Written by Evan Aanerud

Dataro is the “GiveTech” startup’s propensity modeling software that uses machine learning technology to generate predictions about how donors are likely to behave. Dataro’s AI software aims to maximize income and minimize wasted costs, as well as transform the way that fundraisers stay alert and relevant to how donors are engaging with their organizations. Back in the product’s early development, it was selected as one of the top 10 fundraising innovations for not-for-profits in a competitive global open innovation challenge in 2020. Winning the competition built momentum for Dataro’s international plans. Dr. Tim Paris, David Lyndon, and Chris Parver, co-founders of Dataro, were able to raise $2.5 million in seed funding by 2021. With over 100 clients between Australia and the U.S. after three years of operations, Dataro’s rise certainly should be noted.

The Technology & Model

Figure 1: How AI for nonprofits works. Source: Dataro.

Here is how the process works with Dataro’s machine learning tools:

The machine learning software analyzes the fundraising data, splitting it into hundreds of discrete “features” like gift recency and donor age.
The AI searches through these features to find trends, patterns, and relationships. It uses these insights to begin training algorithmic models that can make predictions for the specific desired propensity.
The software evaluates its trained models to find the one that performs the most accurately based on all the complex relationships in the organization’s historical data.
This model is used to generate propensity scores and ranks for each individual donor, showing how likely they are to behave in a certain way and how they compare against the rest of the donors in the database.
These scores and ranks are exported back to the organization’s CRM dashboard for easy use and analytic metrics, giving a real-time AI-driven list of the ideal donors who should be contacted.

Differentiating Dataro

Let’s differentiate Dataro from competitors:

Unlike Boodle.AI, Dataro does not augment data from external third-party sources. Dataro’s predictive models are only trained using the nonprofit’s data. This ultimately takes the guesswork out of fundraising.
Dataro is not a wealth screener. It doesn’t generate artificial wealth scores to grasp a prospect’s capacity. The software predicts donor behavior to help fundraisers prioritize attention on specific donors.
Having originated in Australia, Dataro’s system is built around compliance with the Privacy Act of 1988 and the Australian Privacy Principles. The company encourages and uses anonymized, tokenized data wherever possible, preventing an organization’s data from potential data breaches.

Getting Started with the Software

Figure 2: Dataro’s sign-up page. Source: Dataro.

To better understand the interface and technology, Dataro offers a free-of-charge demo trial with mock organizational data. I will use the demo trial as a basis for my product review below.

Branding and UX Design

When I first opened the Dataro on my laptop, I was immediately impressed with the branding aesthetic and interface design. In terms of its visual identity, the pastel color palette and vector graphics distinguish Dataro as an approachable and brand-conscious platform. Moving around the software feels as if I’m scrolling through my phone. There are no obtrusive pop-ups, delays, or distracting background information. It doesn’t feel like I am navigating an AI-equipped donor platform, and I think that is intentional.

Figure 3: A screenshot the homepage of Dataro’s app. Screenshot by author.

Campaign Results

Under Campaign Results, Dataro users can track basic data such as total raised, net return, return on investment, response rate, and average gift size. For the purpose of this review, I chose the June Direct Mail Appeal to analyze, which has yielded a fantastic ROI of 155% and a response rate of 8.7%.

Moving further down the page, Dataro evaluates its prediction results. Dataro scores approximate the probability of a patron doing a particular action, and the higher the score, the higher the probability of an individual taking the associated action. From the graph below, 79.63% of patrons with a 90 to 80% score responded to the DM appeal, which is just under the threshold of 80%. From the trend of the graph, it appears that the model just slightly underperformed in each score band.

Figure 5: Dataro’s prediction results using the free trial. Source: Dataro.

Dataro ranks order the organization’s donors from most to least likely to take a particular action. A person with Rank 1 is the most likely to act. Here, the graph illustrates that the individuals Dataro ranked from 1 to 1,000 had a collective response rate of 45.5%.

Figure 6: Dataro’s rank results using the free trial. Source: Dataro.

