together create an accurate impression.
| What the demo shows | What buyers may infer | How to keep it accurate |
|---|---|---|
| A chart built from fictional activity | These are actual customer results or typical results | Identify the figures as illustrative beside the chart, and explain the calculation if it matters to evaluation. |
| A guided sequence ending in a completed task | The product completed every step automatically | Distinguish actions performed by the user from actions performed by the software. |
| An integration settings screen | The external system is connected in the demo | State whether the screen shows a connected workflow, sample configuration, or a conceptual illustration. |
| A polished screen for a future capability | The feature is available to purchase and use now | Label it as conceptual or planned rather than presenting it as current functionality. |
| A large financial figure | The vendor has evidence for the implied return | Call the figure a sample input or illustrative calculation, not a customer outcome. |
This is why an accurate disclosure must travel with the content. A buyer might see a screenshot in an email without hearing the presenter explain it. Another might enter an interactive tour halfway through. Put the explanation near the figure or screen it qualifies, not solely in a legal footer or an opening frame.
Build the demo backward from the buyer’s question
The fastest way to make synthetic data misleading is to generate an impressive database first and invent a story afterward. Begin instead with the decision the buyer is trying to make. A product manager might ask whether a workflow supports the team’s process. A marketing executive may want to understand how a feature is experienced. A customer success manager may need to know what an agent sees before contacting an account.
Write the question plainly: Can a customer success manager identify an account that needs attention and assign the next action? Then list the screens needed to answer it. In this hypothetical example, those might be an account overview, an activity trend, related support requests, and a task creation screen. Each record should contribute to the answer. Decorative numbers that cannot be explained add risk without adding proof.
We use a simple planning sequence: buyer question, product action, required records, accurate disclosure. It keeps teams from confusing an attractive data story with evidence that a feature works. For more on planning the path before building screens, see our guide to storyboards and success criteria for product walkthroughs.

A step-by-step framework for creating credible synthetic demo data
1. Define one scenario and its boundaries
Choose a situation the product can actually address. Describe the starting condition, the action a user takes, and the state the viewer should understand at the end. Record what is outside the demo’s scope as well. A walkthrough of finding an account that needs attention, for instance, does not also prove how quickly a buyer can implement the software.
Decide whether the experience is a captured product walkthrough, a live demo environment, or a design concept. Those formats can look similar to a buyer but make different promises. A captured walkthrough illustrates a particular path. A live environment allows the buyer to test behavior beyond that path. A design concept explains an idea that may not be available yet. Name the format when the difference could affect a purchase decision.
2. Create fictional identities, not disguised customers
Make people and organizations clearly fictional within your team’s demo materials. Use consistent names and roles, then review them for unintended matches with real clients. Avoid copying customer notes, messages, addresses, or free-text fields, even if you intend to replace names later. Free text often carries more identifying detail than a tidy account-name field.
Be especially careful with email addresses, links, phone numbers, payment information, credentials, and API keys. A buyer should not be able to contact a person or service merely because a demo value looks plausible. Use controlled test values appropriate to your environment, and check where clickable fields lead before sharing the tour.
3. Map the records that must agree
Write a small data map before filling screens. In a customer success scenario, an account may have an owner, users, activity events, support requests, and follow-up tasks. If a task refers to an account, that account should appear in the account list. If a chart summarizes activity, the displayed period and underlying events should be compatible. If a request is marked resolved, the related narrative should not say it is still awaiting a response.
Check arithmetic as well as relationships. A dashboard total should reconcile with the detail view buyers can inspect. An apparent increase must reflect the dates shown on the chart. Currency, date formats, status labels, and permissions should remain consistent. The goal is not to simulate every production edge case. It is to remove contradictions that undermine the workflow being demonstrated.
4. Include ordinary friction, not an implausibly perfect world
A demo where every account is healthy and every action succeeds immediately leaves little room to show how a product helps. One overdue task or unresolved request can make a scenario easier to understand. Choose exceptions because they illuminate a decision, not because they make the dashboard look busy.
Likewise, do not manufacture a dramatic turnaround solely to create a satisfying ending. If the product helps a user decide what to do next, end with that decision. A follow-up task created in the demo is a visible product action. A renewed customer or recovered revenue is an outcome that requires evidence beyond the demo.
5. Classify what each step proves
Before writing narration, mark every step as one of three types:
- Current product behavior: A feature or action the viewer can evaluate in the product under the relevant conditions.
- Illustrative data: Fictional records, sample metrics, or example inputs used to give the action context.
