Before the days of the internet many companies favoured direct marketing activity due to its ability to track performance. The types of channels we consider are: Direct Mail, Door drops, Press Ads & Inserts (with offer codes)
For these channels, you typically had some sort of “reach” metric – i.e. how many you sent, and you could look at your orders to see how many people responded within a given timeframe of receiving the communication:
- Direct Mail and Door drops could be matched back at address/postcode level
- Press Ads & Inserts could include a coupon used at the point of sale
When the internet came in, it led to a few challenges:
- The cost of reach (i.e. per impression) was usually smaller for digital channels
- Digital has more KPIs, the concept of a click or engagement in the ad (a middle funnel activity)
- Also, it led to the question of challenge of attribution – people may see an DM piece but engage with paid for media
This means that we now have a new qualifying criteria for successful measurement of offline incorporating online as an element of that measurement:
- Can we measure the middle funnel engagement of the channel? (i.e general engagement/website visit)
- What is the overall sales uplift from the activity?
- How does offline work with online channels?
- Does it drive new people online
- Does it help convert those already engaging online
- What is the overall value we can assign to this activity
Let’s review each of these in turn.
Can we measure the middle funnel engagement of the channel?
In simple terms you need to be able to link a website browser, which you would normally capture within tools like Google Analytics, with an individual.
Google Analytics tries to avoid using personal data, so usually this would be a paid for piece of software (like WebFusion). The tool would work in a similar way, however it would link the cookie or website browser to an individual when they buy or engage with tracked communications like email.
It is important that you look for a tool which can also report standard digital metrics, i.e. email platforms often capture some of this information; but often has gaps before the point of contact. We have seen that website personalisation software can fill some of these gaps.
Fundamentally though if you can get a single view of digital activity and a persons offline touch points you can then cut the reports however you require and see that middle level engagement.
Overall, this helps specifically when campaigns don’t perform as they can answer questions like “did the DM activity drive people to the site, and did the site convert it?”
What is the overall sales uplift from the activity?
Offline offers many opportunities for measurement of uplift:
In direct channels you can split cells into test and control and do standard A/B tests, whether this is a group of individuals getting direct mail or a group of postcodes getting door drops. It is important to make sure both cells are comparable in all aspects other than this piece of activity, and the control cell sizes will give you a statistically significant result when comparing the results.
Press & Inserts can be a little trickier as you cannot control postcodes.
You can look at time series analysis, comparing a period on vs off.
However, there are lots of questions around long term brand impacts; i.e. if you stop ads for a month is that truly representative of the loss; as people may still be engaging from content from previous months
Coupons provide a good way of tracking sales that must have been driven by offline. Advanced tracking which allows you to look at the sales with these coupon codes and asses their usage of other (digital) channels; can also provide a reasonable steer. I.e. if 90% of people that use coupons come straight to the site and don’t use other channels – then we can safely assume its uplift.
Adding in additional layers of tracking, for instance using QR codes with UTM parameters or custom landing urls can also give an idea of how many people are engaging specifically with the channel. Once blended with digital data this can give an idea if these people are using other channels or not; which can determine if it is incremental.
How does offline work with online channels?
In the absence of proper tracking, we have to make layers of assumptions around the percentage of people that use channels like PPC after receiving a direct mail or offline piece.
For example, we may match back a direct mail campaign, and then estimate that 30% use PPC brand 10% PPC generic, and the rest is direct.
But we also have to consider, if a person has a catalogue in their home, are they then more/less likely to respond to a display or social ad.
Similarly, we may find that being exposed to online ads may make people more likely to engage with a physical piece of marketing.
The advantage of proper tracking in these instances is that we can easily quantify the number of cases that have these patterns, you should be able to report:
- How many people engage with offline, and use no other tracked marketing channels
- How many buyers had no online activity before receiving offline
- How many buyers had already engaged online, and seem to react to marketing
The pursuit of a single attribution/measurement model, whilst a noble cause, often overlooks the simple concept which many believe that “people respond to an omni channel experience”.
Whilst it is impractical to measure every single possible combination of channels; my view is that you need to be able to quantify the higher-level impact of combinations of marketing.
In many cases there are multiple teams, for example, different channel owners or product owners, who spend so much time looking at their specific measurement; for these cases a better data and measurement strategy can:
- Quantify the amount of overlap; which gives a full range of potential outcomes
- Can assign the total value of crossed campaigns, and a framework to measure these separately to unique touchpoints
- Enable new activities to be measured
Impact of AI
In today’s world we have to ask ourselves what the impact of new modelling or AI tools bring.
It is important to note that AI can only pull-on sources of data that any person can do, i.e. data on the web or in encyclopaedias. Any approaches that rely on AI, really are also using models to fill in any gaps.
Modelled outcomes have to make assumptions on how customer journeys have worked and the uplifts involved. For example, you may say that a person of a specific demographic has a 5% of seeing a direct mail piece and a 10% chance of engaging with an online AD.
Only real data will tell you whether there is a high or low overlap in those two channels and if those synergies lead to more sales.
My view is that AI can help identify good sources of assumptions, or forecasts, however we need humans to decide which assumptions we want to project and scenario plan around.