Financial records people can trust.A platform that’s easier to change.
I find where money leaks between checkout and the books, fix the records underneath, and leave a platform other teams can build on. At Grove, that made the financial records IPO-ready.
Where I help
Find the leaks
Over-refunds, orders that don’t match the ERP, exceptions handled by hand.
Fix the source
Keep the facts of each transaction instead of patching symptoms.
Make it reusable
Billing, pricing and fee capabilities the next team can build on.
Grove · $270M in annual GMV
$3Ma year
in refund leakage closed
About $250K lost every month, adding upRefunds matched what was paid within six months
5%→0.2%
of transactions flagged for reconciliation
1 in 20 needed a manual fix; ~$100K a month written off1 in 500, within a year
−99.5%
realized account-takeover fraud
$430K a year lost to account takeovers~$2K a year within a year, from login and order-flow changes
40%→10%
of platform engineering time spent on upkeep
40% of the team’s time went to upkeepHalved in two years, then halved again
…known for his ability to communicate effectively across the organization, and to create and maintain alignment across some of the most precious projects in the company.
Vanessa Garber, manager at Grove · LinkedIn recommendation
That’s the outcome. Step in to see how it unfolded. About a minute.
One overloaded record, retired one job at a time.
When I started, the shipment record behind Grove’s $270M in annual GMV did nearly everything: charging, refunds, order history, subscriptions, offers, even the cart. Every change was slow, and teams saw different versions of the same order. Each project I owned had to pay off on its own and retire one of those jobs. After the 2022 layoffs I took on what had been a Director’s and two PMs’ scope, and from 2023 I was the only product manager on Grove’s platform.
Each project started with a business need. The product decision was how to meet it while leaving the platform easier to work with.
01
Financial integrity · 2020–2021
Stop reconstructing what the customer paid.
Over-refunds exposed a deeper problem: the records didn’t keep the facts needed to explain or reconcile an order.
Before · two units bought, one at 50% off
$10.00
refunded for the discounted unit. The record kept only a blended total, so Support fell back to full list price.
After · amounts kept per unit
$5.00
refunded: exactly what was paid, read from the stored transaction instead of an expired promotion.
The problem
Records drifted apart between shipment, order, invoice and NetSuite. Finance and Ops needed a principal engineer to reconcile the exceptions every month.
What I decided
Group the real exceptions with Finance and Ops, rank them by money and frequency, and keep the per-unit amount paid at the moment of charge. I sequenced the fixes to meet IPO-readiness deadlines.
~$100K/moin write-offs from order-to-NetSuite mismatches, addressed
2 weeks → 1 dayto clear the monthly exception backlog
“A key ingredient to what has allowed Grove to go public.”
02
Shared data and controls · 2021–2022
Give customers, Support and Finance the same order.
Different teams needed different views of an order, not different versions of what happened.
Order History API
Customer order history
Support tools
Finance reporting
Operations
Analytics
The problem
Customer history, Support tools and financial reports disagreed. Agents logged into customer accounts just to see what the customer saw.
What I decided
Define the facts each consumer needed and make one API the shared source. Keep financial distinctions in the data and tailor only how each screen shows them. SOX access controls for the IPO were sequenced alongside.
40 → 5 hrsmonthly engineering upkeep for the Support plugin
62 → 16open Support tickets, with 100 resolved in 2022
“…his ability to combine a deep understanding of the product’s technical intricacies with exceptional thoughtfulness.”
03
A reusable fee capability · 2022
Build a fee that could outlive its policy.
Grove needed a Supply Chain Fee right away. The lasting decision was how to ship it without hard-coding it.
Fee policy
retiredSupply Chain FeeOpt-in sustainability fee
Fee mechanics
Hard-codedReusable fee capability
TestRoll outRebuildReuse
The policy ended. The capability kept running.
“A great example of a scrappy(ish) test → full rollout to get impact early → build sustainable implementation.”
04
Pricing and offer boundaries · 2022 onward
Make pricing possible to change again.
Four teams had each learned to avoid pricing changes, for four different reasons. I treated that as one business problem, and aligned seven teams on a single pricing source of truth.
