Career Change to Data Analyst at 40

Thirteen years as a CPA. Why leave?

Because I realized I was going to do the exact same thing every January through April for the rest of my working life. Tax season. Every year. The same forms, the same deadlines, the same clients calling at 9 PM on April 14th because they forgot to mention the rental property in Sarasota. I was at a firm called Hadley & Pierce, about 40 people, in West Hartford. Good firm. Good people. I just, I couldn't see myself doing another thirteen years of it.

The thing that actually pushed me over was this one week in February 2024. I was auditing the books for a mid-size manufacturing client, CNC Machining Solutions. They make precision parts for aerospace. And I found something interesting in their cost data. Their scrap rate on titanium parts had gone up 40% in six months but nobody at the company had noticed because the raw materials budget was aggregated at the plant level, not by material type. I flagged it in the audit and the controller, this guy Dennis, he said "huh, I'll look into that." And then nothing happened. I moved on to the next client.

But I kept thinking about it. Because that scrap rate thing, if someone had actually dug into it, could have saved them maybe $200,000 a year. Titanium stock is expensive. And the data was right there. Nobody was looking at it the right way. And I realized that's what I wanted to do. Not audit the numbers. Investigate them. Follow the thread.

How did you make the switch?

Google Data Analytics Certificate first. $39 a month, took me about four months doing it at night after the kids went to bed. It taught me the basics of SQL and Tableau and spreadsheet analysis. Honestly, the spreadsheet stuff I already knew cold because, you know, thirteen years of Excel. The SQL was new. I remember the first time I wrote a JOIN and it worked and I felt like I'd done magic. Which, my husband Aaron thought was very funny because he's a software engineer and to him a JOIN is like tying your shoes.

After the Google cert I did a bootcamp. Thinkful. Ten weeks, $7,500. That one hurt financially. Aaron and I talked about it for a long time. That's a lot of money when you have two kids and a mortgage. But I'd done the math. My salary at Hadley was $87,000 after thirteen years. Data analyst roles in Hartford were listing at $70,000 to $90,000 for mid-level. So even if I came in at the bottom end, I wouldn't take much of a hit, and the ceiling was higher.

The bootcamp was where I learned Python, pandas specifically. And dbt, which I use every day now. But the most useful thing the bootcamp gave me was the portfolio project. I did an analysis of Connecticut property assessment data. Downloaded every residential property assessment in Hartford County from the state's open data portal, about 180,000 records. Built a model that predicted assessment accuracy by neighborhood, and found that assessments in lower-income neighborhoods were systematically higher relative to actual sale prices than assessments in higher-income neighborhoods. Meaning poorer homeowners were overpaying in property taxes relative to their home's actual value.

That project is what got me the interview at my current company. My manager, Deepa, told me later that she'd seen fifty bootcamp portfolios that were all the same Spotify dataset or the same Airbnb analysis. Mine was the only one that used local data and found something that mattered. She said, "You think like an investigator." Which is basically what Dennis at CNC Machining Solutions should have heard from me two years earlier.

What was the hardest part of the transition?

Being bad at something. I know that sounds generic but I mean it very specifically. I was a CPA for thirteen years. I was good. I could read a balance sheet in my sleep. I could spot a depreciation error from three rooms away. I had a reputation. Clients asked for me by name. And then I switched to data analysis and I was the worst person on the team. Literally the worst. For months.

My first week at the insurance company, Deepa asked me to write a query that calculated the loss ratio by product line for the last twelve months. Loss ratio is just claims divided by premiums. Simple concept. I understood the business logic immediately because, you know, I'm a CPA, I can do ratios. But the data lived in four different tables in Snowflake. The claims table, the premiums table, the policy table, and a product dimension table. I had to join them correctly, filter for the right date range, account for policies that were active in some months and not others, and aggregate by product line.

It took me a full day. A colleague named Rafa, who's 26 and has been an analyst for two years, could have done it in forty-five minutes. He was very nice about it. He reviewed my query and said, "This works, but you can simplify the date logic with a BETWEEN clause instead of these two WHERE conditions." And he was right. And I went home that night feeling like the dumbest person in the building. I'm 40 years old. I have a CPA. And a 26-year-old is teaching me how to write a date filter.

