Insanity is doing the same thing over and over again and expecting a different result

– Albert Einstein, Benjamin Franklin, Mark Twain or Rita Mae

Have you ever wondered “why is every marketing campaign in my company executed like it is the first time we are doing this?”

“Why are we not learning from the last year’s Black Friday campaign?”

“Shouldn’t we already know which channel drove most value and focus on that already?”

You are not alone. This frustration with a lack of learning is fundamental, ubiquitous and universal.

It is fundamental in the biological sense that you are wired to detect progress – “am I improving or not.” Anything short of progress causes dissatisfaction with life. This is why many bestselling self help books advise examining your life in seven to ten categories, and making goals for each of them, so that you can demonstrate progress to your brain.

It is ubiquitous. It is not just the Marketing team that suffer from a lack of learning. This malady affects Engineering, HR and Finance teams as well. If the Engineering team is repeatedly missing deadlines or pulling all nighters, then, you know their planning was repeatedly erroneous. If the HR team repeatedly takes too long to make a hiring decision and loses perfect candidates as a result, you know they have an issue. If the Finance team puts in travel restrictions in the final quarter to meet the annual budget every year, you know they too have a learning problem.

It is universal. It plagues big companies like Google, Apple and Microsoft because learning is not a priority. It plagues small companies because they are understaffed.

So, what can one do? What can one person do? What can you do?

Three things: Champion. Understand. Ask.

Every single thing that happens in any company is because there is somebody who wants it done. If you want learning driven marketing done, then, you need to champion it. Championing this is actually very easy. You share your dream for your team with all the stakeholders – that you want to double the effectiveness of the campaign with the same budget as previous year’s; or, execute the campaigns in half the time; or, do twice as many experiments to double the learning. When you demand your stakeholders to expect more from the team, you have influence. You are championing effectively.

Then, ask the right questions. It is surprising how asking the question reveals the path towards the answer. Suppose you are championing doubling the effectiveness of the campaign with the same budget as previous year’s. You spent money in the previous year on Facebook, Youtube, Twitter and Snapchat. You may guess that two of these channels might not have been worth the money spent. But how do you know which? Well, just ask.

Ask who? That is where the third component comes in – Understand how the marketing technology works, and who can do what.

So, there you have it – the framework for implementing a marketing culture based on learning – champion, understand and ask.

A Practical Demonstration

I still remember studying physics in eleventh and twelfth grade. The teacher would talk about theory for a few minutes. It made some sense. And then we would dive into a practical problem. I would be completely lost. I never could apply the theory to the problem the very first time. Other kids would raise their hands and offer a solution. The teacher would praise them and demonstrate the answer for the rest of the class. After a couple rounds of doing this, I would get the hang of it. Then I too could solve the problems.

But I never could answer it the first couple of times for two whole years and it was frustrating. I always wondered how the other kids always knew. Many years later, I met one of the students who always knew the answer. He casually mentioned that their school started a month ahead of our school. Of course!

Keeping a reader like me in mind, I will dive into five examples in this article to demonstrate how to unlock value from data from previous year’s campaigns. These examples will examine the role played by different channels, audience targeting, ad creatives, LTV (Long Term Value) and muti-touch-attribution.

Scenario 1: Measure Effectiveness Of Different Channels In The Previous Year

Let us examine the scenario in which the previous year, you advertised your mobile app on Facebook, Youtube, Google Search, twitter and email. And you want to know which was the best value for money, so that you can prioritize your spending on that.

Insight 1: Every user has a unique ID: This may seem obvious to you, but when you think about it a little bit deeper, you will see a wealth of possibilities open up.

Every time your browser requests a page, or the mobile app connects to the server, it reveals its identity to the server. The server keeps a record of every request it gets and stores it for later analysis. This is how Big Data gets generated.

Big Data analyst can extract all the records related to one particular user and tell you how much money they made for you. All they need is the log data. And where the user came from (after seeing which ad campaign).

Insight 2: Your browser is obliged to report to the server where it came from: This means that the server not only knows your ID but also where you came from. Suppose you were on a Google search result page after searching for some keywords. The server now knows those keywords as well and has it recorded for the Big Data Analyst.