We can dive deeper into this data with Dataro’s prediction segments, listed in the tables below. Dataro gives us insight into the average gift amount, gifts over $500, and the largest gift of those top-ranked 1,000 individuals.

Figure 7: A campaign’s performance results based on Dataro Ranks. Source: Dataro.

My favorite graph in the Dataro app is the Dataro Score Band by Net Returns. This graph illustrates where the organization starts to lose money on its ranked individuals. In this case, supporters ranked 4,000 to 9,815 yielded negative returns for the organization.

Figure 8: A graph on Dataro’s free trial comparing a campaign’s actual performance to its ranking prediction before it ran. Source: Dataro.

The DM Appeal report is just the beginning of Dataro’s impressive array of propensity models. These include a Regular Giving (RG) Churn report, which measures the likelihood that a regular giver will churn/attrite in the next six months. The RG Upgrade, RG Reactivation, and Convert to RG propensity models are all available within Dataro to support the organization’s Regular Giving. Each report contains a similar dashboard as the DM Appeal campaign, a results table, and a table with results segmented using relevant segments from your CRM.

The Campaign Results tab demonstrates Dataro’s accessible design and user-friendly interface. I do wish there was an option under each campaign to view the names of campaign supporters. The dashboard gives a substantial overview of the successes and failures of the campaign, but nonprofit fundraisers often operate from a degree of familiarity and name recognition. It would provide a more holistic and individualized outlook on the campaign and lead fundraisers where to focus their time and money. This isn’t a major issue, since we can see each individual’s rank and score through the Dataro dashboard within the organization’s CRM, which will be covered later.

Prediction Insights

Figure 9: The Prediction Insight Tab. Source: Dataro.

The Prediction Insights tab provides a summary of each campaign’s predictive model, including the score band, the number of supporters in that band, and the percentage makeup of that band in the total cohort. For the DM Appeal campaign, four individuals are 90 to 100% likely to give to a DM appeal, just .01% of the cohort. Dataro’s machine learning pipeline is detailed below.

Figure 10: Infographic of Dataro’s Machine Learning process. Source: Dataro.

The Prediction Insights tab could also benefit from a “See Who” button to list the individuals that Dataro has generated using the predictive model. Especially in a Churn, Activation, or Upgrade campaign, the donor most likely to act is a digestible volume of data and names for the fundraiser to use.

Model Performance

Dataro has a system of checks and balances to test the quality and accuracy of their models. The Model Performance tab details the prediction quality and top predictors for each campaign. The DM Appeal earned a 96% stellar grade according to Dataro’s evaluation, and the model was trained mostly on the frequency of DM gifts, payment method, and total single giving.

In true deep learning nature, the machine learns from the organization’s data. Each Dataro-trained model is unique to the organization’s data and unique to the moment in time. Considering how people are motivated at different times in the year to give (ex. tax season), it’s highly advantageous that Dataro constantly retrains its models to provide the most accurate predictions at each point in time. Dataro’s models are constantly self-adjusting, eliminating mistakes that didn’t accurately predict the right likelihood of action.

It’s impressive to see such transparency behind Dataro’s modeling techniques. I wonder how the volume of available data dictates each model’s performance evaluation. The DM Appeal had a training data size of over 8.7 million, with 1.9 positive cases and 362 total predictors. The data training size isn’t that large for other evaluations, but it still begs the question: will Dataro’s models be as effective on my organization’s smaller data set? More on this below.

Putting a Name to a Propensity Score

Figure 12: Dataro’s Donor Contact Page. Source: Dataro.

Dataro’s CRM integration is where the data starts to come to life. In the organization’s CRM, Dataro’s dashboard appears on the donor’s contact page. The above example is within Salesforce. Here, we can see the propensity scores that are only relevant for the specific contact. Propensity scores are listed on the left (on a scale of 0 to 1) and measure the individual’s likelihood to take a certain action. The propensity ranks, listed on the right, compare the individual’s likelihood to take the action against the entire cohort of donors.