Aaron said, "You knew this would happen." And he was right. I did know. Knowing it and feeling it are very different things.

What transferred from accounting?

More than I expected. The big one is, I know when a number is wrong. Not because I can see the code error. Because I've spent thirteen years looking at financial data and I have an intuition for what's plausible. Last month I ran a query on policy cancellation rates and one product line showed a 3% monthly cancellation rate. Which, if you don't work in insurance, sounds like maybe it's fine? But I knew from the audit work I'd done on insurance clients that 3% monthly is 36% annual churn, which is insane for a commercial property policy. Normal is like 8% to 12% annually. So something was wrong with my query.

Turned out I was including policy amendments as cancellations because both events use the same transaction code in our system, just with different subtype flags. I caught it because the number felt wrong. Rafa, who's better at SQL than me, might not have caught it because he doesn't have thirteen years of staring at insurance financials in his bones.

Deepa told me this is the thing that makes career changers valuable. She said, "I can teach SQL to anyone. I can't teach someone to know that 3% monthly cancellation is wrong." That was the first time since switching that I felt like my previous career wasn't just dead weight on my resume.

What is it?

The grief. That's a strong word but I think it's the right one. I grieved my accounting career. I was good at it. I had clients who trusted me. I had a professional identity that I'd built over thirteen years. And I gave it up to be a junior analyst who can't write a date filter. My CPA license is still active. I pay the renewal fee every year. Aaron asked me once why I bother and I said, "Because if this doesn't work out I need something to go back to." Which is true. But also I just... I can't let it go yet. It's like keeping your ex's number in your phone. You're not going to call. But deleting it feels too final.

Eleven years in journalism. What happened?

The industry collapsed around me. That's the honest version. I was a reporter at the Oregonian covering city government. Before that I was at a smaller paper in Eugene covering cops and courts. I loved the work. Genuinely loved it. There is nothing in the world like knocking on a door and having someone tell you a story that nobody else has heard. That's what reporting is. You find the thing that's hidden and you make it visible.

But the Oregonian went through three rounds of layoffs between 2018 and 2023. The newsroom went from about 120 reporters when I started to maybe 60. The beats kept getting combined. I went from covering just the city council to covering the city council plus the county commission plus the port authority. Three full-time beats crammed into one person. My editors, and I don't blame them because they were stretched too, but they stopped having time for real editing. I'd file a 1,200-word story and get back "looks good" in ten minutes. Nobody was pushing back on my work, which sounds nice until you realize it means nobody's reading it carefully enough to push back.

The last straw was April 2024. They offered buyouts. $25,000 plus four months of health insurance. I was making $52,000 a year. I had been making $52,000 for three years because there were no raises. My wife, Connie, is a pediatric nurse. She makes $78,000. We were fine. Not comfortable, but fine. She looked at me one night and said, "You come home angry every day. That's not fine." She was right.

Why data analysis and not, like, PR or marketing?

Because I did data work as a reporter and didn't know it had a name. In 2021 I did a series on Portland Police Bureau response times. I FOIA'd two years of 911 dispatch data. About 340,000 records. I cleaned it in Excel, which, looking back, was insane. I should have used Python. But I didn't know Python existed as a tool for this. I built pivot tables, calculated median response times by precinct, by call type, by time of day. Found that priority one calls in East Portland had a median response time of 14 minutes compared to 7 minutes in the West Hills. Same city, double the wait time, and the East side has twice the call volume.

That story got picked up by OPB and KGW. I testified at a city council hearing about it. It was probably the best work I did in eleven years. And the core skill was data analysis. I just called it "reporting."

So after the buyout I did the Google Data Analytics Certificate. Learned SQL. Learned the basics of Python and pandas. The SQL clicked immediately because it's just a structured way to ask questions, and asking questions is what I've done for eleven years. The Python was harder. I'm still not good at it. But I can wrangle a dataframe and build a visualization, and at my level that's enough.

How was the job search?

