Managing digital dataWhen digital data is unleashed on an organisation for all to see, it can easily result in a cycle of never-ending questions and confusion. Avoiding this and putting yourself in the position of being able to draw practical conclusions from the myriad data available to you is dependent upon taking a few practical steps.
EducationWhatever industry or sector you operate within, there will be an established vocabulary that everyone is expected to understand and communicate in. I’ll wager that in many businesses, this vocabulary doesn’t include digital marketing or e-commerce, except in isolated pockets. Like it or not, education is a big part of the role of a digital marketer or e-commerce professional. Your job is to educate your colleagues and seniors in a way that they will understand. This needs to start early, else digital data that is circulated is subject to misinterpretation, and this is hard to correct retrospectively.
Define and communicate digital terms in clear and simple terms at the get-go. Evangelise in a non-nerdy way.
What mattersIf everything is important, then nothing is important. It’s not practical to focus on everything all of the time. Here’s a short list of web stats that you could look at. Visitors / Sessions / Pages per Visit / Time on Site / Device Operating System / Screen size / Bounce Rate / Browser / Browser Version / Country / Region / City / Hour / Day / Day of Week / Page Views / Unique Page Views / Revenue / Transactions / Average Order Value / Funnel Steps and Completion Rates / Errors And so it goes on. What really matters to you? What do you really want to know, and will really act on? Most of your web analytics will only come into play when you’re looking at what you need to optimise.
Identify what really matters, and use the rest as diagnostic aids.
AutomationMost tools allow you to automate the extraction of data, and it’s worth doing this, whether this is periodic KPI reporting or extraction for crunching elsewhere. For example, if your chosen tool for number crunching is Excel, it is absolutely worth investing the time and (small) sum of money plugging Google Analytics into Excel directly. Whatever you report from, and in, automation will save a lot of un-necessary work and potential for error.
Schedule extraction and reporting wherever possible, as people will know to expect it, in turn pre-empting questions.
FrequencyThe temptation is to get data on a daily basis, because you can. No. Just no. There is little that can or should be measured daily unless your budgets are significant. For Media spend, absolutely report and adjust this daily, but not your broader analytics. Why not report daily? Are you equipped to react each and every day? Is an individual day measurable in a meaningful way (this relates to Trends)? Many things can impact at a daily level – Weather, Public Holidays, National Events, Sporting Events, Major news events. The list goes on.
Determine a frequency that removes or significantly reduces the impact of short-term anomalies.
TrendsWeb Analytics is more about trends – shifts in behaviour – much less than immediate measures. Measuring steps within an e-commerce funnel on a daily basis is unlikely to be worthwhile unless you’ve just made changes to your funnel. Evaluate periods of time relative to others, ensuring that wherever possible you’re comparing apples with apples.
Watch the tide change rather than focussing on the individual waves.
Choosing the right tool(s) for data collection, analysis and presentation is critical. It may be that one tool can fulfil all of these functions, but it’s unlikely.
Equally important is finding a tool that will allow you to supplement your digital data with data from other sources, such as telephony, back-office systems, field sales, and call centre sales etc.
Have you ever attempted to produce something beautiful from Google Analytics? Did it work?? Maybe, but a tool such as Google Data Studio is designed for the job at hand, and will turn you into a data presentation ninja.
Don’t use a hammer to put a nail in.
Digital measures are rarely, if ever, an exact representation of what actually happened. By this, I mean they rarely record, with 100% accuracy, exactly what happened, all of the time. They can’t. So many technological and behavioural variables come into play meaning that digital data always, always, has a margin of error.
Using your web analytics to produce definitive business MI runs a real risk of under-reporting, and failing to align with what Finance see.
Until all concerned are aware of this unpleasant fact, madness lies ahead. Every business should allow a margin of error in it’s web analytics, though I can almost guarantee that this will bring the FD and/ or CEO out in hives!
Funnel analysis should focus on absolute numbers at the top and bottom, and success/ failure rate ratios in-between.
Encourage everyone to relax a little, and accept that 95% accuracy is healthy and normal.
Web Analytics and broader digital analytics is a deep and potentially mysterious realm. Mis-understanding/ mis-reporting digital data is an absolute breeze. A tool such as Google Analytics is free and has a graphical interface that a 4 year-old could use. As such the wrong people can and will use it to produce reports, taking 2 and 2 and making 3.
Recruit a Web Analyst and give them sole access to report your digital data, so they build and explain everything. Then professionally love and nurture them.
Few digital measures in isolation are particularly useful or actionable (I still hate that word). For example a Bounce Rate tells you almost nothing unless you also look at the traffic being driven into the site, or page(s) and the purpose of the page(s).
If you’ve just launched a heavy-weight acquisition campaign, you can expect your bounce rate to rise as the targeting is likely to increase greater proportions of less relevant traffic (much as you’ll work to ensure that it doesn’t).
Digital data is part of the bigger picture and shouldn’t be used in isolation from other business metrics. You wouldn’t focus just on the oil warning light when driving a car, and ignoring the coolant level, and engine temperature etc.
Use digital data as part of the answer, and not the answer.
There’s no substitute for experience. With experience you’ll know when things just don’t smell right. Trust your nose, and go hunting when your sixth sense kicks in.
A high profile and well respected data scientist – Avinash Kaushik – has a blog entitled Occams Razor. Occam’s Razor (also known as Ockham’s Razor) is a principle that simplistically boils down to “Keep It Simple”. This is a good mantra to hold onto.