Insight 3: The URL can usually be extended by adding key value pairs while pointing to the same destination. The URL looks like https://example.com/uri?u=bcbgmax&t=2018-01-21&loc=90210 … if you observe the url closely, you will notice that it has a number of key=value statements stitched together with ‘&’. For e.g. u=bcbgmax and loc=90210. If you wanted to add src=twitter, you could add it by sticking in an ‘&’. So your new URL would look like https://example.com/uri?u=bcbgmax&t=2018-01-21&loc=90210&src=twitter … Most of the time, the extra addition of key value pair will be harmless and the URL will continue to deliver the same resource. You can use this to your advantage by having a different URL for each campaign.

Insight 4: Answering your question should take less than 1 hour for a competent Big Data Analyst. This is important, because you now know you are not making a big request that needs huge efforts. You are just asking for less than an hour’s time of a self respecting big data analyst. It is hard for someone to say no to spending an hour when the upside is you can save thousands of dollars.

What the Big Data Analyst will need and do: You need to provide the Big Data Analyst with the click URL that was used in marketing campaigns. And the timeframe of the marketing campaign. And how to measure the value of the user (it could just be total time spent on your site, or total money spent). Hopefully, you used different URLs for each campaign. But if not, they will need a list of places where you advertised.

Once the Big Data Analyst has all this, they will run a query and make two lists. One will contain a list of all the users and how much value do they have. Another will contain a list of all the users and the URL from where they came. They will merge these two and give you a file that will have 3 columns – userid, url and value. That is it. Their job is done.

What you will do next: You already know how much money you spent on each campaign. You can use your spreadsheet-chops to extract the channel from the url, find out how many users each channel drove and how much money each source generated in less than an hour. You are now empowered to optimize your spend with just two hours of actual work.

To Do It Again: The best thing is, that doing it again should be very quick. The Big Data Analyst might even set you up with a self serve tool where you put in the time frame and the URL information and get the data yourself. Ask for such a self serve tool. They will be happy to provide you with one.

Caveats: You still don’t know whether some of the users would have any way ended up using your app without you spending money to get them there. For that you would have needed to implement hold out groups. No one can change the past, so, right now, this is the best you can do under the circumstances.

Scenario 2: Measure Effectiveness Of Audience Targeting In The Previous Year

Let us examine the scenario in which the previous year, you had created multiple audience segments on Facebook, and ran different campaigns against them. You want to know which were the best segments based on long term value (not just immediate clicks or installs), so that you can save the creative budget.

Insight 5: Every marketing campaign analysis begins similarly. Imagine that! Just as in the previous scenario, you need to ask a big data analyst for userid – url – value. Then, you need to do a similar analysis as before, this time, extracting the campaign (and thereby the audience target) instead of channel. This is why it is so important to have a unique URL for each call to action, even if it is the same action.

Insight 6: Every audience has a certain reach on Facebook. The target audience is list of people who will see your post primarily. Reach is a much bigger number – it also includes the number of people who will potentially see your post as a secondary effect. Once you know your target audience, go to Facebook, and find out the total reach of that segment. From the previous analysis, you know the top segments. Now, you need to start building your target audience by adding the most effective segments into it, until the total reach numbers meet your campaign goals.

Scenario 3: Measure Effectiveness Of Different Creatives In The Previous Year

Let us examine the scenario in which the previous year, you had created multiple creatives on Facebook, and you want to know which creative had the best success, so that you can prioritize your spending on that.

Insight 7: Clicks without engagement means broken promise. With all this focus on data, it is easy to forget that an advertisement is just a promise – a promise that if you click on the link, you will get what was said in the advertisement title and description. If the ad the best app in its class, but the app crashed the first time it was installed in the user’s device, guess what? the promise made by the ad was broken. You can detect this in your own campaigns by measuring the ratio of clicks to average campaign value.

From previous scenarios, you already know how valuable each user is and which URL they came from. Compute the average value for each creative, and take a hard look at the creatives with low values. They are the ones that broke promises.

Insight 8: View without click means irrelevance: The other list of bad creatives comes from the campaigns with very low clicks. They simply did not connect with the target audience. Either the targeting was bad, or the creative was bad.

Scenario 4: Measure Long Term Value Of Audience Acquired By Each Campaign In The Previous Year

Let us examine the scenario in which the previous year, you advertised your mobile app on Facebook, and you want to know what was the long term value (LTV) of the users that those campaigns acquired.

Insight 8: Long term value is fuzzy: This may seem like a let down, but the reality is, there is no species called ‘average user’. No two people are the same, and no two moments are the same. How the users behaved in the last year (your app might have been #1) is no prediction of how they will behave this year (may be there is a new competing app out there this year).

