Supercharge your predictive AI projects.

Read the Gooder AI white paper, and access this series of videos and articles.

If DIY approaches worked, then DIY would have become common practice – but it's not. Instead, the business problem persists, generally remaining unsolved.

To successfully deploy, your predictive AI team needs a system that converts ML's technical achievements into business value. This constitutes a core, fundamental step in the project lifecycle, one that bridges the gulf between model development and model deployment. It’s a step that must not be deprioritized as simple, secondary, or only a nice-to-have.

Programming a solution from scratch makes about as much sense as programming your own word processor, database – or any enterprise software. Without a tremendous engineering investment, a DIY approach is too limited to solve the problem. Your technical staff could recreate Gooder charts such as profit curves with scripts, BI tools, or even Excel. But time has proven that such bespoke, cursory implementations don’t do the trick. They lack the interactive visualization and deep customizability that is needed to explore business considerations and establish ML valuation credibility. Static charts don't fit the bill.

Instead, simply configure Gooder AI for your project. It’s the first off-the-shelf product for establishing, maximizing, and demonstrating the business value of a predictive AI project. Doing so demands a turnkey solution that empowers your team to:

• Easily customize the setup to accommodate any metric or KPI, and to form-fit to the business needs of most any predictive AI project.

• Explore the potential impact of business-factor uncertainty – by interactively varying their values and seeing how profit curves and other value-oriented visuals morph, smoothly and responsively.

• Visually navigate the tradeoff options across competing KPIs, all with respect to the spectrum of potential decision threshold settings.

The reason that models are rarely valuated in practice – and that most fail to deploy – is that, until now, there has been no ready-made solution. The first-in-class product, Gooder AI stands to overhaul the industry and remedy its dismal deployment record.

Still unsure? See this more detailed list of capabilities and example project workflow illustrating the importance of Gooder AI's comprehensive functionalities.

No. Gooder AI valuates, plans, and sells the deployment of models developed with other software. It doesn't train any model – it doesn't do the ML part of an ML project. Rather, it represents a new category of software solution that supplements model-training software.

Yes. In fact, uncertain business factors are a main reason to use Gooder AI, rather than being a reason not to. They’re a rule not an exception – not a reason to avoid valuating models entirely.

Gooder empowers you to valuate models despite such uncertainties. You may not have direct knowledge of, for example, the monetary loss for each false positive, because it is privy to other business units, or because it would require new investigations or experimental discovery. With Gooder AI, as you alter the value for such variables by moving sliders, you gain instant insights as to how much the uncertainty matters for driving deployment decisions. In this way, you can narrow that range, determining the limits within which the values would have to land for model deployment to be valuable. By viewing how the shape of the curves morph and how other pertinent metrics change, you gain critical intuition as to how big of a difference such factors make, whether a deployment plan may be copasetic nonetheless, or whether some factors are “too uncertain” to move forward without additional efforts to narrow the range of uncertainty.

In fact, even if you already held fairly ideal visibility into the business factors, some of them would inevitably still be subject to potential change or uncertainty – there are always business variables that are subject to such “wiggle room.”

Gooder AI handles most predictive AI projects, across industry sectors, such as for targeting marketing, churn modeling, fraud detection, credit scoring, predictive maintenance – and most all of the virtually endless range of ML use cases. Gooder applicability is “extremely horizontal” the same as ML as a field.

More specifically, Gooder AI can be used for projects where a model drives per-case treatment operational decisions, such as whether to contact, investigate, treat, audit, or medicate. It is specialized for projects that drive a binary decision (yes/no decision). Beyond that, projects that select between a few rather than only two options for each case can still generally leverage Gooder AI by treating each choice as a small number of binary decisions. Gooder AI does not require the models it valuates to be binary, so long as they are meant to drive a binary decision.

Both. For a new project, Gooder AI plans and sells model deployment – addressing the “non-deployment crisis” afflicting most projects.

For an established project that has already reached deployment, Gooder AI monitors performance in terms of business value and guides adjustments to deployment, including refreshing the model and adjusting the decision threshold. You may have already deployed – but thereafter, you're continually re-deploying. Was the initial deployment maximized for business value? Does it still appear to be so over more recent data? Gooder provides the visibility to answer these questions.

Whether developing a preliminary, proof-of-concept pilot model, a mature model potentially ready for deployment, or a refreshed model over newer data, it's the business value that matters. Until now, most of these project stages have evaluated models only in terms of technical metrics. Gooder AI leverages the same test data to also assess business value (cf. the FAQ below on data requirements).

No. Gooder AI represents a new paradigm – planning and selling model deployment in terms of business value – which does mean introducing a new step into your model-development process, but it’s a straightforward one: Open Gooder on each newly-trained model to explore its potential business value. This long-neglected step organically fits into your existing workflow. After all, to evaluate a model, you already have out-of-sample test data. With Gooder, you can use that data to not only evaluate but valuate each model.

Both. Once set up by the project's data scientist, Gooder AI's interactive interface is friendly for non-data scientist stakeholders. The visuals represent the business realities of potential model deployment scenarios. This isn't the “rocket science” part (the model training) – rather, it’s for steering the rocket.

For the data scientists, setting up Gooder is easy. It’s highly configurable in order to be customized to the business particulars of each predictive AI project. Gooder comes with a standard “starter” configuration that can be easily modified. For details, see this video, which overviews Gooder AI's configuration options.

Empowering decision makers with Gooder’s interface is critical for project success. Only by bringing them into the loop to collaborate in detail across a predictive AI project's lifecycle – a practice sometimes known as bizML – can the project achieve a successful deployment. Gooder AI serves as a key component for executing on this collaborative paradigm. The bizML practice was first presented in Gooder CEO Eric Siegel's book The AI Playbook – see the book's website for excerpts, reviews, videos, and more.

