Blog/Give a bad week seven days

Erik · September 2026 · 5 min read

Give a bad week seven days

One of your reps has a bad week. I'd leave it alone, but only for seven days, because that's how long the noise takes to clear.

A sales rep in a headset at a desk, looking at a large calendar with one week of days highlighted

One of your reps has a bad week. Their closing score drops 12 points. Do you coach it?

Probably not, or at least I wouldn't. Sales data is noisy, customers are different, leads are different, and people occasionally just have a crap week. Reacting to every number that moves is a good way to spend your entire life reacting to numbers that move, while simultaneously driving your sales team mad.

There is a question hiding in there though, which is how long you're supposed to ignore it. "Let's see if it happens again" is perfectly sensible management advice, but it's also wonderfully vague. Does again mean tomorrow, next week, next month? At what point does a bad week stop being a bad week and become how someone is actually performing?

We have enough data now that I realised we could have a go at answering that. It looks like the answer is about seven days.

Give it seven days

We looked at more than 30,000 pairs of weekly scores across four sales behaviours and many companies and industries. These aren't weekly scores made from the two calls somebody happened to review for QA: they're based on all the calls we analysed for that rep.

The first thing to know is that weekly scores move around quite a lot even with that coverage. About one in ten moved by eight points or more, so a decent-sized sales team is going to see this fairly regularly. That's important, because you really don't want a manager jumping on every large movement as though something important has just happened.

We then took the larger movements and followed what happened afterwards. A typical large rise was about 12 points. After a week, roughly 60% of that improvement remained. Four weeks later, it was still roughly 60%. Large declines followed almost exactly the same pattern.

What fades from a big weekly move fades in the first seven days

Average score after a big weekly move, in points from where the rep was the week before, for rises and for falls.

+14 +7 0 -7 -14 points from the week before where the rep was the week before rises falls everything that fades, fades here then almost nothing changes week before move week 1 week later 2 weeks later 3 weeks later 4 weeks later

That's the whole reason I'm writing this post. I originally thought the interesting finding was that roughly half of a big weekly movement is still there a month later, and technically that's true. But the much more interesting bit is what happens between week one and week four: almost nothing.

In other words, most of the movement that is going to disappear has already disappeared after a week. What remains then, on average, is pretty much what you're going to be looking at a month later.

If someone's score falls 12 points this week, I'd leave it alone. A substantial part of that decline may disappear without anyone doing anything, and coaching somebody on a fluctuation that was about to correct itself is mostly wasting everyone's time. But I'd absolutely look again next week. If they're still six or seven points down at that point, I wouldn't keep telling myself they had a rough week, because the data gives me very little reason to expect weeks three and four to magically bring them back.

There's an important qualification here. This doesn't tell us that the rep changed. Maybe their lead mix changed, they moved onto a different queue, the product changed, or something else happened around them. We can see that the score moved and stayed moved; we can't establish from this analysis what caused it. But if I'm managing that team, that's fine. I've now got a good reason to go and find out.

Up is apparently the same as down

When a rep suddenly gets worse, we're pretty good at inventing reasons not to believe the number. Tough leads, bad luck, difficult customers, unusual week, let's see what happens next. When the same rep jumps 12 points after our coaching session on Tuesday, we seem to lose some of that healthy statistical scepticism. Look, the coaching worked!

Unfortunately, the data doesn't share our optimism. The rises and falls are remarkably symmetrical: both lose a substantial part of the initial movement during the first week, then mostly stay where they are. So perhaps the rule isn't "don't overreact to a bad week". It's don't believe a big move until you've seen next week, whether it's a result you like or one you don't.

This also changes how I think about weekly coaching. I've argued before that managers should look at performance weekly rather than monthly, mostly because monthly averages hide too much variation. But perhaps the better reason is that weekly data gives you enough time to not react. You see the drop, deliberately do nothing for a week, and give the noise time to remove itself. When you look again seven days later, either the problem has disappeared or it hasn't, and if it hasn't you can investigate while there's still a reasonable chance the rep remembers what changed.

With a monthly average, you don't see the movement and then see what survived; you just see the result after those things have been mashed together. With quarterly QA it's worse again. A change early in the quarter can spend months quietly becoming normal before anyone asks what happened.

There's a fairly important distinction there. Measuring more often doesn't have to mean reacting more often. It can mean being much better at deciding what not to react to.

On Monday

So I wouldn't build an elaborate process around any of this. I'd look at the behaviours I care about each week, flag the unusually large movements in either direction, and wait seven days. If they've bounced back, forget about them. If they haven't, get curious and look at the calls.

Perhaps that's one of the underrated benefits of having lots of data. You can afford to be less excitable about what it tells you. Sometimes the most useful thing your data can tell you is to do nothing, and check again next week.

CallCoach is how we score every call, which is what makes the seven-day test cheap to run on every rep rather than on the two calls somebody sampled. Try it on your own calls and see which of last month's bad weeks were still there a week later.

All CallCoach data in this post is aggregated, anonymised performance data from hundreds of thousands of calls across a variety of industries. We report the trends we spot; full statistical analysis is beyond the scope of a blog post.