I have a smart list of 10,000 leads and would like to batch these in groups of 200, and send every two days rather than sending them all at once.
Using the random sample option means sending it to batches of 2% at a time to start, resulting in close to 30 random sample send steps. Are there any alternative strategies to stagger the send?
Sorry to be glib, but what's the question?
Hi Sanford,
I'm hoping to find an effective method to batch send large lists.
An alternative to using the random sample step.
Yes, I see you added in the question now!
The fundamental problem with Random Sample is that it's inexact. Other than that it's the only effective built-in method.
The alternative is to call a webhook to get exact-size batches (static lists).
Thanks for clarifying!
Hi Ashani
I would also be reluctant to use Random Sample in this way since as you go over multiple flow steps, it's 2% of the remainder.
Eg. if 100 start the campaign.
Step 1: if random sample is 2%, add to list A. 2 people get added.
Step 2: if random sample is 2%, add to list B. Now, 1.96 people get added.
You can see how this starts to mess with things as you get 30 steps. And you will never reach the full 100%. If you put all the percentages as choice steps within one flow step, you're all good though.
There isn't really a great way to do this. It's really a Marketo flaw in functionality.
Another option is to use an A/B test with, say, 10 variations, all with different send times. But the best you could get to there is 10% splits. And I'm unsure how many A/B testing variants Marketo will let you use.
Thanks for the recommendation, Phillip! Agreed- I would increase the percentage as we continue to run out the full list, but as you mentioned this results in too many steps.
I'm also considering setting up a smart campaign with one 2% 'random sample send' step, and then scheduling this as a recurring campaign to run every 2 days. Of course, i'd use a "was sent email -> do nothing" rule within the step to ensure leads don't get sent this twice. And I'd have to work out how long the campaign would need to run until the full list is exhausted.
Definitely would not recommend this method.
The Random Sample is only guaranteed to act as a Random Shuffle -- i.e. non-duplicating sort into rounded % buckets -- if it's executed using one flow step + choices, in a single campaign run. Rather than sending w/Random Sample, you use the RS to sort into 50 static lists.
I'd personally go with the webhook to get true 200-person buckets. But I'm biased as there are many technical tools at my disposal...