When the fundraiser begins to initiate a campaign, they will want to create a report in their CRM using Dataro’s propensity scores. In the video below, Dr. Chris Parver demonstrates the simplicity behind creating a list based on the top 500 ranked individuals likely to give to a DM appeal. Dataro’s models and predictions are also dynamic, updating weekly so the organization will always get the most accurate predictions in real-time.

Figure 13: A demonstration of how to use the tools on Dataro. Source: Dataro.

I cannot evaluate the effectiveness of Dataro's CRM integration capability as I am using the free trial software and not any organizational data. From Dataro's Help Desk videos, it seems like accessing the donor's scores and ranks on their contact page help the fundraiser make meaning of the data. It's important to remember that these scores are predictions and not guarantees of a donor's behavior.

Data Requirements and Integrations

AI and machine learning are only as good as the data they are working with. Machine learning algorithms require large volumes of data in order to learn patterns and behaviors, and to create models based on that analysis. Prospective nonprofits hoping to utilize Dataro or other AI technology may be at a standstill if they don’t have enough usable data.

For Dataro to achieve predictive accuracy, the client needs at least five years of giving history, and at least 10,000 constituent records. Those records don’t all necessarily need to be active, but it is crucial for the system to work on a substantial amount of active data.

Dataro currently works with the following CRM systems: Donorfy, Salesforce, ThankQ, Blackbaud, Raiser’s Edge, NXT, Microsoft Dynamics, and Advanced NG.

Looking through Dataro's current clients, the case studies are powerful applied examples of the technology Dr. Paris said the Royal Flying Doctor Service Victoria increased net revenue in their 2020 Tax Appeal by nearly $35,000 by using Dataro’s propensity scores. But I'm surprised to see very few, if not any artistic clients listed. After speaking with Debbie Stanley, Leader of U.S. Expansion at Dataro, it doesn't seem that Dataro has partnered with many arts organizations. This is in part due to their very recent expansion to the U.S.
I see contradictions between Dataro's mission to democratize the use of AI technology and the limitations inherent within the nonprofit infrastructure. The current CRM systems that Dataro integrates will not work with the systems of many nonprofits. Even if nonprofits have the correct integrated software, the organization likely will not have enough active data records. These issues point to how the emerging use of AI for philanthropy may potentially skew towards better-resourced platforms and nonprofits, bringing AI inequities to mind.

Conclusion

Researching Dataro gave me a certain appreciation for AI-equipped fundraising software that focuses on donor identification. Using donor prediction models puts the fundraiser first and the AI second. AI-powered donor communication systems, on the other hand, do the opposite. I would not advocate for a nonprofit to begin using AI-equipped donor communication tools for several reasons. The ethics of using AI for donor solicitation are tricky to navigate considering authorship issues. Additionally, nonprofit staffs are typically their own experts on donor messaging. Between acknowledgment letters, email campaigns, and day-to-day communications, professional fundraisers know how to speak with their donors. It is the decision-making, tasks like optimizing a list of 5,000+ potential donors for solicitation, that development staffs stand to benefit the most from this software.

I see Dataro as a revolutionary force in nonprofit fundraising. The technology, if truly effective with the right organization’s data, harnesses the potential to shape how we prioritize our time with donors. At this point in time, Dataro is an elite solution. If the product can scale to organizations with smaller datasets and a broader selection of CRM software, then the future is bright for this tech startup.

AI is like hot sauce. A little bit, used in the right place, can be perfect. But dumping too much on or using it everywhere will probably ruin your meal.
Nick Hamlin, AI4Giving

My advice to organizations curious about Dataro or another AI-equipped fundraising platform: approach AI as an experimentation strategy with evaluation metrics baked into the process. We experiment with ideas and projects that are never fulfilled. If the experiment indicates growth for your organization, then move forward and implement the technology into your existing system. AI should not completely reinvent the nonprofit fundraising wheel.

+ Resources

Australian Government. 2014. “Privacy Act 1988.” Legislation.gov.au. Attorney-General’s Department. March 12, 2014. https://www.legislation.gov.au/details/c2014c00076.‌

Bindbeutel, Duncan. “Navigating the Ethics of Using AI for Donor Solicitation.” AMT Lab @ CMU. April 20, 2022. https://amt-lab.org/blog/2022/4/how-can-nonprofits-navigate-the-ethics-of-using-ai-for-donor-solicitation.‌

“Boodle for Nonprofits.” BoodleAI. Accessed April 22, 2022. https://boodle.ai/nonprofits/.

