Analytics and Conversion

5 Tips to Boost the Quality of Your Data

Data-driven marketing. Doesn’t it sound great? You’ve got more data than you can actually use, and more ideas than one could think of. However, before moving forward, one has to make sure they’re looking at the right data. Wrong data often leads to wrong decision making.

If your data collection is faulty, what you see could be a lie!

For instance, this bounce rate graph either tells you the new redesign is performing a lot better than the previous one, or the redesign added a double tracking resulting in a cancelled bounce rate. Which do you think is more accurate?

Data quality is critical when it comes to data-driven marketing. As a marketer, or a computer scientist, one of your responsibilities is to improve your current data quality. You may never reach perfect data, but perfection is not an option here. “Success is a journey, not a destination”. With data, it can be a long journey. Here how it should look:

Massive data quality improvements can be achieved with little effort. Here are 5 tips you can use to avoid common pitfalls.

1. Grow Analytics Maturity

I picture maturity as a ladder; an endless ladder. No matter where you are on the ladder, the goal is to climb one step at a time. Some bars will be missing and some will be broken, but that’s alright.

Analytics maturity is not a one man job; you will have to involve everyone around you including your superiors. To do that, you must believe in it and convince everyone else that data is indeed important.

In fact, it’s more than important, it’s valuable. Data IS value. Jean-Paul Isson (Global VP Predictive Analytics & BI, Monster) said the three-Vs of Big Data lack a fourth: Value.

Data is your gold. The cleaner and more insightful data you have, the more value you’ll get. If it is your responsibility to generate data, then everyone should understand you’re creating value for your company.

One way to adopt this level of maturity is to befriend your IT colleagues if you’re a marketer and vice-versa. A regular data quality audit (let’s say, quarterly) is a good idea to reach this friendship. An audit will include both parties and you will receive good assessment on your current data weaknesses throughout the process.

2. Get a Tag Manager System (TMS) for governing your data collection

Should data be owned and governed by Marketing or by IT? When it comes to data marketing, my guess is that it should be governed by Marketing.

A tag manager is a must for Marketing in order to gain agility and good quality of data which is created from marketing tags, on site and on mobile apps. To learn more, read our previous blog post about that specific topic.

3. Conversion Deduplication

Some TMS vendors offers more than just tag management, they act like a remote control. For instance, you could use your TMS to change your site content without going through a CMS or webmaster. Same thing goes for applications; you do not have to go through the whole validation process.

As for conversion pages, instead of triggering tags for all vendors, the TMS will simply choose one tag based on your attribution model. This helps you avoid getting one conversion attributed to more than one vendor, which could easily happen if a single visitor uses more than one channel before converting.

4. Use a QA or UAT Environment

Let people use QA or UAT (User Acceptance Testing) environments in order to test before going live. Of course, having a staging environment is important, but having marketers access it is even more important. It is fundamental because after all, to err is human, and with QA or UAT one can reduce the chance of making errors.

5. Custom Tag Your Traffic Sources

For site-centric data as in Google Analytics, it is very important to understand acquisition. To do so, make sure all your marketing channels use tagged URLs. Some organic channels are automatically detected and categorized (such as direct, referrals and natural search), but all your paid and CRM marketing efforts (social, email, paid search, display, affiliation, etc.) must be tagged accordingly. I recommend applying different levels of granularity to get better insights.

To reiterate, data will never reach perfection because it is tricky by nature. Nonetheless, there is no real need to be precise and accurate with the numbers when you can focus on trends. For example, instead of stressing on the fact that you have 200 visitors this month (which is not exact), you should look at the trends and deduce that there was 100% increase because last month you only had 100 visitors (which is also not exact).

If I leave you with anything, remember this quote from Alexandre Frantz: “Don’t just be wrong. Be consistently wrong”.

comments powered by Disqus