A great deal has been written about whether, in the Internet age, your business should have a phone number on its website.
On one hand, having a phone number can increase the trustworthiness of your website, help sell potential customers who aren’t comfortable buying online, and allow customers to contact support easily.
The flip side? Phone support costs money.
Many anecdotes support both strategies, but we should be asking, “Where’s the data?”
A tale of two approaches
Here’s what Zappos founder and CEO Tony Hseih had to say about their stance on their phone number:
On many websites the contact information is buried at least five links deep, because the company doesn’t really want to hear from you. And when you find it, it’s a form or an e-mail address.
We take the exact opposite approach.
We put our phone number (it’s 800-927-7671, in case you’d like to call) at the top of every single page of our website, because we actually want to talk to our customers. And we staff our call center 24/7.”
Rackspace is another example of a company that built its brand on customer service, or “fanatical support.”
In an open-letter to customers that announced “fanatical support” for Microsoft Azure, Rackspace CEO Taylor Rhodes said:
We will be available to you by phone or ticket, 24/7 — within minutes. We won’t send you to FAQs or message boards for support. We will be there for your team, backing you up with Fanatical Support.
Jeff Bezos, CEO and co-founder of Amazon, however, has a very different take: “The best customer service is if the customer doesn’t need to call you, doesn’t need to talk to you. It just works.”
And if it couldn’t be more apparent, Rackspace offers two phone numbers after you click the “Call us” link in the menu:
Amazon Web Services? There’s no number in sight (even “Contact Sales” walks you through a decision-tree before offering a phone number):
What does this show?
Not much. These are a handful of anecdotes that prove nothing (other than that you can build a successful business with or without featuring a phone number on your site).
This post is about how you can use testing and data to determine if having a phone number is right for you and your business. I’ll take you through the exact, data-driven method we used.
Before you do anything, gather some data.
At my company, LawnStarter, we allow people to book lawn service online. Both my co-founder and I were inspired early on by Zappos, so we always featured a phone number on the page.
Booking lawn care online is fairly new in the grand scheme of things, so it was a pleasant surprise that most respondents seemed somewhat comfortable booking lawn care online.
Plus, we could probably use some of the tactics described in this article about reluctant online shoppers to push some of the lower scores higher.
Still, roughly 10% were terrified to book lawn care online. That’s evidence enough that we should keep a phone number on our site.
How to test whether to use a phone number on your site
Conceptually, it’s simple. You run an A/B test with version A featuring no phone number and version B displaying the phone number prominently.
The tricky part is attribution. While I strongly recommend connecting all your company’s data together in once place, there’s no need to invest in an elaborate call-tracking setup before you even know if a phone number is worthwhile.
This method uses a forwarding number and a little bit of spreadsheet work to do the test.
Step 1: Set up a tracking number.
The first step is setting up a tracking number. Fortunately, this is easy to do with most VoIP systems.
Please note: Use a tracking number specific to this experiment—not your existing phone number. Phone numbers get scraped by Google and other directories. If you use your regular number, some calls that are part of your test may not originate from your landing page.
Make sure your phone records can be exported to a CSV. Talkdesk and most other VoIP providers allow for this.
For purposes of this tutorial, we’ll use Twilio. This video walks you through the steps of setting up a forwarding number.
Step 2: Run your test.
This part is pretty straightforward. Simply take your tracking number and put it in the desired spot in your preferred A/B testing tool.
Version A: Leave phone number area blank (or remove it).
Version B: Insert your tracking number.
Once you have this set up, integrate the test with Google Analytics, Mixpanel, Segment, or however you’re used to tracking conversions.
Now your test is ready to run.
Step 3: Link your inbound calls with customer sign-ups.
Once you’ve run the test long enough, it’s time to look at the results:
Export your online conversion data from whichever tool you use. Put it into one tab of spreadsheet, like this:
Then, download your customer account data into a CSV and put them into their own tab. (Make sure to include customers’ phone numbers.)
In many cases, phone numbers won’t look uniform. You just need to remove spaces and erroneous characters with the following statement:
If you have 1s preceding your number, you’ll need to take it a step further using a statement like this:
(Note: I like to always multiply string operations on numbers by 1 to keep the data formatted as a number rather than a string.)
Next, download your call data. Fortunately, Twilio keeps call data clean, so you shouldn’t have to do much reformatting.
Simply import the incoming calls associated with your tracking number and paste them into a third tab in your spreadsheet:
Now, we’re going to use VLOOKUP to get call conversions. For those unfamiliar, here’s what Microsoft has to say about a VLOOKUP:
Use VLOOKUP, one of the lookup and reference functions, when you need to find things in a table or a range by row. For example, look up a price of an automotive part by the part number.