Marketingsaw most price changes as too hard to try
Designcouldn’t tell what each displayed price meant
Engineeringfaced a large, expensive rewrite
Merchandisingwanted offers the system couldn’t express
What I decided
Separate how prices are recorded from how orders are charged, so new offers could sit on a stable charging base. Offer rules became explicit: subscription offers take precedence instead of stacking.
What carried forward
A Pricing Adjustment Service scoped for Subscribe & Save’s launch and the tests we expected next, plus a pricing primer so teams stopped rediscovering the rules.
“I am continually impressed as I learn more of the calls you made on what to include vs. not.”
05
Subscribe & Save · 2022–2024
Use longer-term learning to decide what to fund next.
The first read didn’t justify a rollout. Later evidence led to a new investment decision, not a reinterpretation of the old one.
Early read. The first test didn’t justify continuing.
Longer-term signal. With Analytics: established customers ordered less often, but subscribers stayed longer and bought bigger baskets than the control group.
New tests. We moved from a blanket discount to selected brands, with tools for Merchandising and Marketing.
Core program. Grove-funded discounts built the repurchase evidence that brought in vendor-funded discounts.
55% → 85%of products in cart eligible for Subscribe & Save
06
Shopify migration · 2024–2025
Define what should survive the move.
The move to Shopify cut what it cost to run the business. I set the compatibility contract and cutover order across five systems; the earlier platform work made that behavior easier to separate and carry over.
PreserveOrder, billing and Support behavior worth keeping
DropObsolete behavior not worth carrying over
SequenceThe cutover order, and middleware contracts for what remained
The same five habits ran through every project, whether the reason to invest was financial integrity, operator time or a new commercial program. Here’s what they’d look like on your team.
One fix for a problem several teams are working aroundI connect what Finance, Support, Marketing and Engineering each feel before deciding whether it needs separate fixes or one shared capability.
Value now, flexibility only where it’s earnedShip the immediate business outcome, and build in only the flexibility a concrete next step calls for.
A sequence that pays off at every stageI work with Engineering on dependencies so each stage delivers value on its own, not one big bet at the end.
A domain your teams can build on without mePRDs, pricing definitions and lifecycle audits, so teams stop rediscovering the rules.
Priorities people can seeThe decision, the reasoning and the timing, shared where stakeholders meet the work, not only in planning meetings.
Where the numbers come from
5% → 0.2% exceptions, −90% deposit liabilities needing review, ~$100K/month in write-offsManager, 2020 performance review, with peer feedback from the engineering team
$3M a year in refund leakage2020 refund analysis: excess refunds were ~30% of yearly refunds, closed across three projects. The over-refund fix alone was estimated at ~$1.8M a year in my manager’s 2020 review.
“A key ingredient to what has allowed Grove to go public”Manager, 2021 performance review
Upkeep 40% → 20% → 10% of platform engineering time; 40 → 5 hours of Support plugin upkeepPlatform team’s scheduled work as backlog owner: 20% in 2022 (self-reflection), 10% by 2024
Quotes from Vanessa Garber and Alexandra Marchesano GilbreathPublic LinkedIn recommendations
62 → 16 open Support ticketsManager, 2022 performance review
−99.5% realized account-takeover fraud ($430K → ~$2K a year)$430K for 2022 from my manager’s 2023 performance review; later figure from fraud loss reporting. Attempted fraud fell too.
2 weeks → 1 day to clear the monthly exception backlogOperations estimate; it varied month to month
Before platform work, I launched Grove’s first Android app (10% of annual order volume, 500K+ downloads, 4.8 rating), ran growth analysis at thredUP, and built revenue-cycle data systems for healthcare clinics at Sunbelt Medical.
Akhil J Sathyan
01 · Financial integrity · 2020–2021
Stop reconstructing what the customer paid.
Over-refunds exposed a deeper problem: the business was not preserving the transaction facts needed to explain and reconcile an order.
The problem
A customer could buy several quantities of the same product with different discounts. Support could not reliably recover the amount paid for each quantity from an expired promotion or an old sale name. The fallback was to refund full list price.
