Getting started

Do I need a statistics or machine learning background to use PeakyPhi?

No. You describe your process the way you already know it — what you can control, what you measure, and what a good result looks like. The math runs in the background. You see recommendations, confidence levels, and tradeoffs in plain language.

How many experiments do I need before I get useful recommendations?

Typically the first 8–15 experiments are used to map the process space. During this initial design phase, PeakyPhi suggests experiments that efficiently explore the process space before switching to targeted recommendations.

What if I have no historical data at all?

That’s fine. Start with the design phase — PeakyPhi will suggest the first 10–15 experiments to run. You enter the results as they come in and the model builds from there.

Can I import data I already have?

Yes. You can paste or upload existing experiment results as a CSV at any point — during initial setup or later. Each row should contain the input settings and the measured outcomes. PeakyPhi uses it immediately to update the model.

How it works

How does PeakyPhi decide what experiment to suggest next?

PeakyPhi builds a predictive model from your experimental results and recommends the next experiment by balancing expected improvement with remaining uncertainty. The goal is to get the most useful information from every run.

What is the design phase?

Before targeted recommendations can work effectively, the model needs a basic map of the process space. The design phase covers the first set of experiments, selected to explore the useful factor ranges efficiently rather than chase a specific optimum too early. Targeted recommendations begin once that foundation is in place.

Can I optimize for more than one goal at the same time?

Yes. You can define multiple response variables and set independent goals for each — maximize yield, minimize cost, keep impurity below a limit, or hit a specific target value. PeakyPhi balances all objectives simultaneously and shows you the tradeoffs in each recommendation.

What is robustness analysis?

A suggested setting might look optimal in theory but fall apart under real process variation — small fluctuations in temperature, feed ratio, or pH. Robustness analysis stress-tests promising settings with thousands of simulations to identify operating windows that stay stable under typical variation. The goal is a process window you can actually run, not just a point that looks good on paper.

What does model confidence mean?

It reflects how certain the model is about its prediction at the suggested settings. High confidence means the model has seen experiments nearby and the prediction is more reliable. Lower confidence means the region is less explored — the recommendation may be higher-risk, higher-reward.

What happens if an experiment fails or gives an outlier result?

Add it anyway. Failed runs and outliers still carry information — they tell the model which regions don’t work and update the uncertainty estimates. PeakyPhi is built for noisy, real-world data.

Dynamic DOE vs classical DOE

Is PeakyPhi a replacement for DOE?

PeakyPhi builds on the same goal as DOE: learning from experiments efficiently. The difference is that PeakyPhi adapts the next recommendation after each new result instead of relying only on a fixed experiment plan.

How is PeakyPhi different from classical DOE?

Classical DOE usually defines the experiment plan upfront. PeakyPhi starts with an initial exploration phase, then dynamically recommends the next run based on what has already been learned.

Why can this reduce the number of experiments?

Because later experiments are targeted toward promising or uncertain regions instead of following a fixed plan that may include low-value runs.

When should classical DOE still be used?

Classical DOE is still useful for screening, validation, regulated workflows, and situations where a fixed experimental structure is required. PeakyPhi is most useful when experiments are expensive and the team wants adaptive guidance.

Factors, responses, and constraints

What counts as a factor?

Any controllable input your team can set before running an experiment: temperature, pressure, residence time, pH, catalyst loading, feed ratio, machine speed, solvent concentration, and so on. If you can change it and it might affect the outcome, it belongs as a factor.

What counts as a response?

Any measurable outcome you care about: yield, purity, defect rate, viscosity, strength, cycle time, energy use, cost, throughput. You need to be able to measure it after each run.

What are constraints?

Hard limits the optimizer must respect — combinations of factor settings that are unavailable, unsafe, or physically impossible. For example: total solvent fraction must not exceed 80%, or a specific temperature and pressure combination is blocked for safety reasons. Constraints are applied during suggestion generation so recommendations always land inside your feasible space.

Does PeakyPhi support mixture experiments?

Yes. You can define factors as mixture components that must sum to a fixed total (for example, 100% by weight). One component can be set as the remainder, computed automatically so the constraint is always satisfied. The optimizer samples the mixture simplex correctly and the GP excludes the remainder from its inputs to avoid collinearity.

Plans and pricing

What is included in the free plan?

The free plan lets you run a small process problem with up to 3 factors and 1 response, with 3 saved sessions. Bayesian recommendations, PDF reports, and robustness analysis are not available on the free plan. It is intended for exploring the workflow before committing to a subscription.

What does Pro include?

Pro gives you up to 5 factors, 3 responses, targeted experiment recommendations, PDF reports, and robustness analysis. It is designed for teams actively optimizing real processes.

What does Enterprise include?

Enterprise scales to 10 factors, 5 responses, unlimited saved sessions, and everything in Pro plus an admin panel for managing team access. Contact us for pricing.

Can I upgrade from free to Pro later?

Yes. You can upgrade at any time. Your existing project data and experiment history carry over automatically.

Data and security

Is my experiment data secure?

Yes. Your data is stored on private infrastructure, is not shared with other users, and is not used to train any shared model. Each project’s model is built exclusively from your own observations.

What happens to my data if I cancel?

You can export your project data at any time as a CSV. If you cancel your subscription, your data is retained for 30 days before deletion, giving you time to export.

Can multiple team members use the same account?

Multi-user access is available on the Enterprise plan. On Free and Pro, the account is intended for individual use.

Data Security & Privacy

Is my data encrypted?

Yes. All data transmitted between your browser and PeakyPhi is encrypted using HTTPS (TLS), the same standard used by online banking and e-commerce. Your project data is also encrypted at rest on our servers through infrastructure-level encryption provided by Google Cloud Platform, which hosts our service.

Who can see my project data?

Only you and the users you explicitly share a project with. Each project is access-controlled — other users on the platform cannot view or interact with your projects unless you grant them access. Our team does not access your project data unless you contact us for support and explicitly provide permission.

Is my account protected against unauthorised access?

Your account is protected by a username and password. Passwords are stored using industry-standard one-way hashing — we never store or transmit your password in plain text. For additional security, you can enable Two-Factor Authentication (2FA) in your account settings, which requires a one-time code from an authenticator app each time you log in.

What happens if someone tries to brute-force my password?

Our login system enforces rate limits — repeated failed login attempts from the same source are automatically blocked. This prevents automated password-guessing attacks.

Can I share a project with a colleague?

Yes. You can share individual projects with specific users, departments, or your entire organisation. You control who has access and can revoke access at any time from your project settings.

Where is my data stored?

Project data is stored on secure servers. Data is not transferred to or processed by third parties, except as necessary to operate the service, such as payment processing via Stripe if you choose to use online card payment.

Is PeakyPhi compliant with GDPR?

Yes. We process personal data in accordance with the General Data Protection Regulation (GDPR). You can request access to, correction of, or deletion of your personal data at any time by contacting us at privacy@peakyphi.com. For full details, please see our Privacy Policy.

What data does PeakyPhi collect?

We collect the information needed to operate your account, such as email address and username, and the project data you enter into the platform, including factor settings, experimental observations, and results. We do not sell your data to third parties or use it for advertising purposes.

What happens to my data if I close my account?

Upon account deletion, your personal information and project data are removed from our systems. If you need to export your data before closing your account, you can do so at any time using the Export function available on each project.

How do I report a security concern?

If you discover a potential security vulnerability or have a concern about your data, please contact us directly at security@peakyphi.com. We take all reports seriously and aim to respond within 48 hours.

Still have questions?

Reach us at support@peakyphi.com. We respond within one business day.