- Conditional or conceptual material: A configuration dependent on setup, or an idea that is not a currently available capability.
Do not let the categories blur in the copy. If an action requires configuration or a particular plan, explain the condition where the action appears. If the screen represents an idea rather than available software, make that clear before the buyer reaches it. This matters especially when showing integrations, permissions, automation, or features ahead of a release. Our guide to handling pre-release features in demo content covers the related review decisions.
6. Put disclosures at the point of interpretation
Write short notices in the language a buyer needs. For example: Fictional account and activity shown for illustration. Near a chart, add: Sample metrics, not customer results. If an integration is not connected in the scenario, say so beside that step. These are examples of disclosure copy, not a substitute for explaining a meaningful product limitation.
Keep the disclosure visible when a screen is reused outside its original tour. Sales teams often turn one demo screen into a slide, email image, or recap. If removing the opening step also removes the explanation, the disclosure is too fragile. Review the standalone version before distribution.
7. Conduct a buyer-interpretation review
Ask someone who did not build the experience to explore it without a presenter. Then ask what they believe is real, fictional, available now, or dependent on setup. Ask whether any number looked like an actual customer result and whether any step seemed automatic when it was performed by a user.
Listen for the reviewer’s own words instead of prompting them toward the right answer. If their interpretation differs from your intended claim, revise the screen, not just the presenter notes. A disclosure works when the audience understands it, not when the author remembers writing it. A broader interactive walkthrough QA checklist can help with the rest of the publishing review.
How synthetic data changes across demo formats
The same fictional dataset can support several forms of product education, but each format creates a different risk. Choose the format for the question the buyer is asking, then keep the data disclosure attached to the experience.
Self-guided interactive walkthrough
A guided tour is useful when you want to show a specific click path without inviting viewers into an unrestricted environment. The viewer can examine each step at their own pace. The limitation is that a controlled tour proves the illustrated path, not every action a buyer might want to test. Say what the tour shows, and provide a path to a deeper evaluation when needed.
DemoGo is a practical choice for this format. Our desktop demo builder lets SaaS teams capture pages, add steps, customize the walkthrough, and share it without a browser plugin. Teams can self-host their interactive tours on their own website. That publishing control is useful for managing where a buyer sees the disclosure, but it does not replace checking the captured content for private data.
Live demo environment
A live environment lets a presenter respond to questions and demonstrate behavior beyond a predetermined path. It also requires careful preparation. Search results, notifications, connected services, and records outside the planned scenario may appear unexpectedly. Populate a dedicated demo environment, inspect what every role can reach, and rehearse likely detours. If a workflow needs configuration that is already complete in the demo, say so.
Recorded video or static screenshot
A recording is easy to distribute and replay, but the viewer cannot test the workflow. A screenshot is even narrower: it shows a state, not how the product reached that state. Use visible labels on the asset itself. Inspect the full recording for background tabs, filenames, messages, and notifications that could reveal information unrelated to the fictional dataset.
These formats are not substitutes for one another. An interactive tour can make a specific workflow easier to explore; a live evaluation can test variation; a screenshot can orient a reader quickly. The honest choice is the one whose limits match the claim you are making.

Best practices for metrics, AI output, and personalized scenarios
Separate a calculation from evidence of results
A dashboard can accurately demonstrate that software calculates a metric while the underlying values are entirely fictional. Tell buyers both things. If you show a sample return calculation, identify the assumptions and inputs. Do not use the sample output as a benchmark or imply that a buyer will obtain it after purchase.
The same discipline applies to charts that imply causation. A downward trend followed by an upward trend may help explain how a team would investigate and respond. It does not establish that using the product caused the improvement. If you cannot support a causal claim, describe the visible action rather than promising an outcome.
Label generated content according to what it demonstrates
For an AI product, distinguish synthetic inputs, example outputs, and actual software behavior. A carefully edited response may illustrate an ideal format but should not be framed as an unmodified result of a repeatable process unless that is what the buyer can test. If a human selected, corrected, or arranged the output, say what happened when it affects the evaluation.
Keep sample prompts and inputs close to the result they produced. A polished answer without its source material can leave buyers with an inaccurate sense of what the system was given to work with. Buyers evaluating an AI workflow often need to understand the review or approval step as much as the generated text.
Personalize the context without inventing a customer endorsement
It is reasonable to adapt fictional records to an industry or role so a buyer can recognize a relevant workflow. It is not reasonable to imply that a named prospect, their customers, or their team already use your product. For example, imagine adapting task names for a retail buyer. Keep the organization fictional, identify the scenario as illustrative, and avoid importing facts about the buyer that they have not chosen to share.
