The e-commerce conversion rate is undoubtedly advertisers’ preferred indicator, yet it holds many limitations. While in use for the last 15 years by online retailers, it has been interrelated with attribution models. When you use the last click, multi-touch, or even data-driven models, are you actually aware of the limitations of each different conversation rate generated?
Before exposing the limitations of the e-commerce conversion rate, let’s review some basics.
The e-commerce conversion rate is a metric that determines the share of visits that have generated a sale. This “sales generation” notion is based on an attribution model that consists in spreading the weight of a sale on one or more channels.
The limitations of the conversion rate
Let’s illustrate this situation by comparing an e-commerce sale to hiring an employee. No recruiter hires an applicant after reading his/her CV just once. In a conventional hiring process there are several steps to follow, e.g. receiving the resume, a phone interview, a one-to-one interview with HR, then with the present department manager.
Only the last interview with the manager would be eligible to be factored in the calculation of the attribution rate in a final interaction environment. Similarly, in a first interaction attribution environment only the reception of the CV would be eligible for the conversion rate calculation. In a nutshell, the conversion rate would cover up all the unattributed interactions.
Quick reminder, the e-commerce conversion rate is calculated this way:
(Conversions* / Visits) x 100
* Please note that the number of conversions considered is heavily impacted by the type of analysis chosen (by channel, by campaign or even by keyword(s)).
Similarly, the conversion rate generated evolves anytime there is a change in attribution model.
If the attribution model is not very generous with a channel, then its e-commerce conversion rate would be worse off (fewer attributed conversions); conversely, the e-commerce conversion rate is generous towards a channel, if its model is generous.
While making a decision based on an e-commerce conversion rate remains heavily data-oriented, an attribution project is not data-driven in itself.
In that sense, the e-commerce conversion rate has become an attributed piece of information that needs to be made complete thanks to unattributed indicators.
Thus, following a data-driven approach, DCC (data-driven companies) consider the conversion rate rather “old school”, but is there a more reliable alternative?
The “useful visits rate”: THE alternative
What indicator is not affected by attribution? The useful visits rate!
Especially as it takes into consideration visits remanence over several days or weeks.
What is a useful visit?
It simply is a visit that successfully intervenes in the conversion journey.
While the conversion rate considers that a single visit has generated the sale, the useful visits rate takes into account all the visits that have contributed to the sale. Thus, the useful visits rate restores the contributing power of channels that had previously been undervalued by the use of a specific attribution model.
If we take again the example of recruiting an employee, the useful visits rate would consider all stages of the recruitment process. The conversion rate represents a biased vision of performance, while the useful visits rate puts forward more realistic results. Furthermore, using this indicator is a very useful means to compare ones’ channels without any bias.
The useful visits rate is calculated this way:
(Useful visits / Visits) x 100
Over a hundred visits originating from SEA, only 2 lead to an e-commerce sale according to the last click model. Thus the conversion rate is of 2%. This model is not very much in favour of SEA (Search Engine Advertising), contrarily to the first click model that attributes 4 e-commerce sales to this channel. This leads to a 4% conversion rate. The algorithmic model suggested by my SEA agency attributes 10 e-commerce sales to SEA. The conversion rate then reaches 10%. While we are using the same data set, the conversion rate evolves by the sole change in attribution models.
The useful visits rate represents all the visits taking part directly or indirectly in a conversion path. Thus, over the 100 visits generated from SEA, 85 are “useless” visits, i.e. it has generated visits but none has successfully taken part in conversion paths. Conversely, 15 visits are considered as “useful”, as they have helped pushing forward web users in their purchase journey.
Over the 15 visits:
- 1 visit was present at the end of the journey,
- 3 visits initiated the conversion journey,
- 1 visit was autonomous, so it intervened simultaneously at the first and the last visit of the conversion,
- 4 visits were present in the same conversion journey initiated by SEO (Search Engine Optimization), and ended by retargeting.
- However, the 5 remaining visits took part into 4 different conversion paths, these were often in the middle of journeys involving 3, 4 or 5 visits.
Here is a brief layout of conversion journeys that were deemed eligible to be part of the SEA useful visits:
- AFFILIATE > SEA > SEA => Conversion for a total of $120
- SEA > EMAIL > EMAIL => $50
- SEA > SEA > RETARGETING => $60
- SEA > AFFILIATE => $40
- SEA => $35
- SEO > SEA > SEA > SEA > SEA > RETARGETING => $60
- SEO > SEA > SEO > AFFILIATE => $40
- AFFILIATE > SEA > AFFILIATE > SEA > SEO => $100
- EMAIL > EMAIL > SEA > RETARGETING => $75
- SEO > SEA > SEO => $90
Your web analytics tool uses the conversion rate, but what about your other solutions: A/B testing, merchandising optimization, customer live chat, bid management?
Do your ad agencies algorithms optimize the publication of your campaigns, while taking into consideration the contribution of all your channels?