Where does your next euro return the most? fly cockpit shows you.
fly cockpit connects all your channels on one data foundation. You see right away where the next euro returns the most, in the short term and across the whole time a donor stays with you (lifetime value). Your crew of digital employees works on the same foundation. The first one is Daisy, your data science analyst.
- All channels connected
- Annual plan by return
- Crew with Daisy
fly cockpit: a call, 15 minutes, you bring one goal
Watch Daisy at work
An example from the product, shortened. The numbers are made up, not a real run.
Your AI agent for fundraising data, audiences, mailings, reports and donor questions. Let's work through it together.
I've run the analysis. To compare channels fairly, each cohort gets the same amount of time: donors who gave recently haven't had the chance to make a second gift yet.
Cohort: donors whose very first gift was between 24 and 6 months ago. Observation window: exactly 6 months after the first gift, the same for everyone.
| Channel | Newly acquired | Median first gift | Median total value, 6 mo. |
|---|---|---|---|
| direct mail | approx. 11,600 | €28 | €29.40 |
| other | approx. 8,100 | €45 | €46.70 |
| donation form | 14 | €55 | €57.50 |
Is the difference random? No. The chi-squared test for repeat rate and the Mann-Whitney tests for value and timing each return p < 0.0001. Because donation amounts are heavily skewed, I use non-parametric tests rather than means.
Acquisition costs unified: I allocated production, postage, media and agency costs per channel to the donors actually acquired. Year 1 is observed (cohort with first gift 24 to 12 months ago). Years 3 and 5 I project from older cohorts (first gift 2018 to 2021) using a survival curve. "Net" means cumulative donation value per new donor minus acquisition cost.
| Channel | Cost per new donor | Net after 1 yr | Net after 3 yrs | Net after 5 yrs |
|---|---|---|---|---|
| direct mail | €38 | −€4 | +€23 | +€46 |
| other | €52 | +€6 | +€52 | +€89 |
What this means for the CAC question: Direct mail is cheaper per head but only pays back after about 14 months. "Other" costs more, but converts to repeat giving more often and faster, and is in the black after about 11 months. After 5 years, each euro invested returns roughly €1.20 net (direct mail) versus €1.70 (other). Year 5 is a projection with an uncertainty band of about ±15%. Anyone who only looks at first-gift acquisition cost would favour direct mail and leave money on the table over the longer horizon. That requires exactly this unified cost per channel.
Would you like me to save this?
Where data analysis in fundraising often gets stuck
Every question becomes a ticket
The answer comes weeks later from the data team or the agency. By then the print run is printed.
The numbers live in many systems
Database, mailing tool, online numbers: nobody combines them, because it is manual work every time.
A model looks convincing but only guesses
With too few cases, any model returns a list that looks good and costs money.
Flight data: Your data works for you, without a data team of your own
Connect
All channels and sources come together on one data foundation (data lake).
Check and enrich
Every import is checked before anything is used. The analysis runs on anonymous donor IDs.
Train models
On request, Daisy trains a model for your goal and tests it on data she set aside beforehand.
Decide
You see where the next euro returns the most, with the calculation and its limits. The decision stays with you.
Your crew
Digital employees that work on your data foundation. You stay the pilot and decide.
Daisy, data science analyst
Answers questions directly in your data, trains models for a goal and tells you what the result does not prove. An extra resource for your team.
More crew members
More digital employees will join on the same data foundation.
What is in the day-to-day product
From annual planning to the report. These are the screens you work with.
Annual plan: Your budget follows the expected return
You plan your fundraising budget by predicted return, not by last year's numbers. Daisy spreads the budget across mailings, reactivation, second gift and online new donors, and shows you the net return of each scenario.
You compare plans before you commit. Mailing sizes come from the plan, and you export it for finance and management.
The overview: which plan pays off?
Seven scenarios, one question: how many letters are really worth it? Here you see profit, donor retention and retention costs at a glance.
Step 2 is the optimum: 24,600 letters achieve the highest profit (€60,800) with still manageable retention costs of €158 per donor.
| Scenario | Letters | Profit | Lost donors | Avg. gift per lost donor | Cost to retain, per donor |
|---|---|---|---|---|---|
| KI-Vorhersage | 12,400 | 54,800 € | 680 | 22 € | 45 € |
| Schritt 1 | 18,800 | 58,100 € | 310 | 24 € | 89 € |
| Schritt 2 = Optimum | 24,600 | 60,800 € | 145 | 27 € | 158 € |
| Schritt 3 | 30,500 | 59,300 € | 72 | 31 € | 241 € |
| Schritt 4 | 35,200 | 57,100 € | 28 | 34 € | 378 € |
| Schritt 5 | 38,700 | 55,400 € | 9 | 37 € | 489 € |
| Real versendet | 41,000 | 54,100 € | – | – | – |
Example view with illustrative figures, not real records.
