Brain Vibe

marketing muses to stay engaged

Marketing Analytics: Why Trust the Numbers?

marketing analyticsSsshhh.  Come here.  Let me tell you a secret.  Those numbers you just presented.  The presentation you spent hours on.  The meeting that immediately after you got kudos for.  It was crap.

How do you know?  Remember that presentation you gave showing the incredible lift on sales by the optimizations made in your messaging strategy?  Turns out, there was an article that provided data conflicting with your methodology and results.

The standard slice and dice of numbers to prove the ROI of ad and marketing spend doesn’t cut it anymore.  There are plenty of others out there that are ready to pounce on your methodology and offer a better way of doing things.  They’ve found a way to take a complex methodology and turn it into a turn key solution or service that makes it easier to track and positioned as more accurate. The pitch: You have results instantly and can act on them with new messaging, offers, creative and strategy.  Oh, and by the way – check out that lift!  Even if you used an advanced analytic technique, someone has a better one to sell.

What should you trust?  What should your managers and executives trust?  Well, that’s your job – create credibility in the numbers and be prepared to defend the results.

Sorry, there is no silver bullet in any testing methodology or analytic modeling.  Each serves a purpose and each mitigates one issue better than another.  No manager or executive wants to be burdened with the fine points of the statistical model you use.  Less is more after all, and they only care about results.  The answer to the secret is how you position your results and anticipate questions that may be brought up during the presentation or after as it has had time to soak in.  Assume stakeholders will look to better inform themselves after your presentation.  They want multiple data points pointing in the same direction to confirm or disprove your recommendation.

Regardless of whether the stakeholders understand a test or modeling method, you should.  The strengths help you tell the story of results.  The weaknesses are where the astute will hone in.  For instance, if you use a sampling methodology, be prepared to present your insight in a manner that underscores not just the the positive impact noticed.  Your listeners will always look for the tarnish in your insights and you need to show why sampling was still accurate and the effect of selection was not a contributing factor to a change in results if ignored.  This doesn’t have to be in the presentation, but you need to be prepared to put the question to bed.

The other aspect is presenting enough of the information.  Incomplete data is a deal breaker.  In attempts to simplify, there is often the risk of over simplifying.  Again, the balance is providing enough in the presentation to satisfy the obvious needs for information and having the details in your back pocket.  First, this creates the the right focus for discussion.  Second, being up on the details makes  you look prepared and informed.

In the end, it isn’t just the insight you bring to the table.  It is your expertise in presenting your findings and creating confidence that is sustainable beyond the opinions and guidance of others.  Leadership in analytics is as much about the obvious insights as it is in covering your bases.

If you trust the numbers, make your stakeholders as well.


Filed under: business analytics, marketing/advertising, metrics, , , , , ,

Marketing Measurement Primer: Creating Your Reporting System

So, you are thinking about creating an automated reporting system to monitor marketing performance across all your activities: direct, website, internet marketing, social media, e-commerce.  There are all these great tools out there with shiny dashboards and promises of business insight to drive your decisions in one place.  STOP!

It is not as simple as recreating your Excel spreadsheets in a database and copying the graphs to Powerpoint.  Forget what your analysts tell you.  The data isn’t the most important thing in this process.  Your business rules governing your data is the most important.  The business rules stem from your process, your plan, and the point of view of your business.  Make sure your establish your marketing plans and business rules as the cornerstone to your reporting needs BEFORE you go right to the data.

Most business intelligence projects are emerging from mature processes and the topmost functional areas that have application systems capturing and storing data as it happens.  There is a consistency around data elements.  For instance, finance, sales/marketing, customer service all work off customer data with sales transactions and relationship inter-actions.  The reason is that the application manages the data.  Where this breaks down is in the new world of cloud computing and SaaS solutions managing micro components of your business.

Your headaches in understanding marketing performance stems from this increasingly disparate view of your activities.  Traditional marketing is housed in your CRM system (ex. Siebel, SAP).  Website marketing is house in your web analytic suite (ex. Omniture, Webtrends).  E-Commerce might be managed with your E-Commerce provider.  Interactive marketing may be managed by your ad service (ex. MediaPlex, DART) or Google.  Each of these services have their own discrete view of the world and provide you with canned reports.  First and foremost all these various systems need to be bring data into a central repository before any consolidated reporting can happen.  This is where most reporting projects begin and move right to the reporting design and implementation.

Be careful.  This is the pitfall.

