Forecasting Issue Multi Channel

Topic Views - 2307

Fore Casting Issue Multi Channel
Hi All,

I have recently started new role within a company (Travel industry) who had no resource planning function prior. I previously worked for a larger company (Retail) within a resource planning team. Previous role we had several years of data, calls, live chat and a dated but functional WFM.

In my new role I have none of the above, I have at best 2 years worth of call Data, 2018 and so far 2019 and a years worth of email data. Currently we have no WFM and I have been recruited to try and forecast and help better plan for the companies busy bookings period in January, and then the busy summer for customer services.

The company uses Mitel to route Calls and emails to agents, all agents have work in front of them unless set to Busy, emails automatically presented and then calls interrupt.

Challenges I am facing,

Creating an Accurate forecast for each month, for example January in 2018 and 2019 where affected at different times by weather, So how do I create an accurate profile for January and remove the weather related spikes or drops?

I cannot get an accurate AVHT for emails as the system has not been set to factor in the email being paused whilst an agent takes a call. I'm thinking I will have to sit alongside advisers and do a time in motion study?

The company is making major acquisitions regularly which is also altering the trends and forecast's, I'm slowly building a log (not previously done) of when events have happened (including weather and system issues)so that I can try and identify how big or small an influence this makes on the work coming into the contact centre.

So far to create a forecast I have used Jan 18 and 19's data to create a call profile. I've broken down the 2020 full month call forecast (created by using previous years growth and also considering the already forecast bookings target). Then Broken this down to a weekly forecast then down to a daily and then down to 15 minute intervals using a daily call profile based off the last 4 weeks trend and compared against Day on Day previous year.

I have also copied the trends created when marketing have sent out campaigns and have now factored this into the Months forecast. I've used Erlang to then work out the FTE. However I have points where I have 23% occupancy but if I have 1 less staff and I am at 95% occupancy for phones? I have done Erlang at 1 Hour and also at 30 minute intervals and then 15 to create an FTE requirement for each interval.

In Jan we have 1000 calls a day, 950 emails open 9 till 9. I have at the busiest point on calls, 15 agents required to answer phones at an occupancy of 81%, this is fine as I have 24 Seats. But I am yet to factor in email and live chat (Live chat is only 10 - 20 chats a day however)

I know currently just from being in the office the call centre is Overstaffed, ASA is 20 seconds on a call and 1 hour on an email, service levels are at 90% or higher and answer rates nearly always 99%

I have put together a plan for November and December too just so I can test theories, but this is based of just 1 years worth of Data and already this November's call trend is nothing like last years, yet October and September matched 2018. The good news is that at week level I am only -1 - +4% out on call volumes and MTD is currently 2% up on my forecast, its now getting it correct at lower level. Emails are similar.

I am hoping next year I will have a WFM (|Mitel do have a forecasting and resource tool which hopefully I can find out more about and convince management to upgrade too). I am still trying to obtain further data 2017 and before, but at minute am coming up short on all avenues I have been sent to look down.

If anyone has any tips I'd greatly appreciate it, I feel I am missing a simple trick or two. At minute its all excel based.

Kind Regards



Call Centre Helper

Along the right lines
It looks like you are going along the right lines.

Here are a few pointers.
1. Start going back to see if you can build a monthly demand history for the past few years. If you cannot get these off the ACD stats you may be able to get these from the telephone company. This will enable you to create a demand forecast as you go forward.

2. In terms of spikes what you need to do is to remove the spikes from the data and then add them back in. You can do this by comparing the spike with the weeks before and after.

For example There is a spike on Wednesday week 2.

Week1 Week2
Monday 100 110
Tuesday 90 99
Wednesday 85 250
Thursday 80 110
Friday 70 87

So you can see there is a 10% increase in traffic on Monday and Tuesday compared to the week before.

So working backward you can derive the underlying traffic.

Week1 Week2
Monday 100 110
Tuesday 90 99
Wednesday 85 93 + 157 Spike = 250
Thursday 80 88 + 22 Spike = 110
Friday 70 77 + 10 Spike = 87

In this way you can see the underlying demand and you can start to model spikes in the future.


