Your retention rate is the share of the customers you had at the start of a stretch of time who are still with you at the end of it, and the whole quarrel with the number is that it counts the ones who stayed while telling you nothing whatever about the ones who went. Nobody leaves in a meeting. There is no morning where a customer writes in to announce that they are done with you and here are their reasons laid out in order. They just stop, quietly, one at a time, and the stopping does not show up anywhere until somebody goes and counts.
Órla counted by accident. She roasts coffee in a unit outside Kilkenny and sells it on subscription, a bag a fortnight to houses all over the country, and she was deep in the VAT return with the year's figures open in front of her for a reason that had nothing to do with customers. Nine hundred and forty subscribers at the start of the year. Nine hundred and eighty at the end of it. A grand year, she thought, near enough forty ahead. Then she went looking at what she had actually signed up over those twelve months, on account of the marketing spend sitting in the very next column, and it came to four hundred and thirty new subscribers. Forty ahead on four hundred and thirty in. She had lost the guts of four hundred people over the year and never once felt it happen, on account of the headline number going up the whole time.
That is the thing about the leaving, truth be told. It is silent, and it is spread thin, and every month it looks like nothing at all.
The formula, and the trap sat inside it
The sum its self is no trouble. Take the customers you finished the period with, take off the ones you signed up during it, and see what share of the crowd you started with that leaves you. Órla finished on nine hundred and eighty and brought in four hundred and thirty, so five hundred and fifty of the people she started the year with were still there at the end, out of nine hundred and forty, which is a retention rate of fifty eight percent. Written out as a sum it goes ((E minus N) divided by S) times one hundred, where E is what you ended with, N is the new ones, and S is what you started with. Everybody goes and drops the N. That is the trap, and it is not a small one, on account of leaving it out turning a fifty eight percent year into a hundred and four percent year, which is exactly the number Órla had been quietly carrying around in her head since January.
Pick your stretch of time and then stay with it, mind you, since a monthly rate and an annual rate are different animals and a great many people compare them without noticing. A shop losing three percent of its customers a month is losing better than thirty percent of them over the year, on account of the losses compounding on the smaller and smaller crowd left behind rather than adding up flat. Monthly suits a subscription, and yearly suits most of the rest of it. What you cannot do is measure the thing monthly and then quote it annually, which flatters the figure and fools nobody except your own self.
What a good retention rate looks like, and why the benchmark is a poor master
The honest answer is that it depends on what you sell and how often people need it, which is a desperate thing to read when all you wanted was a number to aim at. Órla is down at the hard end of it. Consumer subscriptions churn a great deal faster than anything sold to a business, and Recurly, who see a fair spread of them, have put average monthly churn for consumer subscription businesses up around the five and six percent mark, which reads as a retention rate in the low nineties month on month and a much sorrier one by the far end of the year. Business software sold on subscription lives up in the nineties across a whole year, and the good ones sit above that again. Then there is retail, where nobody ever promised to come back in the first place, so a returning customer counts as a win rather than an expectation.
So the number to beat is your own from last quarter, and after that it is the shape of your own curve. Bain reckoned a long while ago that a five percent lift in retention could move profit by a quarter or more, and the reason that finding has outlived nearly everything published beside it is that keeping a customer costs a fraction of what going out and replacing them costs. Órla knew the replacing cost. It was sat right there in the column beside the one she was reading.
Why the number tells you the wrong thing at the wrong time
Here is the part of it that caught her, and it catches nearly everybody. A retention rate is a rear view mirror. By the time the figure has moved enough to be worth a conversation, the people it is describing went and left months ago, and there is no writing to them now with an offer, they are drinking somebody else's coffee and they are perfectly content about it.
Worse again, the number lumps in people who were never going to stay. A quarter where you ran a heavy discount pulls in a crowd who came for the discount, and when they go at the end of it your retention rate falls through the floor and tells you that something is broken. Nothing is broken at all. You bought the wrong customers and the number cannot tell the two situations apart, on account of it counting heads and not reasons. The same works the other way, and that is the more dangerous of the two. A quarter where you signed almost nobody new will show you a lovely retention rate, since the people left are all long standing and loyal and unlikely to move, and a shop can sit there admiring a rising figure while the business under it quietly narrows.
And there is the averaging, which hides the whole of the interesting part. Órla's fifty eight percent was one number laid across two very different crowds. The houses who had been with her better than two years almost never left. The ones in their first eight weeks left in droves, and they left before they had properly begun, most of them inside the first three bags. One rate, two entirely separate problems, and the single figure she had been carrying told her about neither of them.