“Dataro Demo Account.” Dataro. Accessed April 22, 2022. https://www.app.dataro.io/home

“FAQ – Dataro.” Dataro. Accessed April 22, 2022. https://dataro.io/faq/.

“Fundraising Powered by AI – Wisely Product.” Wisely. Accessed April 22, 2022. https://fundraisewisely.com/wisely-ai-powered-fundraising-software-product/.

“Gravyty First Draft.” n.d. www.gravyty.com. Accessed April 22, 2022. https://www.gravyty.com/firstdraft/.

Kanter, Beth, and Allison Fine. 2020. “AI4Giving: Unlocking Generosity with Artificial Intelligence: The Future of Giving.” Bill & Melinda Gates Foundation.

Martel, Adam. 2021. “How Artificial Intelligence Is Transforming Fundraising.” Education and Career News. March 7, 2021. https://www.educationandcareernews.com/future-of-higher-education/how-artificial-intelligence-is-transforming-fundraising/.‌

Alex Moltzau. 2019. “Artificial Intelligence and Nonprofits.” Medium. September 1, 2019. https://towardsdatascience.com/artificial-intelligence-and-nonprofit-e6cdaaae166f.‌

Parver, Chris. 2021. “Fundraising Analytics: 20 Critical Metrics & How to Use Them – Dataro.” Dataro. March 25, 2021. https://dataro.io/2021/03/25/fundraising-analytics/#future.‌

—————.“Help Desk – Dataro.” Dataro. https://support.dataro.io/hc/en- us/articles/1500004945662-What-are-the-definitions-for-each-Dataro-model-.

—————-. “How Do I Tell If Dataro’s Predictions Worked? – Dataro.” Dataro. 2021. https://support.dataro.io/hc/en-us/articles/1500004945662-What-are-the-definitions-for- each-Dataro-model-.

—————. “Minimum & Suggested Data Requirements– Dataro.” Dataro. 2021. https://support.dataro.io/hc/en-us/articles/1500004945662-What-are-the-definitions-for- each-Dataro-model-.

Paris, Tim. 2022. “Beat the Machine by Dataro.” GoToWebinar. Presented at AI for Fundraising: Learn from an Expert, April 19.

Rizal, Paulo. 2020. “Web Login Service – Stale Request.” December 16, 2020. https://courses.ideate.cmu.edu/62-830/s2022/wp-admin/post.php?post=1415&action=edit.‌

“Predictive Advertising for Modern Agencies and Brands.” BoodleAI. Accessed May 6, 2022. https://boodle.ai/predictive-advertising/.‌

Stanley, Debbi. 2022. Dataro Product Demo Meeting Interview by Evan Aanerud. Dataro.

Thomsen, Simon. 2021. “AI-Led ‘Givetech’ Dataro Lands $2.5 Million Seed Round to Help Charities Innovate Fundraising.” Startup Daily. December 16, 2021. https://www.startupdaily.net/2021/12/ai-led-givetech-dataro-lands-2-5-million-seed- round-to-help-charities-innovate-fundraising/.

Warner, Gregory. 2019. “3 Simple Steps to Break Your Transactional, Spray-And-Pray Fundraising Addiction.” www.linkedin.com. September 10, 2019. https://www.linkedin.com/pulse/3-simple-steps-break-your-transactional-fundraising-addiction-warner-1e/.‌

“What Are Australian Privacy Principles | Privacy Act Guidelines | Imperva.” n.d. Learning Center. https://www.imperva.com/learn/data-security/australian-privacy-principles/.‌

“10 Must-Have Features to Have on an AI-Driven Fundraising Platform.” 2019. BoodleAI. September 19, 2019. https://boodle.ai/blog/artificial-intelligence/10-must-have-features-on-ai-driven-fundraising-platform/.