Support deflection ROI Calculations

(Stephen O'Rowe) #1

Please move me to a relevant thread if needed. I couldn’t find one just on this :no_mouth:

So, my problem is calculations around call deflection of Community. We hear a lot about ways to calculate ROI with the deflection but they all seem intangible or easily disputed and not great to present to your business as they’re pretty loose.

The way I see it, you have 3 main choices for actually calculating deflection from a support ticket, I’ve simplified them below

  1. Count up all the threads on your community where mods or customers have marked as ‘answered’ or marked that this helped them A and multiply by the average cost of handling a ticket Cost.
    (AxCost)=ROI
  2. Analyze if, after someone engaged in the community, did they also raise a support ticket within 48 hours. If yes, failed F deflection. If no, successful S deflection.
    ((S - F) x Cost) = ROI
  3. Simply but not very scientific: Review pageviews P to the community, multiply by support cost Cost, multiple by the margin of error (ME) of .025/0.5
    (PxCost)xME= ROI

I’m wondering if anyone uses something like the above, or can give me some concrete calculations they use?

1 Like

(Richard Millington) #2

Hi @steverowe, you’re not wrong. Almost all approaches contain some very obvious flaws.

The most common approach is usually something along these lines.

In your calculations above, I suspect:

  1. Ignores the people who read the answer and didn’t need to answer the question. This would massively underestimate the value.
  2. Has the same problem as 1.
  3. Would massively overestimate the value as it includes page views to any page of the site not just those with accepted solutions.

In an ideal world, I’d track page views which lasted longer than 30 seconds to questions with an accepted solution, multiply this by the % which were likely successful and then multiply that by the average cost of handling a call/ticket.

This has two VERY big problems. First, 30 seconds is rather arbitrary, it excludes bots but some people just tend to visit a lot of similar questions to get the answer. Second, the % likely to be successful is a VERY subjective number to use. Some people think this is 20%, I suspect it’s closer to 5%. But there isn’t a good methodology for determining if people were successful or not.

Perhaps the best one I’ve seen is when people go to file a ticket they are shown related discussion in the community, if they then don’t file a ticket, it counts as a ticket deflected. This requires a certain level of technology work though.

(it’s also worth noting you’re only talking about the return above and not the return on investment).

1 Like

(Stephen O'Rowe) #3

Hi Richard, thanks for your insight. Glad to hear it’s not a problem I struggle with alone!

Another problem is people not marking their questions as answered which requires a lot of moderation on our side, plus providing workarounds to answers can sometimes leave us wondering if we should mark it as ‘answered’ - (perhaps we need more specific statuses).

What I am thinking about now is analysing on a member level; we can combine individual member’s pageviews/engagement, with whether they raised a ticket with our support team in a set timeframe i.e. they viewed a community conversation (Post +1 comment) and then that account didn’t raise a ticket within 48/72 after the visit = successful visit.

We’d then count up all the successful visits and determine that it deflected a support ticket. We’d also need to add a likelihood of success here too of between 5-20% as you’ve suggested.

What do you think?

1 Like