Joe Robinson: Bald Guy, Marketer, Self-Improver.

Not all “data-driven” approaches are equal. Here’s how to nail it so you can reap the benefits and maximise your first-party data.

Since getting into marketing, every job I’ve ever had has claimed to be “data-driven”. If you apply a broad enough definition, they all were. But like many things in marketing, being data-driven happens on a spectrum.

Is your work really data-driven? Does the data drive your decision making? Or do you just make sure things weren’t a disaster afterwards? It’s not hard to switch from one approach to the other, and the results more than justify the effort.

“Data-driven” vs. “involves data somewhere”

The truth is, lots of marketing teams (or even whole brands) have measurement and tracking in place so they can look at something once a month and know if it’s going okay. If it’s not, they stop. If it is, they keep going. Then they unironically call that data-driven.

I call it the bare minimum. Does that make me cynical?

If you’re only using data after the fact to make sure nothing is on fire, the best you can really call that is “data-reactive”. It’s better than nothing, but it’s never going to get you the impressive results you’re really after.

And with the end of third-party cookies looming, soon it won’t be possible to just throw tactics at the wall and use mountains of user data to decide what worked. The real winners in 2024 and beyond will use their existing data to come up with a clear idea of what they want to happen, test one or two tactics to influence that, and use their first-party data to understand if they succeeded or not. The data-reactive bit is still there, but it’s the last step in a longer process.

Let’s see what that looks like:

Can’t spell “voucher” without “ouch”

Consider your next big promotional activity. What does it look like? Is it a flat discount, based on basket size (e.g. 3-for-2 or “buy 3 items and save 20%”), or a spend threshold (save 15% when you spend £75)? There’s plenty to choose from, and you’ve probably got a rough idea which kinds of promotions work best for your customers.

Let’s consider discounts with a basket size or quantity-based threshold, since there’s a bit more going on.

The data-reactive way of running this campaign is to arbitrarily select a threshold for the discount, run the promotion, and see what happens. As the data comes in, you’ll get a good idea of how things are going.

Working this way will probably drive the sales uplift you’re hoping for without a lot of effort, so it’s easy to stick with it. But there’s a downside to doing things this way. The benefits will probably be short-lived.

You can out-compete a few of your competitors to increase your share of wallet, and incentivise some customers to buy sooner than they would have. But you’re not changing your customers’ overall behaviour. They still place the same orders, usually. And at a lower cost!

Data, take the wheel

So how would doing this in a properly data-driven way improve your end result?

By letting your existing data dictate your discount thresholds.

Let’s say your sitewide AOV is £45, and you’re going to run a 15%-off promotion.
If you set the discount threshold at £50 because it’s a nice round number, that leaves £42.50 in the basket after discounts.
You’ve incentivised the sale, but you’ve also decreased your sitewide AOV with every sale that comes in! No wonder your finance team don’t see the value in marketing.

Sure, some people will spend a little more. But we’ve all done it; you’re £1.50 away from triggering a discount so you trawl the entire website for something that’s 2 quid just so you can get a deal. On average, a deal like this will hurt your AOV and your bottom line.

What’s the alternative?

With a sitewide AOV of £45, you could offer £15% off orders over £60, which gets users £60 worth of stuff for £51.
An extra £15 worth of value for £6 out of pocket? It’s a no-brainer. You get 33% more stuff for 13% more money!
Now, your customers will actively spend more on your products and feel like they’re getting a deal in the process.

And when their delivery arrives full of more things (or better things) than they’re used to, they’ll like that. Some of them will like it enough to make this their new standard order. That’s a healthy bump in lifetime spending thanks to a clever promotion!

Change the norm

You don’t have to restrict this thinking to your promo activity, either.
Play around with your free shipping threshold. Can you raise it slightly for a step-change in AOV?
Remember what we said about trawling the website to trigger a discount? It’s not always a bad thing. If you’re only giving away the cost of shipping to get more value from a customer every time they shop, that’s usually well worthwhile.

This kind of thinking isn’t limited to promotions either; the point is the make your data a planning tool and not just a reporting tool.
I put my foot in my mouth recently by not being data-driven and wasted some time in the process.
I set up a fairly standard split-test within some ads I was running, ready to come back to in 3 or 4 weeks.
When I was ready to view the results, the results weren’t ready for me to view them- my ads hadn’t received nearly enough clicks to draw a conclusion.

If I’d looked at how much traffic these ads were getting before my test, I could have known that without allocating a lot more budget, this was bound to happen.
In the end I re-ran the test using some ads which were getting adequate volume in the first place, and after 4 weeks I could draw a reliable conclusion.

By fucking it up the first time and fixing the second attempt, I was data-reactive. If I’d planned better and nailed it first time, that would have been data-driven.

It’s just a fancy way of saying “planning ahead”

Unless your business is brand new, you’ll have some data on the time you’ve been trading until now.
With it, you can understand what to expect from your products, your customers, your ads- you name it.
You can skip mistakes, which lets you rack up way more wins and far fewer losses.

And the best part? It’s just planning ahead.
It’s just using that data to graduate from guessing, to educated guessing, to proper forecasting, so your assumptions are right more of the time.

When you look back in 12 months time and see how well everything went because you nailed the planning and execution?
That’s the kind of data-reactiveness I’m all for.