What makes the reporting in these disparate systems work is the business rules and processes that generate the data.  How a campaign is set up and tracked in one system can be very different than the other.  Cost management will vary by activity as well as marketing cost may be flat fees or operate through exchange systems with different billing methods (ex: by time, conversion, packaging). You want your campaign plans enforcing the management and reporting of activity across all marketing channels.  If you have not created this framework, you risk accuracy and completeness as feeds can have issues in delivery or have conflicting data that needs to be reconciled and/or adjusted prior to reporting.

As you begin your marketing performance reporting project, bring your sample reports and spreadsheets to the table during business analysis.  But, you also want to be sure you bring your marketing plans and business rules to the table as well.  This will ensure accruacy of reporting and mitigate issues in design, development, and implementation of your new dashboards.

Filed under: business analytics, business intelligence, marketing technology, , , , ,

Social Media and Website Engagement as Business Outcome or KPI?

What is it that we really want to know when we are measuring social media engagement?  It can be an indicator of advocacy, brand affinity, purchase consideration, or actual sales.  In many cases, engagement is considered the outcome showing the value of brand.  The problem that arises in this is that all to often how it is measured has nothing to do with how the value of the brand translates into customer value or initial purchase.

The first problem is that measuring engagement often has more to do with the amount of time spent on site and the amount of view, clicks, and level of content reached.  On the surface, this is a great first step.  When looking closely, flaws abound.  The reason, what is the online experience trying to achieve?  If the purpose is a landing area that drives purchase conversions, then more time on the site and an increase in pathing may actually be an indication of less qualified visitation. If the purpose is education and a first step into creating a customer relationship, then more time on the site, activity, and depth of knowledge seeking can be a good thing,  However, to be realistic, have you looked at your SEO and SEM statistics lately?  My guess is that around 70% of those visiting your site are direct visitors or searching on branded keywords.  That being the case, visitors already know a good deal about you prior to coming to your site and the more time and research they do might also not be a good thing if they are comparison shopping.

Sounds a bit dire, right?

To counter this, web analysts are starting to take a look at measuring actions as they relate to conversion.  Simply spending time on the site, views, or measuring clicks isn’t considered viable and predictive.  However, if desired actions are achieved such as downloading high value content, sharing content, participating in discussions, or taking actions that are highly linked and indicative of purchase behavior, then tracking at this level is more valuable.  Actions can be more connected and aligned to desired results and predict conversion.  Right?  Maybe.  The issue arises of clearly understanding actions that predict conversion to sales or customer value.  The other issue is that measuring the number of actions also isn’t that far off from measuring page views, clicks, and time spent.  It might be more meaningful in that it is a validated initiative, but again, is more actions a good thing?  Once again, as with traditional metrics, at the end of the day, what is your site or landing area intended to do?

Measuring engagement should actually take into account both methods for a hybrid approach.  How this hybrid is determined once again depends on what your desired outcome for the website or online experience should be.  At the simplest level, starting with actions taken and tracked as the foundation of a predictive model is a more sound approach.  These are steps in a desired process for conversion, regardless of what your conversion intent is, that are reliable and accurately measured.  However, actions are part of a process and thus need to be ordered and weighted accordingly.  Processes are relatively linear in fashion and assigning a weight based on the step in the path is important.  It can be a simple distribution or multiplicative, but a step does have relevance and weight.  In a hybrid approach, we also want to introduce the traditional aspects of views, time spent, and clicks.  Starting with views and time spent, leveraging these as coefficients in the model will provide a better perspective on weight on desired actions and ultimately the desired outcome.  Essentially, views act as impressions that influence behavior and time spent introduces the amount of exposure necessary to trigger a desired result.  Taking from online display advertising effectiveness, banner ads as an influencing factor for awareness and conversion increases with exposure even if no action is taken to click through.

That leaves clicks.  This traditional metric introduces a duplicity element that needs reconciliation.  It is important to carefully introduce this measure into the model as it can inflate engagement metrics and thus over forecast results.  Clicks also can be an issue as it is typically a component of measuring effectiveness of ad spend.  What needs to be determined is if the click is associated to intended actions taken on site and avoid double counting or inaccurately measuring ROAS.

Ultimately, engagement is an indicator of a desired outcome and not the outcome.  Combining traditional site tracking methods to weight and adjust models predicated on process actions will create a more accurate predictor of outcomes.

Filed under: business analytics, business intelligence, metrics, social media, social media marketing, , , , , ,



Bookmark and Share

Blog Archive