Quantifying the offline impact, of online marketing, and in turn, your website, is one of the important analytical challenges multi-channel businesses face. This blog post focuses on the multi-channel retail use case.
People shop across channels, but online businesses do not commonly measure marketing return on investment in this way. It’s important not to assume an incremental relationship exists between online and offline. However, discovering either way, is valuable.
This type of insight is often called ‘ROPO’ : Research Online, Purchase Offline. Also sometimes referred to as ‘O2S’ or Online to Store. The ROPO effect is described as a ratio, for example: For every 1 sale online, we create .5 sales offline.
ROPO is one of the ‘big boulders’ of insight, that can be hard to move, for multiple reasons including:
- Siloed data. Inability to connect customer data from online to store, and visa-versa, is commonplace.
- Established return on investment measures and budgets are tough to change, and rightly so!
- A lack of trust in statistical methods, with a healthy skepticism against models that predict scenarios of increased digital spend, sustain inertia.
The implications of this, especially at times of economic uncertainty, are significant. Not knowing what the marginal return on investment of marketing spend beyond the channel, can lead to optimization and investment, that is stifled by a sub-optimal measurement framework.
Measurement as a Competitive Advantage
To give a real-world example, consider a retailer, with e-commerce , and bricks-and-mortar stores.
In this organization, there are many differences of opinion regarding the role of the website in multi-channel, and a conventional wisdom that serves to sustain the status quo, of a siloed approach to marketing measurement.
Traditionally, their primary budgetary spend on digital marketing is in paid search (PPC) advertising, the results of which were measured only for conversions online.
The impact of this, is such that the ROI for PPC was limited to in-channel conversions, based on a limited view of the diminishing marginal return.
Why does this matter? Risk.
The risk to this business is in stifling the right thing to do for all channels, by limiting their measurement model to one channel. For businesses that have not established a multi-channel measure, there are disadvantages. They cannot spend online, to levels that take into account an offline conversion value, meaning that they may lose in advertising bidding auctions against other businesses that have already proven this.
Business A may have quantified, through various experiments and ongoing measurement, what the additional marginal return is offline, for spend online, and adjusted their paid search optimization parameters, and ROI measures, accordingly.
Business B may not have achieved this insight, or been able to prove it with enough confidence to unlock additional investment, and change incumbent ROI frameworks.
In this scenario Business A can profitably grow online market share, whilst benefiting their retail estate. Business B loses out, since they cannot spend beyond their long established curve of diminishing marginal return, which only takes online into account.
What did ‘Business A’ do differently?
What follows is a description of three techniques, that can bring an organization closer to quantifying the offline impact of their online marketing activity.
We say ‘bring closer to’, because measurement is rarely perfect, merely it describes the world in a way that brings you closer to the truth.
These approaches, if explored, can in combination, paint comprehensive picture, as to whether or not there is a relationship between digital spend online, online research behavior and sales in store.
Marketing Mix Modeling (MMM)
This type of study, uses aggregate data, and can be used to quantify the incremental impact of the inputs (e.g. search marketing spend data), to the output (e.g. store footfall or sales). To be successful with this type of modeling, it requires specialized software, or statisticians skilled enough in econometric techniques, and significant volumes of time-series data.
- Historical representative sales traffic and footfall data for online and in store are required
- Consider time lag in the analysis – there may be an effect on the correlation of online activity to in-store.
- Consider seasonality in the analysis – seasonal and promotional peaks will need to be controlled for.
- Articulating the output of such models is important. Influencing skeptical stakeholders will require them to be brought along the journey and have their voice heard.
Known Customer Tracking (using Email as the conduit)
Known customer tracking, in this example, uses the customer’s unique ID, which resides in the CRM system. This is populated in email links and captured in web analytics, then tracked through to sales in all channels.
It is possible to leverage the insights from known customers, via your CRM system. Assuming that a business, has a single customer view ID residing in the CRM system, the ID can be populated in the email links as a parameter. This parameter can be collected by the website analytics to start tracking ‘known visitors’. It is then possible to reconcile, from known visitors online AKA ‘researchers’, to sales offline, by joining the web analytics and offline sales data together.
This gives the raw materials to quantify: from X researchers online, there are Y sales online and Z sales offline. Therefore for every sale online, we achieve Z sales offline (Z can of course be a fractional amount).
One challenge with this technique, is whether or not the business believes that email activity has a particular bias toward offline vs online. Therefore this insight should be used to underpin one or more different measurement techniques. Rather than be relied on in isolation to provide the insight upon which investment decisions are made.
Holdout tests are a great way to take advantage of advertising technology that can geo-target campaigns, such as paid search, enabling marketers to identify like for like regions. Marketers can then up-weight or down-weight spend, tracking results across online and retail.
A business with a large enough retail presence, can split up the country into like for like city regions, for the purpose of conducting an A/B test. For the A group, spend could remain at a baseline, and for the B group spend could be up-weighted.
This requires the orchestration of the geo-targeted advertising campaigns, budget, online spend and conversion data, and offline sales and footfall data.
The resulting online traffic and sales, and store footfall and sales, are compared statistically for each of the regional groupings, A vs B.
The output of this analysis can reveal whether there is a statistical relationship between the digital spend and retail footfall and sales.
Costs will need to be considered; missed opportunity in the case of down-weighted regional spending, or increased budget for up-weighted regions, depending on the design of the test. Therefore adequate skills, technology and methodology are paramount.
The value of insights delivered is high, and so are the risks of delivering ambiguous results, which makes this technique a worthwhile endeavor that should be approached with due care.
Challenge the Measurement Status-quo
There are other techniques in the marketers tool box, such as click to call, and in-store beacons. Developing this level of understanding is not a one and done activity. Invest the time and effort to learn the insights that are valuable and elusive, because in doing so, you can gain a competitive advantage.
Lastly, don’t stop with one experiment, if you can come to similar conclusions using different techniques, the confidence level of those you have to convince, should increase. By conducting three types of measurement, and triangulating the results , you can create a solid foundation upon which to make decisions on how to update your marketing investment, efficiency metrics and ROI measures.