How To Gut Check Marketing (With Your Campaign Data)

Data analysis is a do-or-die for tomorrow.

From job descriptions to business services, the term data-driven is as hot as any phrase in the professional world. People want the proof in their pudding — and the more quantitative the better — because a clear and calculated recipe is easier to dissect and reproduced. It’s not always obvious how this applies in every business, but for brands in online advertising your data-driven strategy could not be more clear, or important.

Given how much data is captured by each online campaign (and the availability of easy-to-use processing software, simple as Excel), there is little excuse for moving from one campaign to the next without thoroughly analyzing the results, be they good or bad. In another post, I provided some examples of the most common data points that you can’t afford to ignore. Below let’s see what actionable insights means in programmatic media by observing an example of how to read into your impression-based data.

A time of day (TOD) example:

By default, when you set up a new campaign (manually) it will spend 24hrs a day. Do this for two weeks however and you see that clicks and purchases don’t necessarily follow the same steady stream. To optimize accordingly simply map your spend and your goals on the same graph over the time of day on one axis. An hour-by-hour graph should reveal exactly where the impressions you buy are leading to results. Usually more important, notice where your spending does not appear to work towards your goals. This is one definition of ad waste, something Forbes calls the silent killer of corporate profits. Below, a light-blue bar graph shows the percentage of spend by hour while a dark blue line graph represents the successes of the campaign, which can be something like clicks (driving site traffic), sales on a conversion page, or anything else (e-mail sign-ups or downloads of your event brochure for example).

(This is a graph of real campaign data)

When not to go on your gut.

It’s clear from the above graph that this campaign’s audience is basically asleep between midnight and 7am. This may seem logical for most businesses but having the data to back it up is key. Too often assumptions are only that. I caution everyone from putting too many eggs in any basket based on gut feelings about their audience behavior. You might feel you know all sorts of meaningful stuff about your brand’s audience and you probably do but, best practice in any case is to test your assumptions out in the real world. Do a gut check. Having a quantified measure on your assumptions will help you understand exactly how significant they are. It will help you allocate spend (and other resources) down to a science and in more meaningful ways. The most interesting data points you garner may save you money or help you reach a more high value audience.

Your data is strictly causal.

Looking back at the graph, why is our spike in conversions between 2pm and 4pm? Could it be a television ad spot taken out at that time? Or maybe our brand’s service is needed most right at that time of day. Keep in mind that your campaign data is not causal. The data will only spell out patterns. Your campaign cannot tell you directly why something is happening, only that it is happening. Understanding why is up to your company’s campaign experts or data interpreter. The marketing automation behind RTB campaigns allows you to unlock more data, but an expert is really needed to take advantage of the new capability and make sense of the stats. It’s going to be their role to imagine and tell the good story of what is going on in those real human lives of your brand’s audience profiles.

Sample sizes require judgement

Testing never ends. It’s hard to say exactly when you have collected enough data to accurately optimize your campaign. As a rule of thumb, I like to wait a full week before I jump to conclusions on bidding optimizations. Depending on what I am gaging, I also prefer to have a million impressions before I manipulate targeting too much. This is always a judgement call, but I prefer to play it conservatively (with more data than less). For example if looking at the best performing days of the week, maybe you want to wait and see a full month of data before you make big moves to cut-back on Sundays. It’s great when you have a full year’s data and can start tracking the impact of your industry’s seasonality or get a real feel for what strategies are worth fine-tuning.

Move slowly, or use meticulous segmentation.

It’s also important not to launch too many optimizations at one time. Do lots of analyses at once and then only make moves on the biggest patterns for that moment. Multiple optimizations can have opposing effects which cancel each other. It becomes more difficult (if not impossible) to gage the impact of each optimization in the presence of many maneuvers. You could wind up with no clue what did or didn’t work. A careful segmentation strategy is another way to mitigate this risk. Segmentation is something like an A/B test, where you can strike up a mini campaign (also know as a line-item) to test any optimization in it’s own controlled environment.

Summary/Conclusion

The only truly failed experience is one in which you don’t learn from the data or evolve your digital strategy. The available campaign data can help you do both. In programmatic media there are mountains of data (typically ~20 data-points for every impression served) and tons of audience patterns when you start drilling into it. Geographic patterns are just as straight-forward as our TOD example above and there are dozens of other segmenting angles to consider. Every optimization can help you reach your audience better and each calculated adjustment can mean the difference between hitting your campaign goals or wasting precious advertising budget.

From our example which considers TOD, you can imagine how a single optimization could save you upwards of 50% of your budget being wasted! This is the simplest form of ad waste for your brand to avoid. I keep a proprietary list of such go-to considerations handy and run regular reports to check the data for the various approaches to optimization. Imagine if each proven method saves you between 5–50% on ad waste. Hiring the right programmatic expert to sift through data & optimize campaigns will pay for itself many times over while also uncovering interesting audience insights that can inform other marketing efforts. Think about how your data can help define better audience segments or develop more effective messaging strategies.

The plug

I inform digital campaigns strategy professionally at Blanket The Web ← drop me a line if you have any questions. I’m new to writing, and it’s scary stuff. Please like, comment, and share this post to keep me going!