# Split your registrant list down the middle.

URL: https://callhush.com/pilot
Last updated: 2026-10-28

Before you commit to anything, we run one webinar cycle with your registrant list split 50/50 at registration. Half get our texts and calls, half get exactly what you run today. Same ads, same week, same offer: and both halves counted the same way.

**The pilot is one webinar cycle with your registrant list split 50/50 at registration.** Half of them get our texts and our calls; the other half get exactly what you run today, with nothing changed. Same ads, same week, same offer, same webinar: and both halves are counted the same way, on show-up rate and on revenue per registrant.

## Will you run a test on my own list before I commit?

Yes, and we propose it first rather than conceding it when you push. (For the technically minded: what this is, in the statistician's vocabulary, is a 50/50 holdout.) Anyone who tries to talk you out of running one is telling you something about themselves.

The reason is not confidence. It is that **no client names, logos or testimonials are published on this site**, there is no results dashboard, and we are not going to invent any of it. The only evidence that would reasonably move you is evidence generated on your own list, in your own week, against your own offer. So that is the thing we sell first.

> **What you would be agreeing to:** One webinar cycle, priced individually after a call. Half your registrants worked, half untouched. Both halves reported to you in the same document, whichever way the numbers come out.

## How does splitting your registrant list down the middle actually work?

1. **Split at registration, not after**: Every registrant is assigned to one of two halves as they come in, before anyone knows how they are going to behave. Assigning after the fact is how a test becomes a story.
2. **One half gets worked**: We text them, we call them before the session, and we work the ones who did not show afterwards. The assistant says what it is, asks why they signed up, and the later call references what they actually said.
3. **The other half changes nothing**: Your existing reminder emails keep running exactly as they do now. The comparison is against what you already do, not against sending nothing: which is the comparison most published reminder research quietly makes.
4. **Both halves counted the same way**: Show-up rate and revenue per registrant, same definitions, same window, both numbers reported side by side. You see the treated half and the untouched half in one document.

Nothing else moves during the cycle. Same ad spend, same creative, same registration page, same replay policy: because the moment two things change at once, neither of us can say which one did the work.

> **Read the result as directional:** Both halves sit in the same webinar, the same chat and often the same retargeting audience, so this is not a clean geographic split and contamination between the arms is structural. A contaminated split should be documented as directional rather than used as the sole basis for a budget decision. That is our own research note on the method, and it applies to our own pilot too.

## Why 50/50, and not the 10% I would rather give you?

Because the small control group is the expensive one, and that is counter-intuitive enough to be worth showing rather than asserting. Commercially you want to hold back as few registrants as possible. Statistically, holding back fewer registrants is what makes the test take forever.

- **2,748**: Registrants needed to detect a +5pp difference in show-up rate when the list is split down the middle (80% power, α=0.05, 30% baseline)
- **7,388**: Registrants needed for the same test when only one in ten is held back
- **~2.7×**: How much faster the even split reaches significance. CallHush power analysis, two-proportion z-test

*The same test, two ways to configure it*

|  | Split down the middle (50/50) | Keep a small control (90/10) |
| --- | --- | --- |
| Registrants needed to detect +5pp | 2,748 | 7,388 |
| Time to a readable answer | Baseline | About 2.7× longer at the same volume |
| Registrants not worked | Half of them, for one cycle | One in ten, for a lot longer |
| How it feels when you sign | Like giving up more | Like giving up less |
| What it actually costs | One cycle of held-back registrants | Months of not knowing, then a re-run |

So the honest configuration and the cheap configuration are the same configuration. That is a rare piece of luck in this business and we are not going to argue you out of the version that reads faster.

The 30% baseline inside that calculation is a planning assumption, not a claim about your list. Where it comes from, and why most published webinar benchmarks are measuring a completely different population, is set out on [webinar show-up rate by traffic source](https://callhush.com/webinar-show-up-rate-by-traffic-source).

## What gets measured: show-up rate, or revenue per registrant?

Both, reported together, because show-up rate on its own is a vanity metric. Filling a session with people who were never going to buy is a worse outcome than a smaller number who attended live because they had already told someone, in their own words, why they signed up.

- **Show-up rate**, treated half versus untouched half, on the same definition of attendance for both.
- **Revenue per registrant**, which is the number that decides whether any of this was worth doing.
- **What your registrants said about why they signed up**: in their own words, which is the part of the pilot that has value even if the show-up rate disappoints.

> **Why we will not promise you a number:** We are not going to tell you the lift in advance. Proving an incremental effect to invoice-grade certainty takes more registrants than one client produces in a reasonable window, which is exactly why nobody in this category publishes a clean one. The pilot buys you proof and it buys us credibility. It does not buy either of us a guarantee, and a vendor who offers you one is quoting a number they cannot have measured.

## If it fails, how will we both know within 30 days?

Within one cycle you will know what happened on your list. Whether you will know it to the standard a statistician would sign off depends entirely on your volume: at 500 registrants a week the figures above land in a matter of weeks, and on a smaller list they do not land inside a month at all. We will tell you which of those you are before you start, not in the readout.

If the treated half does not beat the untouched half, that is the answer. You get the same document either way, you are month to month, and there is no term to serve out while we explain the result away.

- **None**: Client names, logos or testimonials published on this site. No case studies, no results dashboard, and none invented
- **One cycle**: The length of the pilot: one webinar cycle, priced individually after a call
- **Month to month**: The commitment after it. Shortening the term is the first thing we concede, never the price

The two things people usually want before booking are the wording and the money. The wording is on [what it says](https://callhush.com/what-it-says), the deal shape is on [how pricing works](https://callhush.com/pricing), and our standing rule about publishing the calls that went badly is on [hear real calls](https://callhush.com/hear-real-calls). When you are ready to run the split, [book a call](https://ainora.lt/contact?from=callhush).

## Frequently asked questions

### Do I have to sign a contract to run the pilot?

No term. The pilot is one webinar cycle, priced individually after a call, and what follows it is month to month. Shortening the commitment is the first thing we concede in a negotiation, never the price: an annual lock-in tells a buyer we know something about our own churn.

### What does the pilot cost?

You get the number on the first call, in the sentence after you ask. We do not publish it on the website and we do not wait to hear your ad spend before answering. The reasoning, and the full shape of the deal, is on the pricing page.

### Can I give you 10% of the list instead of half?

You can, and we will tell you it is the expensive option. Detecting a five-percentage-point difference in show-up rate needs about 2,748 registrants when the list is split down the middle and about 7,388 when only one in ten is held back. The split that feels cheapest is the one that takes roughly 2.7 times as long to read.

### What if the treated half does worse?

Then that is the result and you get it as it came out, in the same readout we would have shown you if it had gone the other way. We do not have a mechanism for making a bad result look good, and inventing one would cost us far more than one lost client.

### Do you bill me on the difference between the two halves?

No. Billing on measured incremental lift needs a permanent control group, more registrants than one client produces in a reasonable window, and a baseline dispute with no neutral referee to settle it. We bill on total attended calls. The pilot is a credibility exercise at onboarding, not an invoicing mechanism.

### Who approves what the assistant says during the pilot?

You do, every line, in writing, before anything sends, and changes go through the same approval mid-campaign. The exact opening line, including the part where it says it is an assistant, is published word for word on the What It Says page.

### What happens if one of my registrants complains?

A complaint escalation pauses the system and notifies you, and a breach of that voids the contract. We would rather give you that than a money-back guarantee, because money back is worthless here: you cannot refund a reputation.

---

CallHush is operated by AINORA, MB (Vilnius, Lithuania). Book a call: https://ainora.lt/contact?from=callhush