Where the leaving actually shows up first
Long before anybody churns they go quiet, and before they go quiet they nearly always write in about something. That is the part that never makes it into the metrics conversation.
Órla went back through the mail for the ones who had gone, twenty or so of them, picked at random out of the year. It took her an evening. Nine of the twenty had written to the shop at some point in their last two months, and not about anything dramatic either. A delivery that came Thursday instead of Tuesday. Somebody wanting to know about a grind setting. The one that stopped Órla dead was a woman who had asked twice, eleven days apart, whether she could skip a fortnight while she was away, who got no reply either time, and who cancelled the day after the second asking.
None of those were complaints, not really. They were people trying to stay, the lot of them, and a retention rate does not record a single one of them, on account of it only counting the moment somebody is already gone. The mail knew months before the number did.
A worked quarter, so the sum is not left in the air
Say you started the quarter with two hundred customers, signed up sixty over the three months, and finished with two hundred and ten. Take the sixty new ones off the two hundred and ten and you are left with a hundred and fifty of your original crowd, out of the two hundred you began with, which is seventy five percent retention for the quarter and a quarter in which you lost fifty people while your headline count went up by ten. Numbers made up for the sake of showing the working, mind you, but that shape turns up in real businesses constantly.
Then do the bit almost nobody does. Split those fifty into the ones who had been with you over a year and the ones who had not made it past their second order, and count each group its own self. If the loss sits with the newcomers you have an onboarding problem and no amount of loyalty work will touch it. Where it sits with the long standing crowd instead, something changed lately, and those are the ones to go and ask while they will still take your call.
Where a tool like Maxdesk fits, said without the sales voice
I would as soon say plainly that this is the sort of shop we built Maxdesk for than have the piece pretend it wandered onto the subject. Proper help desk software will not lift your retention rate on its own, and anybody telling you different is selling harder than I am. What it does is make sure the woman asking twice about skipping a fortnight gets an answer the first time she asks.
Maxdesk takes the support mail a shop already gets, on the address the customers already write to, and puts the whole lot in one shared inbox where messages become tickets with an owner on them instead of lines in a pile two people squint at. Each customer's history stays gathered together, so the second time somebody raises the same delivery you can see it is the second time and treat it that way. It ships a knowledge base for the questions that arrive week in, week out, the grind settings and the pause-my-order and the where is it, so those stop eating the hours meant for the awkward ones. And you can set the reply times you mean to hold your own self to and let it mind the clock, which is the whole difference between a customer who got an answer on Tuesday and a customer who cancelled on the eleventh day.
It does that on a free plan with nothing charged per head, which suits a shop that takes on a hand for the Christmas rush and lets them go again in January. Fair is fair, there is a trade in the free plan and you may as well hear it here. The free workspace carries some supporting ads, puts a small Maxdesk name at the foot of the mail going out, and keeps three months of history rather than the whole of forever, so if you want to go back over a year of threads the way Órla did, that is what the paid plans are there for.
The part this does not fix its self
Fair is fair a second time, so here are the limits. Maxdesk is email support and only that. There is no phone line behind it, nor a chat bubble in the corner of your website, nor a portal bolted on the side, so if your customers need a live voice inside the minute then a tool built around email is not your answer and I would be doing you no favour dressing it up as one.
And no tool anywhere fixes retention, which is worth saying flat before somebody buys something on the strength of a blog post. People leave over the price, or the product, or because they have moved house and the delivery no longer suits, and the finest answered mailbox in the country will not hold a customer whose coffee turned up stale. Support is one of the things that keeps people, and it is nowhere near the only one, no two ways about it. What changes is only this much. The signals stop being buried in a mailbox for eleven days, and the leaving stops being a thing you find out about your own self in March while you are doing the VAT.
Where to start, if you have never worked yours out
Do the cheap thing first. Take last quarter, write down what you started with, what you finished with, and what you signed up in between, and do the sum your own self on the back of something. Most owners doing that for the first time go and find the figure is well under what they had assumed, and the gap between the two is the more useful number of the pair.
Then go and read the last two months of mail for five customers who left. Not fifty, five. If three of the five wrote in about something before they went, and one of them never got an answer at all, you have found the cheapest retention work available to you, and it costs nothing but seeing the message in time. You can put Maxdesk on the mailbox you already have and watch it run against your own real threads inside a few minutes, which is a more honest test of whether any of this is for you than reading another page about it would ever be.