The online world is full of amazing success stories on how fancy analytics tools have turned around a business issue. They all sound cool and suddenly your management or team buys into the concept of having analytics and some real data-driven marketing happening. Now you are facing the dilemma of where to actually start.
Frankly speaking, starting small is often a winning strategy. Here I mean that you should first have some quick wins, something which can serve as a proof of concept. Why? The more time you spend on building the ultimate marketing analytics project the more you put “all your eggs in the same basket”. Basically, you are risking that if that one big project fails you may never be allowed to continue down the path. There are plenty of opportunities you have for the first marketing analytics projects which are easy to set up and have a high chance of success. Once you have proved that you can bring bottom-line results you will certainly get more support from management or just more energy in the team.
Suggestions for the first marketing analytics projects
Some general best practices
- Use data you already have. It is important not to get bogged down with digging data up. With this approach, you are also removing the risk of your project failing for lack of data or data quality issues. If you have at least a Google Analytics account set up, that is already a treasure chest of data.
- Make sure the business questions are clear and your team is aligned with it. It is imperative that your stakeholders understand what business issues you want to solve at first. Make it as narrow as possible so that expectation management becomes easy.
- Be clear on how you will measure success. Connected to the above it is also key to clarify the KPI you are targeting. Is it the cost of acquisition? Could it be the total revenue coming from the webshop? Would the number of acquired customers be good? It does not matter what you pick, but it is important to clearly work towards it and it should be an important KPI for the business. You might want to handle risks during the project, but do not pick a secondary metric as a measure of success, because you will lose the importance of your project and with that the whole point of introducing marketing analytics.
- Try not being dependent on IT. It is difficult to achieve during the first marketing analytics projects, but for speed and actually proving that your concept is future proof it is better if you can leave IT out. The marketing team should be able to manage its own experiments and analytics without heavy IT involvement anyways, so your proof-of-concept projects should be able to showcase that.
The simplest thing you can do is to set up a few A/B tests. You can use Google Optimizer, which is free and simple to set up. If you pick the objectives and areas of test right you can bring in results pretty easily. Just a few things to consider when managing your A/B test:
- Pick something significant to test. I’m not trying to push you to change the entire layout of your website, you can change important things even by modifying some colors. What is important that you do not do these changes on obscure parts of your site. Impact a form or a CTA, something you believe is key to drive conversions. This way you will be able to show results actually impacting the bottom line with your first marketing analytics projects.
- Make sure you have a big enough sample size. Use this calculator to make sure that the difference in results for the versions is significant. Don’t make rash judgments, give enough time to your tests to produce a significant result.
- Plan more than 1 test. It can happen any time that an A/B test is inconclusive or that your challenger (B option) performs worse than the original. It is still a valuable test since you do have learnings, but at the same time, it is more difficult to explain to management. I’m sure it is easy to come up with multiple ideas, so you have a portfolio to ensure that you really ship insights management can understand.
I have written a few pieces already on cohort analysis (simple approach, cohorts for retention) if you want to dig deeper, but basically, it is a tool for assessing a portfolio over time. Some of the most usual questions you can answer:
- What is the retention rate of customers acquired through different channels?
- How fast am I losing customers after acquiring them?
- What is the average purchase (and over time re-purchase) amount of customers coming from different sources?
- How fast the impact of different campaigns flatten out after they have been finished?
You can put down your own questions with 2 elements:
- What is your reference/starting point or rule(s) of segmentation? This could be acquisition channel, the finish of a campaign or time of acquisition.
- You should have questions related to results over time. This is why it is good if you can keep repeating your cohort analysis, so you can keep tabs on your portfolio
Preparing a good cohort analysis requires data structured for it, so this type of assessment might take some time and effort upon starting up. Also, you may need to experiment what kind of cohorts are the most useful for your business, so prepare for re-running your assessment multiple time. Due to this a cohort analysis is probably the most time and resource intensive among the first marketing analytics projects I have introduced here.
The beauty of cohort analysis is that you can give a health check of your portfolio in one chart. That is a pretty powerful tool to show to your management. Once they get the gist of it, they will love it.
Setting up control groups on campaigns
I have seen at many companies that systematically using control groups for campaigns is not in the daily operating routine. Due to this, they are unable to tell if their campaigns are driving extra value above the natural behavior of their customers. This is why the introduction of control groups is one of the best first marketing analytics projects.
Simply introducing control groups to your campaigns will give you very valuable insights. You will be able to challenge the intuition driven assessments that for example X percent of conversion rate is good or being below or above average is what is constituting success. You will be able to get to the point of actually showing incremental results and compare campaigns based on that. In my experience, you will find a lot of the times that some of your seemingly best-performing campaigns are actually not bringing anything extra to the table and some other which seems to have marginal results actually get at least some of the “unmovable” customers or prospects moving.
The next step from this is to actually introduce some real return-on-investment calculation based on purely the uplift of the campaigns. Once you get to this point you can really change the landscape of what activities you are doing. At the same time, this activity will give you a constant credibility.
The biggest challenge I have faced, actually not once, when introducing control groups, is basically the fear of missing out. By not targeting or soliciting some customers you seem to be missing out on some revenue, but I’m sure you can convince your management to look at it more like an investment. Also, as I mentioned before, start small. It also helps if you start with activities which are more widely debated, so you can bring in proof of concept. As you enlarge the set of campaigns with control groups and uncover new insight this challenge will solve itself.
The way forward
Once you have one or more of these projects done, I’m sure you will be on the journey of becoming more data-driven and also you will have more resources at your disposal to maintain your tools and bring in new solutions. I have a few parting advice to help you manage your growth:
- Manage your portfolio. Always have some quick wins coming in, do not disappear into big projects only. You need to keep up the visibility of results and even of all your big bets work out, it is better safe than sorry.
- Reduce maintenance costs. The first time you do basic projects they may be kind of “quick and dirty” and that is all fine. Once they become a usual practice you need to make sure that maintaining them does not require a lot of resources. Believe me, once analytics start working out people and management will always keep asking for more and they tend to forget that now usual practices were quite unusual a while back. Do yourself a favor and invest into streamlining your regular tools and processes to give you room for trying new things.
- Some of these tactics will never get old. What I am saying is that it is best to make sure that your stakeholders understand that it is not always the new tools, algorithms and shiny stuff which brings in the results. Many times it is speed what is the best in old and tried tools. They are reliable, everybody can interpret the results after a while and it is also easy and fast for you to bring insights to the table. So, just don’t forget to use the basics, they will help you a lot.
I’m interested in your thoughts on how to start doing some analytics, so let me know your ideas in the comments!