Selection: The mailing ends where a letter brings in less than it costs
Selection ranks every donor inside a mailing by expected value. The model learns from your own donation history, and every result feeds the next ranking.
You cut mailings below break-even. The postage goes into reactivation or second-gift mailings, where the next euro returns more.
34M donation decisions predicted and optimized, 380+ campaigns in the DACH region.
Selection: choose donors.
Price Points: run for all eligible donors.
Send Lists: finalise the selection.
▼ Advanced settings Engagement model active
Optional. With engagement data enabled, the selection uses the engagement model, which additionally weights how donors have responded to this topic and incentive in the past.
Settings
The ROI focus controls whether the export can use a recommendation list. At 100 % ROI focus, recommended donors are those whose projected donation value per mailing (DMV) covers the letter cost.
Optional. A run without a name is listed by its start time.
Expected results
Selected contacts
The breakdown shows which donor segments the selection includes.
| Segment | Contacts | Share | Avg. DMV |
|---|
Audiences: Build segments and reuse them everywhere
You build segments such as at-risk monthly donors, first-time donors from the last 12 months or reactivation candidates. You use the same audience in mailings, campaigns and reports.
Start from ready-made templates.
Daisy checks every template against your donor data and shows right away how many people match. You use a template with one click.
Major donors
Giving rhythm interrupted
Upgrade candidates
Single-gift donors
Christmas-only donors
Regular giving · active
Example view with illustrative figures, not real records.
Reports and dashboards: Last year, campaigns and ROI
Year-on-year comparison, campaign performance and ROI at a glance. You export the numbers as CSV for finance and management.
Annual donor count by gift category (last 5 years)
Data as of 28 Jul 2026, 16:40
Calculation notes
Example data. A donor counts once per year in the category of their total.Gifts by channel (current year vs. last year)
Data as of 28 Jul 2026, 16:40
Calculation notes
Example data. A donor counts once per year in the category of their total.Churn-risk segments over time
Data as of 28 Jul 2026, 16:40
Calculation notes
Example data. A donor counts once per year in the category of their total.Monthly active donors (last 24 months)
Data as of 28 Jul 2026, 16:40
Calculation notes
Example data. A donor counts once per year in the category of their total.Second-gift rate by channel (new)
Data as of 28 Jul 2026, 16:40
Calculation notes
Example data. A donor counts once per year in the category of their total.Example view with illustrative figures, not real records.
Getting your data in: Three ways, one fits your setup
Manual import: You upload your export, the AI maps the columns to the data model, you review and confirm. Open API: Your CRM delivers data continuously. We help you integrate: With an unusual system, an odd export or little IT time, we work out a solution with you. So far we have always found one.
Every import is checked before anything is used. The analysis runs on anonymous donor IDs, and personal data stays with you.
Import data you can trust.
Upload a file, map your columns and check the result in the preview. Nothing changes in your data until the numbers add up and you agree.
Which data do you want to import?
Pick one. We recommend importing gifts first. You can add the others afterwards.
How does the data reach us?
Drop a file here or
Map columns
Preview
Checks
sftp.example-npo.orgPath/export/gifts_*.csvScheduledaily at 02:00Last nights
| Mon, 02:00 | ✓ successful | 412 rows |
| Sun, 02:00 | ✓ successful | 388 rows |
| Sat, 02:00 | ✓ successful | 1,027 rows |
New gifts land in your data straight away, with no file and no waiting.
📎 raffle_2026.xlsx
Connect your own MCP server
You connect the tool yourself, and Daisy can work with it afterwards.
Recent imports Show
| Date | Data | Route | Rows | Status |
|---|
Example view with illustrative figures, not a real import.
Explorer: Look into your own data
You see what was imported and how it was checked before you rely on a number.
Data explorer
Browse contacts, gifts and contracts. Filter and export the records you need.
| Donor ID ↕ | Country ↕ | Language ↕ | Gender ↕ | Audience ↕ | Birth year ↕ | Location ↕ | Sponsorship since | Favourite topic | Newsletter |
|---|
Example view with illustrative figures, not real records.