=VLOOKUP(Value you want to look up, range where you want to lookup the value, the column number in the range containing the return value, Exact Match or Approximate Match – indicated as 0/FALSE or 1/TRUE).
We’re going to use VLOOKUP to check whether the incoming caller number is also found in our customers list.
See all those cells with phone numbers in Column C? Those are customers. The error cells indicate incoming calls that didn’t result in conversions.
Let’s make it a little cleaner by using an an IF() and ISERROR() statement.
Much cleaner. It’s time to get a few data points out of here. We want:
- Total number of calls. Number of incoming calls we received.
- Total number of callers. Number of unique incoming numbers that called us.
- Caller conversion rate. Number of converted customers over the total number of callers.
You can see the formulas here:
We’ll go ahead and cut and paste these formulas into the “Summary” tab:
Step 4: Combine and check your results.
Modify your original summary tab as follows to combine the results and compare conversion rates:
In this fictitious example, we saw a significantly lower online conversion rate (for the variation with the phone number) but a higher overall conversion.
But the test isn’t finished yet.
Step 5: Account for revenue.
Phone sales aren’t free—you pay a human to make the sale, and you can’t exclude this in your analysis.
To get a general sense of how much of a true win this test was, we’re going to incorporate a few more data points:
- Lifetime value of a customer (net of all marketing and support costs);
- Number of sales reps required;
- Monthly cost per rep (i.e. their salary);
- Test period (one month).
The calculations we did are as follows:
- Unadjusted total LTV = LTV per customer * total conversions;
- Sale cost = sales reps * monthly cost per rep * test period;
- Adjusted total LTV = unadjusted total LTV – sale cost.
As you can see in the spreadsheet, we got a 9% increase in conversions but only a 2% increase in total LTV.
Next Steps: It’s not just a conversion rate problem
If we were making a simple tweak to the UI, it wouldn’t be a complicated decision to execute. However, phone support requires people, time, resources, and costs.
Here are six next steps to help you hone in on a data-driven decision.
1. Sensitivity analysis
You could do a significance test on revenue and call it quits.
But that isn’t the point. The point of testing phone versus no phone is to get a general sense of whether a phone number is a large-enough opportunity to account for the overhead.
A sensitivity analysis—”a technique used to determine how different values of an independent variable will impact a particular dependent variable under a given set of assumptions”—can help evaluate the real opportunity.
Here are some of the “what if” scenarios you’ll want to evaluate:
- An increase in phone conversions. Your phone conversion probably wasn’t optimal for the first go-round. What does total LTV look like if you can increase it?
- Decrease in rep cost. What does it look like if you decide to have your sales reps overseas at a fraction of the cost?
- Hiding the phone number. Hide the phone number at the bottom of the page, hoping to maintain your online conversion but allowing the Luddites to seek out their preferred method.
2. Device testing
3. True cost of human sales
You’re probably thinking, “Ryan, you made some pretty simple assumptions there with that analysis.” Aaaand you’re right. Hiring a sales rep costs time and money.
There’s turnover. You have to train them. While a rep may be able to handle 30 calls per day on average, you’ll likely have to increase the number of reps to handle peak times.
As the team scales, you’ll have to do call reviews and eventually hire management. Scaling a sales team and maintaining, not to mention improving, your conversion rate is no easy task.
You’ll have to account for this in your financial plan, but that’s another post.
4. Accounting for increased support costs
If you don’t currently offer phone support and suddenly put a phone number on your site, you’re going to get phone calls. You have two options:
- Start offering phone support. This changes your business model and associated costs.
- Tell customers you can’t help them. This creates a bad experience
If you have phone support and remove the number, you’ll want to make sure that customers can still solve their problems.
5. Testing infrastructure
The method I presented is quick, dirty, and requires no overhead.
But should you decide that having a phone number is right for your business, it will inevitably be part of every A/B test you do going forward. You’ll want to make sure you create the systems to track results on an ongoing basis.
For more info on this, check out a couple of resources:
Personally, I hate being subject to the limitations of out-of-the-box analytics tools. So we simply set up a MySQL table that contains all of our tracking numbers. We put all of our internal data—as well as data from third-party tools—into one Redshift database, then join it up and visualize it in Tableau.
6. Your core brand identity
Finally, you have to think about the type of brand you want to build. Do you want to be Zappos, Amazon, or something in between? The phone number goes a long way toward that determination.
A phone number on a website can be a game-changing opportunity, a huge cost center, or simply inconsequential.
And as with most things in digital marketing, you can’t rely on other companies’ case studies or “best practices.” You have to test it yourself.
Working on something related to this? Post a comment in the CXL community!