Your Sales Dropped 20%: What Would You Analyze First?
Why did sales drop 20%? It’s the question that lands on your desk on a Monday morning, right after someone forwards last month’s numbers and Slack starts buzzing. Someone is already typing “what happened??” in a tone that makes it clear they want an answer before lunch, not a thoughtful investigation by Friday.
Table of Contents
Monday morning. Someone forwards you last month’s numbers, and there it is — sales down 20%, compared to whatever baseline someone picked. Slack starts buzzing within the hour. Someone’s already typing “what happened??” in a tone that makes it clear they want an answer before lunch, not a thoughtful investigation by Friday.
This happens constantly, and how you handle the first hour of it says more about you as an analyst than almost anything else you’ll do that week. A 20% drop isn’t a root cause. It’s a symptom, full stop. Your actual job isn’t to explain the number away — it’s to figure out what’s really going on underneath it, which means fighting the urge to blurt out a plausible-sounding theory before you’ve even confirmed the number is real.
We use this exact scenario constantly in Ankashram‘s practical training, mostly because it looks deceptively simple and falls apart the second someone rushes it. So let’s actually walk through what a seasoned analyst checks, roughly in the order they’d check it — and honestly, the order matters more than people realize.
Read More : 10 Best Data Analytics Tools for Data Analysts in 2026: A Practical Guide
First: Is the Number Even Real?
I know this sounds almost insulting to bring up first, but skipping it is exactly how people burn an entire day chasing a problem that never existed. Before you go anywhere near “why did sales drop,” confirm that sales actually dropped.
Did someone pull the wrong date range? Is there a timezone quirk shoving a few hours of transactions into the wrong reporting bucket? Did a pipeline silently break somewhere, quietly dropping records instead of loudly failing? Did the definition of “sales” quietly shift — gross versus net, before or after refunds — without anyone updating the dashboard everyone’s now panicking over?
I’ve seen this exact scenario play out for real: a whole leadership meeting gets called, everyone’s stressed, and it turns out an ETL job broke two days earlier. Nobody’s business actually declined. The data just stopped arriving correctly. A ten-minute cross-check against a second, independent source — raw transaction logs, a finance export, anything not sharing the same pipeline — saves you from analyzing a crisis that was never real.
Read More : Online vs Offline Data Analytics Course Which Is Actually Better?
Second: Compared to What, Exactly?
Once you’ve confirmed the drop is genuinely there, ask the deceptively simple question nobody asks out loud: compared to what?
“Down 20%” versus last month means something totally different than “down 20%” versus the same month last year, or versus a forecast someone set six months ago. If last month happened to be your best month ever — say, December for most retail — then a rough January by comparison isn’t a crisis, it’s just the calendar doing what calendars do. But a 20% drop against the same month last year? That’s a much heavier signal, because seasonality’s mostly stripped out of that comparison.
A lot of bad, panicked analysis starts right here — comparing against the wrong baseline and then spending the rest of the day building a story on a comparison that was never fair in the first place. Nail this down before touching anything else.
Read More : Most In-Demand Data Analytics Skills in 2026-2027
Third: Is It Everywhere, or Just Somewhere?
Assuming the drop’s real and you’ve picked a fair comparison, next comes segmentation, and honestly, this single step eliminates more wrong guesses than everything else on this list combined.
Split it by region — is one market tanking while everywhere else looks fine, or is it genuinely nationwide? Split it by product — did one item flop while the rest of the catalog held steady? Split it by channel — online holding steady while in-store craters, or the other way around?
Here’s why this matters so much: a 20% aggregate drop caused by one struggling region tells a completely different story than the same 20% spread evenly across every region, product, and channel. One points at something local and fixable — a store closed, a competitor opened nearby, a regional supply hiccup. The other points at something systemic — a pricing shift, a brand-wide problem, a broader market pullback. Same headline number, wildly different response needed.
This is a big part of what we hammer on in Ankashram’s case studies, honestly — don’t just accept the aggregate number as the story. Nine times out of ten, the real story’s hiding underneath it, and you only find it by breaking things apart.
Read More : Data Analyst Course: Syllabus, Roles and What It Actually Costs in 2026
Fourth: Rule Out the Boring External Stuff
Before diving deep into your own metrics, check the obvious external explanations, because they’re common and they’re fast to rule out. Did a holiday land on a different date this year, shifting sales across a reporting boundary? Did a competitor launch a big promotion right in this window? Is there something bigger going on economically that’s hitting your customers specifically, not just you?
And don’t overlook supply. Did you actually have the product to sell? A stockout on your best-seller or a shipping delay can look exactly like a demand problem from the outside, when really it’s a supply problem wearing a demand problem’s costume. The fix is entirely different depending on which one it actually is, so this is worth ruling out early rather than assuming.
