Social Engagement Routing Models
By Edwin Margulies
What routing model should we use? This is one of the biggest questions posed by social care practitioners during the project planning stages. But choosing the right model may not be as intuitive as you think. Here we'll take a look at three popular routing models and their benefits to make the choice a little easier.
Routing Models: Not a Religion
I've heard some pretty evangelical screeds on what routing "has to be" for social engagement. That usually happens when someone takes a position based on what they use, not what's possible. In other words, if all you have is a hammer - everything starts to look like a nail.
Of course, the best choice is the one that best fits your needs. Although the nomenclature may change from vendor to vendor, here are the top three routing models to choose from:
• Universal Queue Engine (ACD Model)
• Supervisory Filtering (Controlled Agent Filters)
• Agent “Cherry Picking” (Agent Autonomy Model)
Universal Queue Engine (ACD Model)
With a universal queue or ACD (Automatic Call Distributor) model, each social post item is routed based on a variety of attributes. These attributes are both agent and queue-based but also include attributes about the social posts themselves. On the agent side, these attributes may include the agent's availability, the number of outreached items an agent may have already received in the past hour, and the agent's skill. Even the level of skill proficiency can be taken into account with some systems. On the social post side, information like the publication date, author, and content are used as routing criteria.
With this ACD model, the agent does not decide what social post item to work on. Instead, this is decided by the universal queue engine (ACD) and each item is then pushed to the agents one at a time.
For this model to work, the social engagement platform must provide real time information the universal queue platform. For example, the social engagement platform will have to tell the ACD all about the social posts that are coming in from the social web. For example, the social engagement platform may provide spam scores for each item, sentiment, author influence, and even cluster information. Of course, none of this is possible if the social engagement platform does not have an NLP (Natural Language Processing) engine to do advanced filtering.
The universal queue engine to use its own algorithms and pre-set rules to determine which agent should get a particular post. For example, the ACD could be set to send all angry customers who are posting about a particular product to the "retention" queue or to a specific individual who is skilled on that product. The level of sophistication in your routing tables is really up to you as a user or practitioner. Many ACDs are set up to simply route based on availability or who has taken the least number of transactions. All of this is programmable.
The are several benefits to using a universal queue model for routing. First, you can "blend" media types. This means the ACD can send a phone call to an agent, then after the call is done, it can send a social post item, or even an email. This is beneficial because the non-real time items can be distributed in such a way that the "peaks and valleys" are taken out of the equation. That is to say that emails and social items can be pushed to agents who have inbound telephone call downtime.
The second benefit of the universal queue model is the ability to piggy-back on supervisory coaching and oversight that already exists in the enterprise. If your social engagement team is growing, this is important because a ratio of 20:1 agents to supervisors is a best practice. It makes sense to cross-train both agents and supervisors if you are going with a universal queue model for economy of scale.
Although not a full-blown ACD, supervisory filtering provides you with a fairly sophisticated routing approach. In this model, supervisors are able to change the filters that govern what is displayed on each agent’s screen. This means supervisors can be selective about the types of social post each agent gets to work on.
For example, the Supervisor can set an agent’s filter so he sees only angry customers who’s social posts are on a certain subject. Likewise filters can be set to show a certain workflow disposition or relevancy score. Let's say you are running a grocery chain. You could make all social post items about rotten fruit with angry customers who were referred to an escalation group. This is not a full-blown ACD model, but it allows supervisors to create special filters for each agent so he or she has more control over the workgroup.
There are a few benefits of the supervisory filtering approach. First, it's very hands-on and allows the supervisors to quickly adapt to on-the-fly changes. This is especially useful for smaller groups of agents who are not "blended" agents but rather dedicated to social media. Second, you can deploy this model quickly and there is no systems integration or dependencies on other platforms.
Agent Cherry Picking
Agent “Cherry Picking” is another model altogether. Here, agents are able to view all of the social post items from a common queue. Agents get to choose which items they will pick out of the queue. This is typically done on a first come, first served basis. This means as each social post item is taken by an agent, it is locked so other agents cannot engage with that customer.
If your social engagement for customer care platform does not do native Spam filtering, clustering, author sentiment, or author influence, the type of routing model almost becomes moot. That's because without advanced filtering, it's kind of a free-for-all anyway. However, if you have a system with NLP (Natural Language Processing) that handles sophisticated filtering, agent cherry picking affords you the same level of routing as supervisory filtering. The only difference is you are putting the controls in the hands of the agents.
The benefit to this "agent autonomous" approach is that you are able to give experienced agents some latitude in filtering their own queues. Agents that are being on-boarded or who are not very experienced will need more supervision. It is possible with some platforms to "mix and match" these routing models. For example, you can use supervisory filtering for new and inexperienced agents and cherry picking for the veterans.
No matter what routing model you choose for your customer care initiative, consider the flexibility and benefit of a hybrid model that allows you to make changes on the fly. If you have a traditional contact center operation with blended agents, using a universal queue model is a great idea so they can all work on social post items. If you have a group of care specialists who are dedicated to social, you can use supervisory filtering on the side of an ACD. During transitions from dedicated to blended agents, you may need all three models, so consider platforms that allow you to operate all three.