Brutal. I applied to something like sixty positions over five months. Got maybe eight first-round interviews. The problem is, my resume says "Reporter, Oregonian" for eleven years. Hiring managers see that and they don't see an analyst. They see a writer. I had multiple recruiters tell me, "Your background is really interesting but we're looking for someone with more technical experience." Which means, we want someone who's been writing SQL in a corporate setting, not someone who FOIA'd police data in Excel.

The job I got, I got because of the police response time project. I put it in my portfolio. Full write-up with the methodology, the SQL I rewrote it in after learning SQL, the visualizations in Tableau. My manager, a woman named Sandra, she's a former political science PhD who also came to data sideways. She told me in the interview, "I don't care that you learned SQL six months ago. You found a story in 340,000 rows that changed a policy conversation. Most analysts with five years of experience can't do that."

That interview is the only time in the entire job search that someone saw my journalism as a strength instead of a gap.

Ten months in. What's the day to day like compared to reporting?

So the company I'm at sells outdoor gear online. About 400 employees. I'm on a three-person analytics team. My day is mostly writing SQL queries to answer questions from the marketing and merchandising teams. Things like, what's the repeat purchase rate for customers who bought a tent versus customers who bought a sleeping bag. Or, which email campaigns in the last quarter drove the highest average order value. Standard e-commerce analytics.

The thing I'm good at, the thing that transferred from journalism, is asking the follow-up question. Our marketing director, Pauline, asked me last month for a report on email open rates by segment. I built the report. Then I looked at it and noticed that the "lapsed customer" segment had a 31% open rate, which was higher than the "active customer" segment at 24%. That's backwards from what you'd expect. Lapsed customers should be less engaged, not more.

So I dug in. Turns out our email platform, Klaviyo, was miscategorizing some customers as "lapsed" because they'd bought through our wholesale channel, which uses a different order system. So they looked inactive in our DTC data but they were actually buying consistently through their local gear shop. 2,300 customers in the lapsed segment were actually active. Which means Pauline's team had been sending them "we miss you" emails with 20% off coupons. Giving a discount to people who were already buying at full price. I estimated it was costing us about $15,000 a month in unnecessary discounts.

Pauline's face when I showed her that. She just went quiet for a second and then said, "How long has this been happening?" Seven months. Since the Klaviyo migration. Nobody had questioned the segment definitions because the open rates looked healthy. I questioned them because 31% open rate from lapsed customers set off the same alarm in my brain that a 14-minute police response time did. The number wasn't wrong. It was suspicious. And suspicious numbers are the beginning of a story.

What's the biggest difference between the two careers?

Impact versus visibility. In journalism, when I published the police response time story, 40,000 people read it. I got emails. I got stopped at the grocery store. A city councilwoman quoted my data in a floor speech. That felt like impact and it looked like impact.

At the e-commerce company, finding the Klaviyo segment error saved $15,000 a month. That's $180,000 a year. But nobody outside the marketing team knows about it. It's not on a website. Nobody stopped me at the grocery store. Pauline thanked me in a Slack message and our director mentioned it in a quarterly review. That's it. The impact is real. The visibility is gone.

I think about that a lot. Whether impact without visibility is enough. My wife says I'm being dramatic. She might be right. $180,000 in savings is more concrete than any news story I ever wrote. But I could point to the news story. I could show my kids. I can't show them a Slack message from Pauline that says "good catch."

What's yours?

I miss writing. Like, painfully. Journalism is a writing job. Every day you construct sentences, build narratives, explain complex things in simple language. Data analysis is not a writing job. It's a query job with occasional writing. I write Slack messages and Notion docs and SQL comments. None of it is writing the way a 1,200-word story about police response times is writing.

I started a blog about a month ago. Just for me. I write about Portland stuff, local things, whatever I'm interested in. I posted one piece about a weird zoning variance on my old beat and got eight readers. Eight. At the Oregonian I'd get 40,000 on a good story. But those eight readers are more than zero, and the act of constructing a paragraph that works, of finding the right word and putting it in the right place, that still lights up the same part of my brain. I don't know if I'll keep doing it. But right now, typing a blog post at 10 PM after my kids are in bed is the closest thing I have to what reporting used to feel like.