Most of what people actually want is not one campaign. “Mail them their cart back, and catch them when
they return” is two campaigns plus the rules that carry a person from the first to the second — and those
rules are the half that is easy to get silently wrong.
A program is a guided path: pick the outcome, answer a few questions, and Personyze creates the
campaigns already wired to each other. It lives under Programs in the menu.
Each outcome shows the shape it builds before you choose it — Email → On the site, On the site, Email → Email — and whether it comes with a holdout or no test. “Pick a shape below and you will have two or three staging drafts in about a minute — nothing reaches a visitor until you publish them.”
publish them. You can run one, look at what it made, and delete the lot.
What you can pick
| Outcome | The ask, as somebody would put it |
|---|---|
| Recover an abandoned cart | They filled a cart and left. Mail them their own cart back, and catch them when they return. |
| Win back people who stopped coming | They used to visit and then stopped. Mail them, and greet the ones who come back differently. |
| Follow an email up on the site | Mail them, and show something different to the ones who click through. |
| Mail the ones who did not open | Send it again, only to the people the first one missed. |
| Get a discount code into their hands | I have a code. I want the right people to see it, and to know whether it earned anything. |
| Collect leads from the site | Turn readers into addresses I can mail — without a form nobody fills in. |
| Say something to first-time visitors | Tell them what we do, without shouting it at regulars. |
| Prove a campaign is worth running | Not just which version wins — whether running it at all beats running nothing. |
| Let people ask instead of leaving | Somebody has a question my pages answer badly. I would rather they asked than gave up. |
| Put an ask box on your pages | People cannot find things. Give them somewhere to type the question. |
How it goes
- Pick the outcome. Each one states the ask it answers and why it is built that way.
- Answer the questions. Which design, what it says, who it is for — described in your own words
or built as rules. Leave the audience empty to target everybody. Every step shows the rules it will use before anything
exists. - Name it, and press create. Nothing is created until you press the button.
- Finish the campaigns. The program screen says what was created and what is left to do, and links
each campaign into the ordinary editor.
What it actually makes
Ordinary campaigns. “Everything above is a normal campaign. Nothing here owns them.” There is no
program campaign type, no program action type, and nothing different happens at serving time. What the program
contributes is the wiring — and a tag that groups them.
That last point has a consequence worth knowing: because a program is a tag, you can rename or delete that
tag like any other, and what remains is simply the campaigns. Nothing a program makes is undeletable.
The wiring, which is the point
Carrying a person from one step to the next is where hand-built multi-campaign setups go wrong. The program writes
these rules for you:
- Email to site. The email’s outbound links are tagged for the program, and the site campaign
targets people who arrived carrying that tag. It works across sessions and needs no identity resolution. - Email to email. The second send targets — or excludes — the audience of the first, by
what happened to it: delivered, opened, clicked. - Site to site. Within one visit, the second step keys off the first having matched.
rule. On-site rules see the current session and on-site behaviour only — they cannot see email at all. The
tagged link is the way, and it is what the program uses. A hand-built version of this targets nobody while looking
perfectly configured.
Testing what you built
Steps that support it are created with variant groups already split, so the program is an A/B test from the first
day. Where the split leaves a slice untouched, that slice is your holdout — the baseline that
answers “did running this beat running nothing?”. Give every visitor a variant and the screen says so
plainly: “No holdout — nothing to prove the campaign beat running nothing.”
Prove a campaign is worth running exists for exactly this, if what you want to measure is the campaign
rather than the wording.
If it stops part way
Campaigns are created one at a time, because step two’s targeting has to name step one — and a batch
cannot reference what it is still inserting. If something fails midway, what was created stays created and the screen
tells you which parts exist rather than reporting a clean failure that would be untrue about your account. Finish it in
the editor, or delete it and run the program again.
Related
- A/B testing — splits, control groups and reading a verdict.
- Tags — what a program is stored as.
- Emails sent by Personyze — the email half of most
programs. - Testing & QA — a program arrives in staging; this is how you publish it
safely.