We’ve learned that being successful, truly successful, with text analytics requires specific patterns in case teams’ processes. What’s more, customers who take ownership of their analytics success appear to travel very similar paths, and we’d like to spend a bit of time today exploring a few of these trends in the form of three useful tips.
Successful analytics users:
1. Know their data.
Knowing what documents you have and how you expect them to behave in an analytics setting is an essential step in your project planning. We don’t see projects go south very often, but when we do, it is almost always because someone didn’t ask enough questions about the documents with which they’d be working.
For example, they:
- Ask if the case hinges on numbers, or if a single word in a contract will determine relevance.
- Clean up their text with regular expressions to optimize email threading.
- Develop a strategy to identify documents, such as those with poor OCR, which will require separate, manual review.
2. Ask, don’t guess.
Unless you are a seasoned veteran, there comes a time when you might not be sure what to do next. Even the simplest tools might give results that are open to multiple interpretations, or have workflows that perhaps could be tweaked to achieve better results.
In these situations, the more successful users take a minute to stop and ask a technologist—on your team, at a partner, or with your software vendor—to help them interpret their results and assess next actions. This approach is far more preferable to forging ahead and hoping for the best.
For example, they:
- Select and groom an internal analytics guru to guide the rest of the team. Having your own point person who masters the tools and achieves certification status not only streamlines and shortens your organization’s decision-making process, but also adds prestige to your team both internally and externally.
- Seek out workflow experts before acting. This is decidedly an “ounce of prevention vs. pound of cure” situation. For example, it’s very common for newer users to seek advice early and often during their initial analytics efforts. Once they navigate the process a few times, the need for support drops off precipitously.
3. Practice, practice, practice.
It is said (ok, by me), that your first day of review is like your first pancake: You feed it to the dog (I don’t have one, but let’s pretend). You then adjust your temperature and move forward.
The same applies to your first use of analytics. You might get a good result, but only after regular use do you get into a good rhythm of what works and what to expect. Once you have a better feel for how all the pieces fit together, you’ll be able to improvise new workflows and focus on expected results.
Some folks wait for a really big case to get their feet wet in analytics, but this is not necessary. It is much simpler to cut your teeth on many smaller cases, especially when it comes to structured analytics, such as email threading or textual near-duplicate detection.
For example, successful teams:
- Thread everything that contains emails, regardless of the project size. We see teams running email threading on data sets as small as 5,000 documents. There is still incredible value to be found in these smaller cases, and the proficiency you gain will only further ensure your success in the bigger cases.
- Have internal learning events. One of the more impressive pieces of feedback we’ve received from customers is how they share their experiences internally. Some have regularly scheduled lunch and learns, while others do case studies or retro-analyses on cases once the project closes. Regardless of how you share lessons learned with your colleagues, sharing practical tips and results is a great shorthand type of practice.
- Find ways to incorporate analytics into their existing workflows. The most skilled analytics users inevitably see past the myth that analytics usage is an either/or situation. There are many great hybrid options here, such as performing keyword searches with the help of conceptual keyword expansion; using cluster visualization to prioritize data during a manual review; or streamlining a QC review of privileged documents with near-duplicate detection.
Analytics features are not inherently any more difficult to use than other tools. As with any new technology, two of the main reasons some folks perceive them as more difficult stems from a lack of exposure and a lack of practice. These obstacles are only temporary and easy to overcome, as we’ve seen time and again with our customers.
Constantine Pappas is a licensed attorney with more than 15 years of legal experience. He is a member of kCura’s customer success team, helping Relativity users build workflows for analytics and computer-assisted review.
- How 1 Curious Experiment Dispelled 4 Analytics Doubts
- An Attorney’s Answers to 3 Questions on Analytics and e-Discovery
- Throwback: Speaking the Same Language as Your Team’s Analytics Guru