Client retention

Share of clients retained over the trailing year. Of the clients who had an active engagement a year ago, the share that still have one today, based on engagement start and end dates.

Keeping a client is several times cheaper than replacing one, and most growth should come from the accounts you already have, which only works if they stay. The floor is 80% retained over the year, against typical annual attrition of 10% to 15%. It sits on this board against floor 80%.

This accounts for what the accounts that left were billing, across one account in all.

Each account appears once. An account is on the book while an engagement covers the date, so leaving means the last engagement closed inside the period and no other one covers its end. The largest single loss here was $215k. That figure is what the account billed in the period it left, so it understates a full year of it, deliberately: this section measures what the period lost, not what a renewal would have been worth. A client is on the book when an engagement covers the date. An engagement with no closing date is treated as still running, which is the one place an absent value is read permissively here.

The thing to understand about retention is that it reports on decisions already taken. A client who left this quarter decided months earlier, usually somewhere between a delivery that slipped and a conversation nobody had. So the causes of this number are not inside this number: they sit in delivery, in response times, and in how much attention each account was actually getting, which is mostly a function of how many accounts its lead was carrying. On this workspace the accounts that left were Foundry Athletics. They are named above with what each was billing and when it closed, which is what turns those three causes from a list into a question you can put to the people who ran them.

Do not start with the accounts that left. Start with the ones still here that look the same: a slipped delivery, a quiet quarter, or a lead carrying more accounts than anybody else.

This number cannot be moved backwards, and the only value left in it is as a description of what a departure looked like HERE. That description is now available: the accounts above closed inside this period, and what they had in common is the question worth an hour. The at-risk accounts still here are knowable this week from delivery and contact. Read it alongside income per account executive as well: a retention dip with a heavy book per seller is the overload pattern, and its fix is rebalancing rather than effort.

This closes no measured driver above, and does not claim to.

Manage the inputs weekly and keep this as the quarterly scoreboard it is. At-risk accounts, response times and the account load behind them are all knowable long before a renewal is.

The structural mistake with retention is treating it as a metric to manage, which it cannot be, because it moves after the fact. What can be managed is the set of signals that precede it, and the one most often missed is workload: a lead carrying more book than the band allows drops the quiet accounts first, and the quiet accounts are precisely the ones that leave without ever complaining. Putting a standing review on those signals is the change. The product can support it once it can name the accounts; the review itself is yours.

This closes no measured driver above, and does not claim to.

Retention is the compounding one. Replacing a client costs several times what keeping it costs, and the replacement arrives with no context, a ramp, and a first project priced to win rather than to earn. A book that churns quietly therefore runs faster to stand still, and none of that effort shows up as a line anybody reviews. It also poisons the tile next to it: a firm replacing clients reads a rising new-revenue share and can mistake it for growth, which is the same set of facts told as good news.