Further discrepancies appeared as records moved through shipment, order, invoice and financial systems. Differences in quantities, discounts and tax meant Finance and Operations needed engineering help to reconcile exceptions.
What I decided
I grouped real exceptions with Finance, Operations and Engineering to identify recurring causes. I prioritized by financial impact, frequency and the effort required to resolve them, including work that needed a principal engineer.
We needed to retain what the customer actually paid at the quantity level, rather than reconstruct it later. With Engineering, I sequenced changes to charging, order records and downstream propagation so we could fix the pipeline incrementally while meeting IPO-readiness priorities.
The result
From 2020 to early 2021, transactions flagged for reconciliation exceptions fell from about 5% to 0.2%, and customer-deposit liabilities needing manual review fell 90%. The work eliminated the full-price over-refund behavior.
Operations reported that clearing the monthly reconciliation exception backlog went from roughly two weeks to about one day of elapsed time. Removing recurring investigations freed principal-engineering capacity for higher-value work.
The decision. Preserve the facts of the transaction instead of depending on a later reconstruction of pricing rules.
What it left behind. Quantity-level discounts and paid amounts became usable for refunds, customer explanations and financial reconciliation.
02 · Shared data and controls · 2021–2022
Give customers, Support and Finance the same order to reason about.
Different teams needed different views of the order—not different versions of what had happened.
The problem
Customer order history, Support-agent tools, manager views and financial reporting did not all show the same information. Agents sometimes accessed customer accounts just to understand what the customer was seeing.
Vendor-funded discounts, promotional offers and customer benefits had different business meanings. Finance needed those distinctions, but customers needed a clear explanation of the price they paid.
What I decided
I worked with the teams to define the facts each consumer needed and the capabilities that could not be lost during migration. Order History API v3 became a shared source of order information for customer and internal surfaces.
We preserved financial distinctions in the data while tailoring how each surface presented them. I documented the business rules and tooling requirements in PRDs, so consumers could adopt the shared source without rediscovering its meaning.
For SOX controls ahead of the IPO, I sequenced the removal of customer-account access and interim logging of administrative access while the shared-data and tooling work continued.
The result
Shared order facts supported customer history, Support, Finance, Operations and Analytics. Operational and analytical views could use the same transaction definitions instead of reconciling competing representations.
By 2022, the Support plugin had completed its move from Shipment to OMS. Recurring engineering maintenance fell from 40 hours to five hours per month.
The decision. Agree on business meaning and required detail, then migrate consumers without losing essential behavior.
What it left behind. The same underlying order facts supported customer explanations, internal tooling, refunds and analysis.
03 · A reusable fee capability · 2022 and later
Build a fee capability that could outlive the original policy.
Grove needed a Supply Chain Fee immediately. The lasting product decision was how to support the policy without permanently hard-coding it.
The immediate business need
We needed to add a fee line item to orders and support eligibility rules and exclusions. I balanced the urgency of the rollout with the cost of building sustainable infrastructure.
What I decided
We started with a limited test, rolled out the fee to realize value, then invested in a sustainable implementation. I defined the scope for applying and representing fees without repeating special-case logic for each new policy.
The reusable capability separated fee mechanics from the commercial decision about which fee to charge and when.
What carried forward
The Supply Chain Fee was later removed. Its infrastructure was reused for an opt-in sustainability fee: a different customer proposition built on the same capability.
The decision. Deliver the immediate policy, then invest in reusable mechanics where the continuing business need justified it.
What it left behind. The original fee could end while the infrastructure remained useful for another proposition.
04 · Pricing and offer boundaries · 2022 onward
Make pricing possible to change again.
Different teams had learned to avoid pricing for different reasons. I treated that shared constraint as a business problem.
The problem across teams
Marketing treated many pricing changes as too difficult to pursue. Design lacked clarity about what different prices represented. Engineering faced a large, expensive problem. Merchandising wanted capabilities that the existing systems made hard to explore.