None of this needs fancy analysis. Mostly it just needs someone to ask around and check a calendar instead of assuming the entire answer lives in the sales table.
Fifth: Walk the Funnel
If nothing external explains it and the problem genuinely looks internal, walk the funnel top to bottom. Did traffic actually drop — fewer people showing up at all — or did traffic hold steady while conversion fell?
These point to completely different problems. Traffic dropping means something upstream broke — an ad campaign underperformed, marketing spend got cut, search rankings slipped. Traffic holding steady while conversion drops means something’s actively getting in the way of people who were already interested — a clunky checkout, a price hike nobody communicated well, fewer payment options, long lines driving people out the door.
Break the funnel into stages and check each one against its own history, and suddenly “why did sales drop” turns into something much more answerable, like “why did conversion specifically fall in this one segment during this one window.” That’s a question you can actually solve.
Sixth: New Customers, or Existing Ones Leaving?
Here’s a distinction people skip way too often, and it genuinely changes everything downstream. Is the drop coming from fewer new customers walking in the door, or from your existing customers buying less — or leaving entirely?
An acquisition problem points at marketing, awareness, top-of-funnel stuff. A retention problem points somewhere closer to the product itself, or customer experience, or a competitor quietly pulling your existing base away. And the fixes genuinely diverge here — throwing more ad budget at a retention problem is basically pouring water into a bucket with a hole in the bottom. You’re just paying to refill customers who are leaving just as fast out the back.
This is exactly the kind of moment where business understanding matters as much as anything technical — and it’s one of the things Ankashram leans on hard in how it teaches this stuff, because the SQL query to find the answer is easy. Knowing which question to ask in the first place is the actual skill.
Seventh: Did Something Change on Your End?
Worth explicitly checking — did anything shift internally right before the drop kicked in? A quiet price increase. A discount code that expired and nobody renewed. A new shipping fee that wasn’t there last month. Sometimes the answer’s embarrassingly close to home, and customers on the fence just quietly decided it wasn’t worth it anymore.
Lay the exact start date of the drop against a timeline of internal changes — pricing, site updates, policy tweaks, even staffing changes — and you’ll often find a smoking gun sitting right there. When a specific internal change lines up neatly with when the drop started, chase that hard before reaching for anything more elaborate.
Read More : How Data Analytics Helps Businesses Think With Data
Putting It Together
Notice what didn’t happen anywhere in this process: nobody reached for a fancy predictive model or a machine learning pipeline on day one. This got solved — or at least narrowed down dramatically — through confirming the number, segmenting it properly, ruling out the boring external stuff, walking the funnel, and lining up a timeline. Most real sales-drop investigations go exactly this way. Patient, structured, a little unglamorous.
This is genuinely the kind of scenario-based thinking Ankashram tries to build into how people actually learn this field, because textbook exercises rarely capture how messy a real question feels when it lands in your inbox on a Monday. “Sales dropped 20%, what happened?” isn’t really a SQL problem, or a stats problem. It’s a thinking problem that happens to use SQL and stats along the way to get there.
The analysts who handle this well aren’t necessarily the sharpest technical people in the room. They’re the ones who resist the urge to answer immediately, who ask “compared to what?” before opening a single chart, and who segment before they theorize. Speed matters — sure, someone genuinely wants an answer by lunch. But a fast wrong answer costs way more than a properly diagnosed one that took until the end of day to land.
Frequently Asked Questions
Q1.What should I check first when sales suddenly drop?
Whether the number’s even accurate. Check for reporting mistakes, a broken data pipeline, a timezone issue, or a quietly changed definition of “sales.” An uncomfortable number of scary-looking drops turn out to be data problems, not business ones.
Q2.How do I know if a drop is just seasonal or an actual problem?
Compared against the same period last year, not just last month — that removes most seasonal noise. And check whether a holiday shifted between reporting windows, since that alone can fake a drop that isn’t really there.
Q3.Should I segment before or after looking at the overall number?
Right after confirming the number’s real segment as early as you can. Splitting by region, product, and channel almost always shows the problem’s concentrated somewhere specific rather than spread evenly — and that completely changes what you dig into next.
Q4.How do I figure out if it’s an acquisition problem or a retention problem?
Check whether fewer new customers came in, or whether your existing customers are buying less or leaving outright. These need very different fixes, so separating them early stops you from wasting effort solving the wrong problem.
Q5.Do I need machine learning or advanced stats for something like this?
Usually not, at least not at the start. Most sales-drop investigations get solved through solid segmentation, timeline comparisons, and funnel analysis — the kind of fundamentals covered in Ankashram’s practical case studies. Save the heavier modeling for once you’ve actually narrowed down where the problem lives.