Optimisation techniques are used in a variety of business contexts to find the best combinations that deliver the desired results, often measured in terms of value added from maximising revenues, minimising expenditures, or both. In marketing, a frequent problem that's well suited to optimisation is when one has:
- Many customers (often millions of them) ,
- Multiple potential offers,
- Rules that determine the number and frequency of offers,
- Resource constraints (such as budget), and
- The need to maximise something, such as sales or profit.
Optimisation quickly gets complicated because to find the best solution, one needs to consider all the combinations that exist, and with millions of customers and hundreds of communications, optimisation is not an easy problem. SAS solves this with SAS Marketing Optimization (SAS MO), but the principles surrounding the challenge that I am outlining here are relevant to any of the so called ‘large scale’ optimisation challenges.
Often people think that they need many models to justify using optimisation techniques, or to get the best out of the optimisation. So whilst the latter is partially true, I really don’t think that the former is, especially when considering what one is trying to achieve. I would even go as far as to say that you can get significant performance improvements with optimisation, even if you have only a few, or even no models, (none, zilch, zero!). Let me try and justify this:
The uplift in optimisation solutions, when compared to how a BAU (Business as Usual) approach is performing depends on three core factors:
- How good is the BAU approach? If the current approach is to choose the offer(s) and customer combinations is weak (e.g. a simple campaign or customer hierarchy prioritisation), then SAS MO will give a big uplift.
- How many customers and campaigns are there? Big customer files with lots of potential decisions are hard problems to solve, so true optimisation techniques like SAS MO can give a big uplift.
- How many constraints and contact rules are there, and how complex are they? One or two constraints are easy to manage usually, and simple contact rules (e.g. 'no more than one offer, this month, and just optimise this month'). When they get complex (e.g. 100 campaigns, each with a volume constraint, and daily, weekly, monthly contact rules across multiple channels), then many other optimisation solutions grind to a halt, but SAS MO can thrive on this complexity
Even before considering predictive models and the quality of their outcomes, if one of the three core factors is present, there is a good chance of significant uplift coming from the optimisation approach alone. Let me give an example of a client solution I worked with once, which included the following challenge:
There were 100 campaigns, occurring on a daily basis, across a customer base of 30 million customers, and the goal was to optimise one month of campaigns. This is effectively over 3,000 campaigns, across 30 million customers. This is complex – and is challenge number 2 above.
- There were multiple campaign rules across multiple channels, across days, weeks, and months, across different campaign groups. This is also complex, described as challenge number 3 above.
- Their goal was to try and maximise the number of communications that they wished to send to their customer base - essentially trying to fulfill Marketing’s desire to execute as many of their campaigns, as fully as possible. We could all argue that in a perfect world, marketing would be focusing on response, relevance, value, etc. but sometimes the old adage that ‘more is better’ is still the driver!
Their BAU approach was one of a "hierarchical approach," which means that they would:
- Identify the most important campaign, and select all customers that were eligible for this campaign, or the maximum volume allowed was reached.
- Treat these contacts as if they had occurred – i.e. as a contact history.
- Identify the next most important campaign, and again select all customers eligible for this campaign.
- Repeat steps 2 and 3 until all campaigns had been considered.
There were no response models since targeting was based on selection criteria set by Marketing. The complexity of the problem meant that they were only reaching 50% of the target market on average, and campaigns that were very low in the hierarchy would get far lower than this, despite them being potentially very relevant to some customers.
By applying optimisation techniques, with the goal of maximising the number of contacts (with the inherent assumption that more contacts meant more response), the complexity was managed, and the proportion of the target market that was reached was significantly increased from 50% to over 75%. There were notable side benefits too, with a simpler process and a more informed view of the campaign volumes prior to execution.
The performance increase in lead volume also resulted in an increase in response, and therefore revenue, thus creating greater value for the business. More satisfying still was that the success of the relatively simple KPI of lead volume delivered business buy in to a roadmap of improvement. This mapped out development of response and revenue models that then drove new KPIs that were maximised, delivering even greater value to the business. This greater leap may not have been reached without first getting to the ‘stepping stone’ of increased lead volume.
Optimisation is clearly a technique that can help today’s marketers deliver greater value to their business. This is still the case even if another approach to address multiple customers, offers, rules and constraints is already in place. Moreover, optimisation techniques can deliver greater value even when few models have been developed by providing a stepping stone for other models to derive the value from analytics. This stepping stone is not always in the same place for all businesses. My experience tells me that the closer this stepping stone is to the current way of working, the greater the likelihood of successfully reaching it, and the greater the leap that a business can make from there.
Optimisation is not just the technique that facilitates this, but also provides the impetus to the business to deliver more customer focused analytics too. When applied to marketing, optimisation creates value by allowing confident data-driven decisions to meet business goals. Knowing that - if you are not using SAS Marketing Optimization today, can you come up with a good reason why not?