Personalization should change the explanation, not conceal functional gaps. If an industry-specific workflow requires a configuration that is absent from the demonstration, disclose the requirement. For a fuller approach, read our guide to adapting demos for different industries without faking features.
Keep privacy and publishing control in the same review
Privacy work does not end when the dataset is fictional. Check what was captured around the product screen: account menus, support conversations, background notifications, document titles, links, and embedded media. Review both the source environment and the published asset. A safe source page can become an unsafe recording if an unrelated notification appears during capture.
Decide who can approve the data, product claims, and publication location. Product can confirm that the depicted capability is accurate. Marketing can check the buyer-facing message. A privacy or security reviewer can inspect exposure risks. An owner should be responsible for updating or removing a demo when the interface or feature availability changes. Our demo review workflow guide goes deeper on coordinating those checks.
Self-hosting gives a team control over where a tour is published, not automatic control over everything shown inside it. Review access, page context, and removal procedures as part of your own publishing process. The question is not merely, “Is the data fake?” It is also, “Who can see this asset, and what might they reasonably believe after seeing it?”
A pre-publication checklist for synthetic-data demos
- Scenario: The tour answers a defined buyer question rather than presenting unrelated screens.
- Identity: Fictional people and organizations do not copy real customer records or contact details.
- Consistency: Relationships, dates, statuses, and visible totals agree across screens.
- Capability: Current, setup-dependent, and conceptual steps are distinguished where relevant.
- Metrics: Sample numbers are identified as illustrative, not customer outcomes or expected results.
- Disclosure: Labels remain understandable in the full tour and in any reused screenshot or recording.
- Capture: Notifications, background content, filenames, links, and credentials have been inspected.
- Interpretation: A reviewer unfamiliar with the project understands what is fictional and what the software actually does.
- Ownership: Someone is responsible for revisiting the asset when the product changes.
Run this check again after a meaningful interface or workflow change. A tour can become misleading without anyone editing its copy: a once-current feature may move, a required setup step may change, or an old screen may imply availability that no longer holds.
Frequently asked questions
Is synthetic data the same as anonymized customer data?
No. Synthetic demo data is created for the scenario. Anonymized data starts with real records and is transformed to reduce identification risk. For a public-facing demo, changing names in a customer export is not a reliable substitute for building a fictional dataset, because other details may still point back to the source.
Should every screen say the data is fictional?
Not necessarily. The buyer should encounter a clear explanation at the start, and a screen containing a consequential metric or claim should carry the context needed to interpret it. If a screen will be shared separately, its disclosure must stand alone. Repetition is useful when it prevents a false inference, not when it crowds out the workflow.
Can a demo show a projected return or savings figure?
Yes, as an illustrative calculation with visible assumptions and a clear explanation that it is not an actual customer result or a guaranteed outcome. If the buyer needs a forecast, discuss the inputs from their own environment. A fabricated figure cannot establish what their organization will achieve.
Can we use synthetic data for an integration demo?
Yes, but distinguish the sample records from the state of the integration. Show whether the external system is actually connected in the demonstration, whether configuration is required, and which steps a buyer would need to validate in their own environment. Fictional data should not imply that an unsupported connection exists.
What if buyers want to explore beyond the guided path?
Tell them what the guided experience covers and arrange an evaluation suited to their questions. A controlled walkthrough is valuable for understanding a workflow, but it is not a live account. The difference is especially important when buyers need to test permissions, unusual data, or integration behavior.
Does self-hosting make a synthetic-data demo safe by itself?
No. Hosting control addresses where the tour is published. It does not remove sensitive information accidentally captured in a screen or fix an inaccurate claim. Review the content, access, and buyer interpretation separately.
Make the fictional scenario useful, and the product claim precise
The strongest synthetic-data demos do not try to pass invented records off as evidence. They use a believable scenario to make real product behavior easy to inspect. Give each number a purpose, disclose what it represents, and stop the story at the point the software can actually demonstrate. If a buyer wants proof of implementation results, that is a different conversation from exploring a sample workflow.
For a self-guided version of that approach, DemoGo lets your team build step-by-step SaaS walkthroughs on desktop, without a browser plugin, and self-host the finished tour. You can start with the freemium version, use a carefully reviewed fictional dataset, and give buyers a clear path from the example they explored to a conversation about their own requirements.