How does your model keep learning?
Daisy tests ideas in loops and keeps only what improves the prediction.
Your model keeps learning in loops.
Each round adds one idea to the best model so far. Daisy measures it on held-out contacts the model has never seen. If AUC improves, the idea stays. If it gets worse, it goes.
AUC measures how well the model ranks donors (0.5 = chance, 1.0 = perfect).
| # | Experiment | AUC | Change | Status |
|---|
Result after 10 rounds
Your effort
Name the goal and approve the result. Daisy does the trying.
What the model knows now
The three features that helped all say the same thing in different ways: the shape of the giving history matters more than the raw data. The log transform smooths outliers. Giving rate per day adjusts for pauses of different length. The square root of the number of gifts dampens the pull of frequent donors. A reactivated person is not an active donor, so raw frequency helps less than its shape.
Two frozen audiences
| Audience | AUC | Saved |
|---|---|---|
| Reactivation · top 25,000 by model score | 0.688 | ✓ Original model |
| Reactivation · top 25,000 by model score v2 | 0.712 | ✓ After the loop |
Frozen means the selection stays as it was. Later you can compare which version did better.
Example view with illustrative figures, not a real model. Real values depend on your data.
Give her a goal
Pick a goal, and Daisy trains and tests a model. Worked example with made-up numbers.
What you get: answers you can check
- How do newly won donors develop per channel, and what are they worth after 1, 3 and 5 years?
- How big should our house mailings be in 2027? Where does the print run no longer pay off?
- Which one-time donors are most likely to become monthly donors, and which channel reaches them best?
These are example questions from the product. Daisy shows every step and tests models on held-out data before giving you a list. A simple rule of thumb serves as the comparison.
An example of an honest no: with 9 bequests in 6 years, a model would only guess. Daisy then gives you no list and suggests asking about interest in bequests in the next mailings and retraining in 12 months. This example uses invented numbers and is not a real run.
Daisy works with anonymous donor IDs. Personal data stays in your house.
What fundraisers who work with it say
“I played around with Daisy for a few days and I'm thrilled. […] It now feels as if the AI understands my needs and my data correctly.”
“We are working with flyraising quite intensively now, and it is honestly fun to chat with Daisy and look at the forecasts and analyses.
It's really great that so many points are being implemented so quickly.”
Quotes from customer messages, translated from German, anonymized and lightly shortened.
Frequently asked questions
What is fly cockpit?
fly cockpit is our platform for fundraising data in day-to-day use. Your channels sit on one data foundation, you plan by expected return, and Daisy answers questions directly in your data.
What is an AI agent?
An agent does not just answer one question. It plans steps, calculates and carries them out. Daisy spins up her own server when needed and trains a model for your goal.
What is a data lake?
A shared data foundation where all your sources come together: donations, audiences, mailings and reports.
Does my IT have to approve anything?
No. Daisy works with anonymous donor IDs. Personal data stays in your house.
Can the AI guarantee donations?
No. Daisy shows you probabilities and limits. Whether a letter or call really triggers something is only shown by a small test with a control group.
How do I get started?
With the Tailwind Check: 2 exports, 15 minutes, result in 3 days. Then you decide whether to continue.
Am I even allowed to share my data?
For the Tailwind Check you do not need addresses, email addresses or names. We work with anonymous donor IDs. If personal data should flow in day-to-day use after all, a data processing agreement (DPA) covers it.
Do audiences stay current, or do they go stale?
Audiences with time conditions always recalculate from today. If you set “no donation in the last 12 months”, two months from now they only contain people for whom that is true then. Anyone who has donated in the meantime drops out automatically. There are also fixed lists that always keep the same people.
How does flyraising handle communication stops and block lists?
Your blocks from the CRM come in as custom fields, for example no mailing in January or at most one mailing per month. If the send date falls into a block, the selection removes those people automatically.
Can I see in the export why someone is in the selection?
Yes. The export shows per person which audience brought them in, plus rank, expected donation probability and personalized donation suggestions. That lets you assign send groups and mailing versions for the lettershop.
Does Daisy understand our own terms?
You can store definitions, for example what reactivation means at your organization. Daisy then works with your terms and answers questions about your own segments.
Which goal should your crew tackle?
Pick a time. Bring a goal, 15 minutes are enough.
See your annual plan with your own numbers →fly cockpit: a call, 15 minutes, you bring one goal