Since LTV is not an exact science, it makes sense to approach the problem from a practical point of view – how much am I willing to spend to acquire one user? I share here my approach to calculating long term value.

My approach to LTV focuses on three choices.

Choice 1: How long is the long term? The technology moves so fast that when the user switches phones, they will probably not install your app. Phones have two year contracts – so, a user on an average will be with you for one year. So, I would choose one year as the long term in my LTV calculations.

Choice 2: What is the value? This is never straightforward. Do you pick total revenue made in the last year? What if you paid some money to acquire the traffic? Do you consider the salaries paid to employees and factor in the cost of hardware? My favorite is “average margin per user” (AMPU). It is simply revenue minus costs divided by total users. The rationale is “cost is proportional to users; revenue is proportional to users; when there are a large number of users already, then, profit (or loss) will be proportional to users”.

Choice 3: What is the retention rate? A fundamental law of manufacturing is that things either fail early or last really long. Have you ever bought a new computer, only to have its hard disk crash in the first week? Similarly, a lot of users drop off in the first week. They click on an ad, download the app, open it for a minute, get interrupted, get busy and forget they ever had the app. Since we are interested in the LTV where our long term is one year, we need to find out how many users still use our app one year later. So, assuming the data are available, pick a day a year ago, and identify all the users. Then, calculate which of those users were still using the app in the last month. That ratio is the retention rate. It will typically be a fraction – like 1%.

LTV Formula: LTV = AMPU x Retention Rate

That is it. It is a simple formula that makes sense.

Armed with LTV, you can now proceed to measure the campaign effectiveness in terms of the LTV. While you cannot calculate AMPU for individual users acquired from each campaign, you can definitely measure the retention rate from each. Every campaign will bring different quality of users. A good campaign has a higher retention rate than the average retention rate.

Scenario 5: Measure Multi Touch Attribution For Each Campaign In The Previous Year

Let us examine the scenario in which the previous year, you advertised your mobile app on multiple channels multiple times, and you want to know what was the impact of each time you touched the user with an ad.

Let me begin the discussion of Multi Touch Attribution (MTA) with an anecdote.

Mr. Kuber was an extreme miser with a great analytical mind. He was consumed with a get-rich-quick-scheme as he ate four slices of bread. The first three slices of bread were totally useless – they did not fill him at all – he still felt hungry after eating them. But the fourth slice had something magical about it – eating the fourth one satisfied his hunger. If only he could identify the fourth slice correctly, he could save 75% of costs and not eat those other 3 worthless slices! He conducted many experiments on himself and his employees, that resulted in years of agony and extreme hunger. Finally, he discovered the principles of multi touch attribution, before succumbing to diabetes.

Insight 9: The purpose of MTA is to figure out how much to pay for each touch. The way you touch the user could be through email, search ad, facebook stream or youtube video. There is no point is spending the money first, and then back calculate how much each touch was worth. Instead, the value is at the time of spending  – you may be willing to pay a bit more to show the ad the third time to a user than to spend it on a totally new user. So, the previous year’s data are useless. Unless you got lucky with sequential rollouts. The rest of this scenario deals with how to execute the campaigns this time so that you can have insights the next year.

Insight 10: The discipline to track every ad served is a pre-requisite for MTA. When all the ads have impression and click trackers in them, the MTA Measurement Vendor will be able to provide you with a list of ads seen or clicked by your users before they converted. It gets difficult with walled-gardens such as Facebook and Google who refuse to honor each other’s trackers.

Week 1: Suppose you have a email newsletter – and you introduce the app in one of the emails to 10% of the audience randomly selected. A few people open them. You have your first touch. In the first week, you can measure how many converted based solely on this single touch.

Week 2: You now introduce the app to 30% of your newsletter audience. You obtain the list of email addresses of openers. You split the list into three. One list is uploaded to Facebook custom audience, another to Youtube custom audience and the third to both. You target each group with appropriate ads, and track the conversion. You now have enough information to make a decision on how much it was worth to spend on each campaign.

Week 4: Measure the retention rate for the users acquired by each type of touch. Two weeks is a short time, but many users drop by then. Use all of this information to come up with a strategy on how much you are willing to pay for each campaign.

Final Thoughts

This article has empowered you with everything you need to know about Marketing Technology to get you started with unlocking insights using data from past marketing campaigns. Look back to some of the examples when you get stuck for inspiration.