On the other hand, if a data scientist’s customer or stakeholder has entirely delegated the model-deployment strategy, they might never use Gooder. Instead, the data scientist could, for example, use Gooder to establish a few viable deployment options to pass by their client.

Not in the usual sense of the word. Gooder usually strengthens the perception of an ML model, rather than weakening it. The main outcome of using Gooder – and its main purpose – is to empower you to maximize deployed value and to prove that value to your customers, colleagues, and other decision makers. Users of Gooder generally experience its elucidating interface as a validation of business value that they already intuitively believed was there.

This drives deployment. Gooder's value-oriented lens on model performance provides vital evidence to help you convince others and ensure that your model gets deployed – and that it gets deployed well.

On the other hand, some “audits” help rather than hurt. Audits can be oriented toward unearthing, proving, and communicating potential value – placing a spotlight on an initiative's purpose and value so that the value will be realized. Moreover, in some cases Gooder might help you by revealing an addressable weakness in a model. For these reasons, some organizations are considering incorporating Gooder AI into certain auditing practices.

No. The only way to pursue business value during model development is to appraise its business value along the way. And the only way to make prudent business decisions as to whether to deploy, which model to deploy, and precisely how to deploy, is to drive those decisions according to business value.

Explicitly planning for value increases value. It is possible that a model only evaluated technically could turn out to realize value if deployed – but that value would have been left unnecessarily to luck, since the process wouldn't have explicitly optimized for value. What's worse, the value would typically be nil, since most models that aren't valuated aren't deployed at all. Technical performance fails to compel stakeholders.

Gooder AI also maintains ongoing value after deployment. By monitoring performance in business terms, changes to the model or to its deployment particulars (such as the decision boundary) can be driven to maximize business value. ML projects must be continually revisited and potentially redeployed, so model valuation is a must not only pre-deployment, but also pre-redeployment.

Gooder AI stands as a first-in-class, a category of one – for now. It represents a new kind of software for the predictive AI space, and there is not yet any product that directly competes with it. Machine learning solutions, of which there are many, focus on evaluating the technical performance of models. Gooder AI augments those tools by doing what they cannot do: estimating the potential business value in terms of KPIs, while also empowering users – by way of an interactive interface – to manage uncertain business factors and navigate tradeoffs between competing KPIs.

Gooder AI handles any model, from logistic regression and decision trees to deep neural networks and ensemble models like XGBoost – no matter which paid or open-source ML tool was used to create the model. Gooder achieves this boundless applicability because it does not need the model itself, it only needs the model’s scores on test data. See the following FAQ on data requirements for more.

Gooder AI uses the same data that the data scientist has already prepared for out-of-sample model-evaluation: scored test data. Test data consists of the independent (aka input) variables and the labels (aka actuals or dependent variable, i.e., the thing being predicted). By adding on a column for the model's output – that is, the probabilistic score for each case – it becomes scored test data. To compare multiple models, a column must be added for each model's scores.

Gooder AI doesn't need the model itself. It doesn't care how a model works – it only cares how well it works, which is represented by the model’s performance on the test data.

This data requirement is the same as that already needed for standard model evaluation. It is the same data needed to calculate technical metrics such as precision, recall, accuracy, AUC, F-score, lift, confusion matrices, etc. Gooder repurposes that data to estimate the business value that will be realized in deployment.

Gooder AI provides critical visibility into potential business value across the ML lifecycle. Fortunately, the necessary data is typically available at all stages of a predictive AI project, including the development of preliminary, proof-of-concept pilot models, mature models potentially ready for deployment, and, after deployment, models refreshed over newer data. When refreshing, Gooder AI provides a champion/challenger comparison so that the updated model can be assessed in terms of relative business value rather than only in terms of the usual raw technical assessments.

The test data provided to Gooder should typically include between one and a few independent variables (input variables) that could help with projecting the model's value, such as the size (dollar amount) of each transaction for fraud detection. Generally, the majority of independent variables aren't used by Gooder, although it's okay to leave them in.

To view example datasets input to Gooder, see this short video.

When training models with popular Python libraries, Gooder’s data requirements can be fulfilled automatically. The gooder-ai package opens Gooder AI on models developed with Python libraries such as XGBoost, scikit-learn, and PyTorch. It does so by fulfilling this data requirement under the hood.

Yes, Gooder AI is SaaS. It runs inside your browser rather than as a stand-alone app. However, once you've opened the product in your browser, you may optionally select “offline mode” so that your data and all project specifics remain strictly local, without going to the cloud. In “offline mode,” you can load your data locally and save your Gooder AI visualizations locally. As with any data-related project, there can be advantages to allowing cloud access – so, if you choose not to select “offline mode,” you will be permitted to share your project data, as well as your Gooder AI visualizations, through the cloud via Amazon Web Services.

Note also that the necessary data, described in the previous FAQ above, is relatively innocuous, since Gooder does need any identity-revealing variables to establish model performance.

Gooder AI is free to use. This is to facilitate the emerging adoption of the new paradigm that it enables: bridging the gap between model development and model deployment by maximizing and selling the potential business value of ML models. As a company, we are pressing the industry to widely adopt this paradigm as a best practice for predictive AI projects. Paid premium features will be added to Gooder AI at a later date.

Learn more about Gooder AI and start using Gooder AI hands-on today.

Yes. In addition to the Gooder AI software product, we as a company provide hands-on assistance in leveraging Gooder to plan, sell, and greenlight the deployment of your predictive AI projects. Our assistance can provide a small boost to help you get started, or ongoing support. For more information, visit the Gooder AI professional services webpage.