With OfferFit’s AI-powered experimentation, marketers can test hundreds of variables at once — personalizing timing, content variants, and message frequency for every customer — while still delivering messages through Braze. In this episode of Thread the Needle, Ray El Khoury, a Solution Architect at Stitch, walks through how to integrate OfferFit with your Braze instance to unlock smarter, continuous-optimization campaigns, including:
Transcript:
Hi everyone. I'm Ray. I'm a Solution Architect here at Stitch. Today I'm going to show you how to set up an OfferFit campaign that integrates with Braze. If you're doing traditional A/B testing right now, this is going to blow your mind.
So normally you would test one thing at a time, right? Maybe you test subject line A versus subject line B, or you'll test sending at 10:00 AM versus 2:00 PM. With OfferFit, you can test hundreds of variables at the same time for hundreds and thousands of different customers. And how it works is the cool part. OfferFit sits on top of your existing Braze setup. So you are still sending through Braze, but now OfferFit is making individual recommendations for each person.
It looks at all your customer data, figures out exactly what's best for each person, and then tells Braze exactly what to send and when to send it. The whole thing creates this continuous learning loop Braze sends the message, the customer either clicks or doesn't click, engages with the message. That data gets sent back to OfferFit and then OfferFit gets smarter about what to recommend next time.
So there are two ways you can use OfferFit. There's a self-serve version, which is what I'm going to show you here today. And then there's a full service version where OfferFit basically builds custom models for you and it has a bit more custom integration.The self-serve version is perfect if you're looking to optimize things like clicks, opens, conversions, everything you're already getting in Braze. The full service version is more complicated. Stuff like lifetime value retention, acquisition, where you need custom metrics, custom integrations, and custom models.
Instead of just having a control group and a treatment group OfferFit, actually splits your audience into three groups. You've got your main group that's getting the recommendations, the AI based recommendations, and you've got a randomized control group that gets random recommendations. This is so you can see how much better the AI's doing. Then you've got a third group, the holdout group, which gets nothing. That is your baseline. When you're starting out with OfferFit, usually you set up a 10% audience to test with. And then within our audience, you split out 80% to get AI recommendations, 10% for the randomized control, and then 10% for the holdout. And as you grow, you increase that 10% to 15-20% as you scale to make sure that the model's learning as it grows.
And now I'm gonna jump into OfferFit. This is the self-serve version. Come over here to the top left. We're gonna create a new experimenter. We're gonna give our experimenter a name. I'm gonna call this one Stitch Demo, for this video's purposes. Now you're gonna need to connect it to Braze this, or you select Braze as your platform and obviously that's what we're using. So the one thing to note here is that you're gonna be sharing data between OfferFit and Braze. So make sure you've cleared this up with your legal team or any sort of internal processes you have. Next you're gonna need to input your Braze API credentials. We're gonna connect new credentials. And we're gonna connect Braze So for me, I'm gonna be using our Braze URL for this demo. That’s US-03, but you can find your Braze URL by clicking this link.
Next we're gonna enter an API key. This API key has permissions to send campaigns, extract, segment data, templates, and so on. You can find out more information on what you need to set up here, but if you have this information is usually found under your settings, APIs and identifiers, API Keys location. So I've created one for this demo so we can use it going forward. Next, we're gonna need to create an API campaign. So I've gone ahead and done that. Created an API campaign, selected an email as my messaging channel, and then that gives me a message variation id. So I got the campaign ID here, I'm going to input it here, and then my message variation ID, 815.
Now I need to get my app key. So that is located under settings, app settings. I'm going to get, you know, for example, this one to connect to Braze. Now I need to set up a discipline. So I'm gonna use Stitch as an example and then no reply@Stitch.cx. This would be whatever your setup is in Braze. So that concludes, that completes our integration step. That was very simple.
Now that you're connected, you can now select your audience, see how it's pulling all my Braze segments. So this is connected directly to Braze This is everything I have in Braze So let's just select engaged recent users as our segment that I want to import to our OfferFit. Next is we're gonna divide the audience up into the control and OfferFit groups. So I'll stick with 10% and 90%, but you can increase it to whatever you feel comfortable with. But as you get higher, it can slow down learning and reduce leverage for optimization. So for this purpose, I'm gonna stick with 10%. You can even go lower than that if you need.
Next is we're gonna be setting up the fun part, the email variance. These are all the variables we're gonna be testing. So we're gonna set up base templates. We're gonna need one variant per template that we want to test, and you can import these directly from Braze, or if you have HTML files, you can just upload them directly. I'm just gonna import some of the templates we have in our Braze. So you can see I've got a couple of welcome emails. So I'm gonna do just these three templates. And then I'm gonna create variants per template. So you can see here, I can create the subject line as a variant, my hero image, and our CTA. So now I have three different variants for this version.
Before I suggest subject lines, do you have any guidance? So for example, here's where you can share preferences on tones or topics, humor and. Just humor would be the one for this one, let's say. And then here you can give it more instructions on different subject lines. Direct but funny. For this demo's purposes, we're just gonna add a couple of our own, "Welcome to Stitch", and then for CTAs, we're gonna generate a couple here. "Join now" and "Read more." And then if you have any images, you can import images from your media files, media library through Braze, and then you can continue. And then it'll have all these variants that you created per this one template, right? And so you can see how you can create different variants per email template. That is how the Braze Action Bank grows and has all the different tests that it can run. So for this demo’s purposes, I'm just gonna delete these so we can move ahead.
Next is we're gonna be setting up the frequency controls. This is super important 'cause you don't wanna spam people. So let's say I just wanna send one email per week, and I've said that here, right? So you can think about this similar to Braze frequency cap, but more focused on OfferFit. So I'm gonna select two per week since I only wanted to send them twice per week. And I only wanna send from Tuesday, Monday to Friday, not on weekends. So this is where all that's set up.
Next is time controls. We'll choose the time of day to send these emails. So we're gonna start with the time zone. We are in Eastern, so that's the easy one to pick for me. So you can choose the time of day, which OfferFit can send emails. This is again, based in Eastern time, so I would do something around 10 to 8:00 PM but you can configure whatever makes sense for you and your marketing plans.
And then finally, here are the guardrails. This is where you can get really specific about content rules. Maybe the same person shouldn't get the same exact email more than once, or maybe you should make sure subject lines are always different. You can set all that up here so you can, within seven days, you should not receive the base email more than one time, right? Subject line variance sends, and so on. So for the purpose of this demo, I'm just gonna skip ahead and that completes our self-serve campaign setup. Here's where you review everything once more, one more time. Once everything looks good, you just hit go live and OfferFit starts doing its thing.
So basically that's it. The setup here is pretty straightforward. The results can be incredible. OfferFit has seen campaigns improve their performance by 20 to 30% or more. Just by letting the AI figure out the optimal combination of timing, content frequency for each individual person. The key thing here to remember is to start simple — maybe just test timing and subject lines first. Then gradually add more dimensions as you get more comfortable with how it works. And you can always set those guardrails. They're here to protect you and they're here to protect your brand and make sure you're not over communicating with people.
Thank you for watching.
