lundi 23 janvier 2012

How to use Agile methods to deploy a CRM system

Despite the user community's urgent calls for this or that functionality, customer relationship management (CRM) is still enterprise software that needs to be done with an architecture and a plan.

Deployments of unstable new functionality (or pushing existing functionality to unprepared users) can diminish the credibility of the CRM system. This can mean serious backsliding of user adoption, unwinding the virtuous cycle that is at the core of CRM success.

How to manage this issue? You won't be surprised that I advocate incremental deployments and Agile project methodologies, as they are the fastest (and cheapest) way to get functionality safely deployed. That said, iterative delivery is necessary but not sufficient. To establish sensible CRM priorities, you need to add the following principles:

Data First, Functionality Second
There's no point in building (or even switching on) new functionality if the underlying data is incomplete, dirty, or riddled with duplicates. While nobody can afford perfectionism, getting relevant system tables cleaned up is a precursor to progress. In nearly every table, the error rate needs to be below five percent, and even one percent for critical fields and pointers. Pretty much the only area where you can tolerate as much as a 20 percent error rate is in leads, as its speed of "information rot" makes higher standards uneconomical.

Accounts and Contacts are the Most Important "Static" Data
In most CRM data structures, the accounts table is the top of an information pyramid, with a dozen or more child tables pointing to it. If your customers are large multinational corporations, the account records themselves may be part of a hierarchy. So making sure that your account records are right (and agreeing with the data in other systems) is a key milestone. Fortunately, the account records themselves shouldn't change all that often.

Relating to the accounts are the contact and contact role records, which need to be accurate and free of duplicates to make sure that calls, emails, and action items are properly tracked. Unfortunately, in 80 percent of the CRM databases we see, the contact role record (it's just a pointer, really) is empty, rendering almost any marketing effectiveness or serious pipeline analysis impossible. As this field can only be filled in by humans (either sales or telesales), the only solutions are incentives or other behaviour modification techniques.

Opportunities and Cases are the Most Critical "Transactional" Data
All too often, sales reps manage their boss by managing information. Ok, hiding it. Real deals are not in the system, but fake ones are. This means executive management cannot see what's going on with the pipeline until it's too late to fix it. Watch for symptoms like "submarine deals" that pop up unpredictably during the quarter close, or pie-in-the-sky pipeline that shrivels up all too predictably as the quarter progresses.

While it's easy to come down hard on these practices, it's important not to push behavior modification too fast. The sales reps will engage in passive-aggressive resistance, and may end up gaming the system in an even more misleading way. Expect progress on this front to take six months or more, and measure success by gradual improvements in forecasting accuracy.

On the customer service side of the house, cases (aka incidents or service calls) are the most pivotal data. The problem with cases is not that they aren't in the system - the issue is incompleteness, as they don't reflect reality.

Every CSR or service tech needs to be on the system all day, updating the case records with status changes they might previously have kept in Excel or on paper. Every "truck roll" or customer call needs to be recorded as a task so you can see the sequence and the real accumulated time required to resolve issues.

Integration is the Hard Nut, but Enables the Most Value
The more serious your CRM system, the more other systems it will need to integrate with. When a sales or service rep can see the big picture (that may be stored in a half-dozen outside systems) of the customer situation, they can get the customer happy and paying sooner. When executives get closer to a 360° view, they make better business decisions.

But every point of integration is an invitation to data quality and duplication issues. While off-the-shelf, point-to-point adaptors are widely available for CRM systems, you'll almost certainly want to use a real integration server to handle transformation workflows, network timeouts, two-phase commits, and compensating transactions to unravel error conditions. Beware: integrations tend to be highly susceptible to subtle new bugs whenever any of the systems does a patch or upgrade. Watch out particularly for duplicate record creation or foreign-character set incompatibilities.

Until Reports are Clean, Your Data Isn't
CRM users, particularly executives, tend to ask for reports and dashboards first. Unfortunately, these are highly visible ways to expose every one of your data quality problems. Some of these will be digital noise -- misspellings, transposed dates, bad pointers -- that can be quickly found and corrected using pivot tables and other digests. Some of the problems will be more insidious, caused by business processes that cause "semantic smear" or by integration problems that can take weeks to troubleshoot and correct.

Our advice here is to keep users away from reports until you're really sure about data quality. The easier the CRM system's report wizards are, the bigger an issue this is. Early on, reports should be produced only by an analyst who really knows the data model and the things to filter out.

Users Must Believe the CRM System is For Them
Finally, the features you invest in should be the ones that help the users do their job better. Although they know management will be using the system to measure things, they must not believe that the CRM system is principally a spying machine for micro-managers. So spend money on things that save users a few mouse-clicks on every transaction, or features that will make their job just a little bit easier all day long.

We even encourage eye-candy features that don't actually add a lot of value, but make the system a little bit more fun for the users or the customer.

Aucun commentaire:

Enregistrer un commentaire