Individually, some customer-facing inconsistencies looked too small to justify investment. Together with financial correctness and the commercial ideas teams were abandoning, they exposed a broader constraint on the business.
What I decided
I brought together customer needs, each team’s critical requirements and the behavior that had to survive. With Engineering, I chose the least complexity that met those needs and negotiated trade-offs where they conflicted. I evaluated scope and sequencing against the value we needed to deliver within the roadmap.
We separated pricing representation and historical records from charging responsibilities. That let us build new offer behavior on a stable charging foundation. For Subscribe & Save, I scoped a Pricing Adjustment Service around launch needs and the future iterations we expected to test.
Simplify behavior, not just code
Legacy offers could compete for the same slot, making one activation unexpectedly replace another. Stacking subscription offers with other discounts also produced discounts beyond what the business intended. We made the rules explicit: subscription offers took precedence rather than accumulating additional offers.
For high-complexity legacy behavior, I worked with Marketing to assess usage and agree on simpler replacements. Where necessary, we accepted a temporary capability gap while delivering the replacement in the subscription-offer system.
The leverage for other teams
I created a pricing primer and audited how pricing, product information and ordering appeared across the customer lifecycle, including first and repeat orders. Those definitions and requirements gave Design and the subscription teams a common foundation instead of requiring them to reconstruct the domain for each change.
The decision. Make business rules explicit and simplify them where possible, rather than carrying every legacy behavior into the next system.
What it left behind. Clearer pricing boundaries, reusable adjustment capabilities and shared documentation supported later subscription and site-experience work.
05 · Subscribe & Save · learning over a longer horizon
Use longer-term learning to decide what to fund next.
An early read did not justify broad rollout. Later evidence led to a new investment decision—not simply a different interpretation of the same launch.
The initial decision
The initial Subscribe & Save test did not appear to justify continuing the program. I owned the platform work spanning pricing, interoperability with existing offers and NetSuite, data models, subscription transitions, pricing displays and messaging dependencies.
What changed our view
I worked with Analytics on longer-term purchasing behavior among established customers. They were ordering less frequently, but the subscription experience showed a promising signal of stronger persistence and larger baskets compared with the control group.
We used that signal to make the case for reopening investment and running further tests. Those tests strengthened the case for moving forward.
What I changed in the investment
I supported the infrastructure variations needed for subsequent tests. We moved from a blanket discount across subscription-eligible products to supporting selected brands and products, with tools for Merchandising to manage eligibility and Marketing to run subscription-focused campaigns.
We used Grove-funded discounts to build product-level repurchase evidence. That helped Merchandising make the case for vendor-funded discounts and broader participation.
What carried forward
Subscribe & Save became a core shopping program. The platform could support successive discount strategies and a different funding model without rebuilding subscription pricing for each iteration.
The decision. Use an exploratory signal to fund further testing, then build the capabilities needed to act on the learning.
What it left behind. Selective discount eligibility, operating tools and repurchase evidence supported an expanding subscription program.
06 · Shopify and remaining retirement · 2024–2025
Define what should survive the move to Shopify.
The later migration aimed to reduce the technology investment required to run the business. The earlier platform work made existing behavior easier to separate and carry forward.
My contribution
I set the compatibility contract and cutover order for Grove’s migration off its in-house platform to Shopify, deciding which behaviors across order, billing and Support to preserve, which to drop and in what sequence.
Clearer data representations and business rules let us define middleware contracts for the capabilities that would remain. My contribution focused on those existing-system boundaries within the broader migration.
The decision. Preserve valuable business capabilities without carrying every obsolete behavior into the replacement platform.
What it left behind. The earlier work supplied both reusable capabilities and the domain understanding needed to define integration boundaries.
Walkthrough · illustrative amounts, tax and shipping left out
A refund needs the amount paid, not the old promotion name.
A customer buys two units of a $10 product. One is full price, one is 50% off: $15 paid in total. They return the discounted unit.
Unit 1 · no discount$10.00paid
Unit 2 · returned$5.0050% off
The fix wasn’t a better screen. It meant capturing the pricing facts when the order was charged and carrying them to every system that needed them.