The first step of being heard by Product is actually putting yourself in their shoes to better understand their challenges. Imagine you’re busy compiling data from multiple stakeholders, it can be difficult to know who to listen to first. If Mary from Sales says customers need one thing, Bob from Support says customers need another, and a customer interview reveals a third possibility – who’s right? When it comes to influencing product decisions, we need to come armed with trustworthy, meaningful data.
Most support leaders inherently know what customers want, we have a deep understanding of the needs of our users, and we care deeply about getting that information into the hands of decision makers. But retrieving meaningful data from thousands of user conversations is difficult. And when arguing with, say, sales, we have to compete with cold, hard, revenue figures.
The good news is that our daily interactions with users give us the data we need to be influential. If we track data in the right ways, we can then go to product with more confidence and have a positive effect on development strategy.
The key to making support more influential in cross-departmental disagreements is to come to the table with quantified, contextualized, curated data.
Wrangling this data from support conversations is something we’re trying to make easier at Idiomatic. Here’s how we approach this challenge:
How many other people have this same problem?
Contents
Quantify your data
Imagine you’ve landed a new role as a smartphone designer. What would be more helpful in your search to design the perfect phone:
- A few people think screens should be bigger than they are today.
- 24% of our customers have written in complaining about not being able to read small print on their screens.
The first one is an offhand remark. The second is useful information. If almost a quarter of our customers have difficulty with visibility, that’s definitely something to address in the new smart phone. Maybe we can improve zoom, or make the screen bigger. It’s a really difficult statistic to ignore when we’re asking what customers want in a new model.
Support teams frequently struggle to quantify data from customer conversations in a meaningful way. Most smaller teams will rely heavily on ticket tagging. When a customer contacts support about a specific issue, the agent can attach a pre-defined ticket tag. Then, when compiling feedback for other departments, support can pull a report of the number of conversations filed under that tag.
There’s two problems with that though. First, you need to know what you’re looking for; you can’t tag for everything, you need to be focused and pre-define the tags. Second of all, manual data review can be massively time consuming.
For example, Upwork’s Customer Experience team used to have three people reviewing hundreds of cases a week to pull out actionable insights. The manual effort was yielding an anecdotal and imprecise sketch of the customer experience. When they switched to using Idiomatic, they were surprised how quickly they could get even clearer insights.
Joe Wang, Director of CX at Upwork, says:
“Idiomatic gave us the customer voice insights we needed to improve support operations while eliminating the need for manual reviews.”
Instead of relying on the manual tagging and analyzing of support conversations, Upwork can see every trend brought right to the surface.
Most tools that analyze feedback use Natural Language Processing (NLP) technology. This can be helpful but the problem of lack of specifics still exists. For example, if I tell Joe at Upwork that there has been a spike in payments issues detected by my NLP, someone on his team still has to do a deep dive to figure out what’s going on. However, if I tell Joe that there has been a 15% increase in freelancers finishing the job and then the payment failing due to incorrect bank details, he already has some idea what to do to fix this problem. To get to this level of detail, Idiomatic creates our own training data then uses supervised machine learning to train custom models to label this feedback specifically, identifying trends you didn’t even think to look for (or have tags for), and turns customer feedback into data that helps resolve disagreements.
Regardless of how you do it, quantifying support conversation data makes it easier for other departments to act on support driven insights. If tagging is your chosen method then check out our getting started with tagging guide.
But what is really going on?
Contextualize your data
When presented with data, most logical people will want to dig in deeper. They’ll want to know the “why” behind the numbers. Who are these people? What led them to think this way? Can we characterize these users further? That “why” can often be uncovered by analyzing second level trends, or contextual patterns that emerge from within the subset of data. Secondly, context becomes even more compelling when quantitative data is backed up with meaningful qualitative stories.
Combining the quantitative with the qualitative is the true secret to driving product decisions with support data.
Lindsay, Head of CX Ops at Slack uses Idiomatic to provide data on customer issues to their engineering team. She loves how quick it is to dig deeper into trends that the AI finds:
“I answered my CTO’s question about the root cause of an issue in the meeting instead of having to wait and do a manual audit of tickets.”
If you’re using AI to surface trends, it’s easy to leave the cross referencing to the computer. There’s no need to pull up every ticket with the relevant tag for further analysis. The AI will have already identified the second level trends within the subset of tickets. Product owners can dive deeper with a single click. This transparency brings even more weight to the arguments customer support teams have been making all along.
Even if you aren’t using Idiomatic It’s key to keep examples all in one place. While Product Owners aren’t easily convinced by 2 or 3 anecdotal stories, they will want to read 20 tickets to have a better idea of what customers are talking about. If you’re using tagging, linking statistics to the tickets responsible for those statistics can be difficult. It’s crucial that you do the hard work of putting all these tickets into one place instead of just pulling 2 or 3 examples because product owners are unlikely to start searching through help desk conversations to find the necessary tickets that it will take to convince them.
To settle disagreements of what customers really want, teams need easy access to second level trend analysis and qualitative stories.
Why should I care?
Curate engaging data
Finally, voice of the customer data needs to be engaging. You can have all the quantified data you want, but if it no one reads it, you’re not going to make a difference. Curation of data can help make reports more relevant to each department.
To us, curation means:
- Trimming out irrelevant or misleading data
- Aligning feedback with current company goals
- Reducing noise or less important trends
For example, one way Idiomatic curates engaging data is through the use of sentiment analysis. NLP classifies customer conversations by emotion and tone, so teams can pull out the most extreme feedback to distribute.
Intercom uses idiomatic to send VoC data across the company in a weekly report. Often this kind of weekly report is difficult to get engagement from. But Sian Townsend, Director of Research at Intercom says they’ve had good results getting the relevant information in front of everyone:
“We’ve got loads of interest in the Idiomatic data from across the company! It’s so exciting to see Intercomrades’ enthusiastic reactions to the weekly reports.”
Enthusiasm? About reports? When you’re providing the information colleagues need to do their job better, that’s a very possible reaction to a report.
Don’t bring a knife to a gunfight
Maybe equating a product discussion to a gunfight is a bit extreme. But the advice is sound. When you’re advocating for the needs of your customers, it’s your responsibility to prepare properly. If you try to influence product decisions with opinions, don’t be surprised when support loses its seat at the table.
Preparation takes time, but if you put that time in you can get great results. Instead of bringing gut feelings, unsubstantiated claims and anecdotal feedback, support teams need to develop quantified, transparent data to backup their requests. AI can help with that, but step one is recognizing that it’s going to take effort and making a commitment to invest resources in preparing yourself for success. Good luck!
If you’d like to learn more, reach out to us and get your custom demo!