Call Centre Helper

Are you really overstaffed?
You say
"I know currently just from being in the office the call centre is Overstaffed, ASA is 20 seconds on a call and 1 hour on an email, service levels are at 90% or higher and answer rates nearly always 99%"

If ASA is 20 seconds that sounds like quite low. What is your Service Level Target?

Sorry its taken so long to reply.

Thank you for the tip, I planned Jan using various articles from here and following the method you mentioned above. Unfortunately I have got it wrong, thankfully not horrifically and we have not lost contacts, but SLA's have been missed.

Comparing Jan 18 and Jan 19 to what has happened neither have followed a similar pattern, and like wise looking at the trends from the December running upto Jan I cannot see anything from a weekly level that appears to have affected/predicted the different contact trends.

Interestingly if I combine my month total contacts for Email and Voice is only 2% out, the split between voice and emails though is clearly out, One factor is that they where not counting previously carried forward emails and the way they have configured mitel meant this information has never been recorded.

I tried to re forecast factoring in the change in the split of contacts between voice/email only for it to swing completely the other way to the point the last two weeks calls follow my original forecast better than my new 'corrected' forecast.

The company is continually making changes to the website which is seeing an improvement in website conversion (CC is making a smaller % of overall bookings). Another factor is the acquisitions, its difficult with how many have been made over the last few years (trying to work out what attributes to each acquisition) and how they vary in size/time what change/affect they have made on the whole business.

Outside of peak the Contact Centre is very much over staffed in November/December voice had an ASA of 8 seconds, and 0 abandoned calls (currently there is no interest in AR). SLA for voice is 80% in 20 seconds (reason is management read that it was an industry standard) I currently think it needs reviewing. There also appears to be no concern as to budget and staffing costs.... I have come from somewhere where this was very much a concern, and feel that we actually need to be paying attention to it, as I am sure soon it will be questioned at a point where fixing it becomes a major challenge.

In peak different story, we are missing SLA sat at 75% on calls answered in 20 seconds and emails 56% (target 80% in 1 hour). We struggle to clear the back log from when we close to open (200 emails at 9:00am opening) we have pulled 8am email starts on some dates which has helped but not cleared the problem. Although now I am logging this for 2021 we will have a better idea of what this workload will be.

There's a lot of things I am seeing that need looking into, biggest struggle I am finding is working out where to start from.

- Shrinkage - never been recorded and currently not a way to accurately record this - I think its very high/plus agents can hide
- No WFM
- Struggling to get accurate email data, the carried forward issue (not being logged), Emails are handled/interrupted by inbound calls, agents also pulling several emails at once and handling at once leading to inaccurate handling times.

At the point where so much going on struggling to get my head around exactly where we are. Already have management wanting to change opening times to pick up extra bookings, when at the minute I don't think we even really know how we are really performing and where we are at.

Gotta love a challenge

Kind Regards



Call Centre Helper

Months vs weeks
One of the reasons why you can forecast better at monthly level is that you can average out the peaks and throughs, so that can provide a better base.

Lots of people do high level forecast -> weekly forecast -> daily forecast to deal with this.

In terms of forecasting a WFM system can help but not solve the problem. Most people forecast with spreadsheets + WFM. Where WFM comes into its own is scheduling.

One factor to watch out for with monthly forecasting is the number of working days in the month.

in 2018 we had more web visits in August than in September, which is odd considering August is usually quieter with people on holiday. The difference was that in August there were 31 days and 8 weekend days = 23 working days. September started on a Saturday and ended on a Sunday so there were 30 days and 10 weekend days = 20 working days.

Hi Jonty. I'...
Hi Jonty.

I'll have a look at see what the impact is of the number of weekends in the month as that would definitely affect us. Busiest booking day is Sunday evening for bookings enquiries.

I'm looking more at weekly level, and comparing weeks to previous years. (whilst noting any holidays or differences I.E first week of year). Hoping to try find some trends to look at, at minute Jan 18, Jan 19 and this year have all varied. Marketing and Acquisitions are going to send more data about prior years so I can try and see what has influenced volumes.

Kind Regards


Want to add a comment?

Not found what you were looking for?

1. Try searching through our site.
2. Still not got an answer?

Why not ask the Call Centre Helper Community? Click here to ask your question