BA, UX, PM
A fellow web developer introduced me to Alex, CEO and founder of an appliance repair services company with several offices across the US. Alex started the business back in the times of paper documents workflow. Now he had to go through a tough process of adaptation to the digital age.
wow, such a direct dependency! The company makes 90% of revenue as a subcontractor. And the volume of contracts the general contractor provides ultimately depends on a complex customer service score.
CEO believes office operation is the bottleneck, because:— customer service score largely depends on office operation— customer service team productivity is low; the team is stressed— growing the team from 5 to 8 members didn’t have a tangible effect
Investigate customer service team workflows and multiple complaints about software. Propose and implement solutions that will increase customer service score by at least 30%.
I needed to clearly understand 2 things:a. what is the lifecycle of an order in the companyb. how customer service score algorithm worksSo first of all I’ve built order lifecycle map. This is how a heavily simplifed map looks like:
A hugely simplified order lifecycle map. In reality it may take up to 240 steps from start to finish.
To complete this task I had to go through every level in organization, which was very complicated: each team member was heavily multitasked, and each talk would cause more stress afterwards. If not my superweapon, pulling guys out of their tables would have been next to impossible.
Cafe down the road. My superweapon. Makes people talk.
While building the map I was figuring out what were the core metrics for customer service score. I have managed to meet branch managers from 2 largest general contractor companies. The score algorithm was complicated, yet there were several factors that largely affected the score:
General contractors will tolerate a lot of things, but not communication issues. Each time a customer contacts general contractor saying she failed to reach subcontractor, it will negatively affect subcontractor’s score. General contractor will occasionally try to reach subcontractor as well and fail to do so (in this case the penalty is even more severe). So I’ve started observing how the team handles the calls:
Calls go through the circle until available agent is reached.
Some smaller issues causing bigger problems.
When I was about to give up search for a reasonably priced out of the box solution, I found Nextiva. This SaaS piece met all the requirements to address the pain points I’ve discovered:
From the demo use Nextiva left an impression of quite user friendly service with a comprehensive analytics tool; and importantly, it was affordable and easy to migrate to.
+ 3 weeks after migration I have checked a weekly average time on hold. We have managed to go from 10min 42s down to 3min 15s (+267.6% faster).+ With prioritized queues the team missed no calls from general contractors.+ 6 weeks after our customer service success rate went 32% up, 2% more than initially planned already.
Having the initial goal accomplished in the very first iteration we could afford digging deeper into order lifecycle map, as the initial phase of work left me assuming that...
“The entire data workflow is uneffective”.
The assumption turned into a hypothesis as I have discovered that many sets of data make a very long way from one entity to another. Here’s an example of work order data that includes from 8 to 12 items (customer address, name, type of appliance, serial number, order number and etc.):
Apart from redundant manual labor, this process caused multiple smaller issues along the way, such as technician copies info with mistakes or forgets to attach a picture, or the manager not understanding handwriting or copying technician’s mistakes to the report.I have managed to convince CEO to make deeper changes rather than tackle manifestations of a larger problem. So our updated plan was to streamline the process and lay foundation for future automation.
The team has been using Lua messenger as a center point of conversations, so I was willing to explore the direction of a modular messenger playing a central role in the team’s processes. As a result, I have managed to optimize many of processes with Slack API, Zapier and a bunch of webhooks. This is where our full-stack web dev stepped in.Here’s the process we’ve mapped out and integrated:
Agents spend no time on going through emails for copy pasting.
1. Configured emails We have configured the inbox filters so new orders get their own labels. Each labeled email goes through a Zapier parser that picks up what we need, cutting rest.
2. Created pre-filled web forms We have designed a web form that pushed invoice routine to online space. Parser sends data to our server, where we generate a form pre-filled with data at the unique URL.
No copy-pasting + data pre-filled = faster workflow for technicians.
Minimizing manual inputs: agents only set time and assign technicians.
3. Created notifications & events I used Zapier to create events for a separate calendar. If the order is approved and the appointment scheduled, technician gets a Slack notification of a new event with the link to the invoice.
4. Streamlined order submission When manager submits a report, no copy pasting is needed anymore — she copies and pastes all the required data in the report in 1 click.
The manager copies and pastes the values in the relevant fields in 1 click.
We have managed to save approximately 12 minutes on each order processing throughout all the lifecycle and almost completely eliminated manual input issues. We have managed to slightly accelerate reports submission speed which eventually resulted in +7% increase in customer service score.Importantly, 3 weeks after going full-scale all the affected departments were largely positive about the changes, admitting an increase in productivity.
We’ve made a bunch of experiments that generated extra revenue. Below are some of the examples. 1. Launched email campaign: +72 orders We pulled emails from old invoices and matched them with the existing reporting system on insurance side. We launched a quick incentive-driven campaign through Mailchimp that generated 72 orders (minimum $160 revenue each). 2. Configured auto-acceptance of x2 tarrif orders: +7 orders/mo avg. We used SliqSubmitter to automatically accept rush orders sent through email. These time-sensitive work orders are billed at double price, adding more customer service points. With a properly configured autoclicker we dramatically increased our chances to use these opportunities. 3. Launched a ‘best price checker’: any deal = best deal The company relied on several 3rd party vendors as for parts supply. We have implemented a simple iframe-based web interface that allowed comparing and choosing the best deal way faster, at one place.
Our team contributed to a total of 37% increase in customer service score and went 7% beyond the target goal in just 2 months.A series of wins built a higher level of trust and cleared the road towards building a custom project management tool tailored to the specific needs of the company.