How to Cut Costs Without Hurting Growth

Executive summary

• To cut costs without hurting growth, you need a model of what each dollar produced. A general ledger only records where it went.
• Company-level margin hides the answer. Contribution margin and fully loaded cost-to-serve show which customers and products carry their own cost.
• Spend bends. Plotting marginal return by category shows where the next dollar stops working, and where the last one still does.
• Direct savings are easy to calculate. The damage shows up downstream, in capacity, cycle times, implementation speed and retention.
• Pilot the cut in part of the business, measure it against a control group, and set rollback triggers before it becomes structural.

For years, Kraft Heinz looked like a cost-cutting success story. After the 2015 merger, the company became known for aggressive zero-based budgeting, stripping expenses out of the business and pushing margins higher.

Then the numbers started telling a different story. Revenue barely moved: $26.3 billion in 2016, $26.1 billion in 2017 and $26.3 billion in 2018. In February 2019, Kraft Heinz announced a $15.4 billion write-down tied largely to Kraft and Oscar Mayer, and its shares fell roughly 27% in a single day. 

Analysts began questioning whether years of aggressive cost reduction had gone too far, particularly as the company faced changing consumer preferences and needed investment in its brands and products.

The problem was not cutting costs. It was knowing which costs could disappear without weakening the economics that produced future revenue. That is the harder question for finance: before removing a dollar of spend, how do you determine what that dollar actually contributes to growth?

To cut costs without hurting growth, stop starting with the GL

A general ledger groups spend by what it was called when it was paid: salaries, software, professional fees, travel. That structure says nothing about what the money produced. Cost reviews built on GL exports drift toward the categories that are easiest to defend cutting, so travel, contractors and tooling absorb the reduction while the spend carrying the growth plan survives untouched.

The rebuild starts by attaching spend to the activity that consumes it. Time-driven activity-based costing does this without a six-month study: estimate each team's practical capacity in hours, identify the handful of activities that consume it, and assign a cost per unit of activity. A support function costing $2.4 million a year becomes a cost per ticket, split by product, customer tier and ticket type.

Then classify each cost by how it behaves. Most costs finance calls variable are step costs. An implementation team absorbs volume until it saturates, then needs two more people. Removing 15% of that function releases no cash until you cross a step down, and it consumes headroom the growth plan already assumed.

Company-level profit hides the costs you can actually remove

Consolidated gross margin averages things that behave nothing alike. Build contribution margin where decisions get made: product line, customer segment, channel, acquisition cohort. Load the full cost to serve into it, including implementation, support, customer success coverage, payment processing, returns and the account management time that gets allocated nowhere.

The results tend to reorder the business. An enterprise segment showing 74% gross margin can land in the low thirties on contribution once you count three solution engineers per deal, a nine-week implementation and quarterly business reviews. Pair that with CAC payback and retention by cohort and the picture sharpens: a segment that pays back in 11 months and renews at 118% net revenue retention is the last place to look for savings.

Capacity utilization completes the view. Two teams can carry identical cost lines at 62% and 94% utilization and represent completely different decisions.

Every spend category has a return curve. Most teams never plot it

Spend does not convert to output at a constant rate. Plot incremental output against incremental spend by category across the last eight to twelve quarters. In most companies the first tranche of paid acquisition returns a CAC two to three times better than the last tranche, so the top of that budget and the bottom of it are different decisions wearing the same name.

Driver trees make the chain explicit: SDR headcount drives meetings booked, which drives qualified pipeline, which drives closed revenue at a known conversion rate and cycle time. Once that chain exists, the marginal return of the next hire can be calculated the same way you would judge the incremental return on any spend request.

Lag is where this usually breaks. Cutting SDR capacity shows up in pipeline within six weeks and in revenue two quarters later. Brand and product investment can take three to four quarters to register. Compare the spend series to the outcome series at the correct lag, or the analysis will report a free cut whose effect has not arrived yet.

Model the second-order effect before you book the saving

A saving has two numbers: the direct cost removed and the economic consequence downstream. The second one rarely appears in the cost memo. Removing two implementation engineers saves $340,000 a year. If implementation stretches from nine weeks to thirteen, revenue recognition slips a quarter, first-year churn rises, and the reference customers the sales team depends on stop being produced.

Run each candidate through a driver-based model with minimum viable capacity defined for every function: the volume at which it breaks, and the bottleneck it sits in front of. Justifying a permanent reduction on a single planning scenario is how finance ends up defending a number it cannot support two quarters later.

Then rank the candidates on four axes: size of the saving, risk to growth, reversibility, and quality of the evidence behind the estimate. Cuts with strong evidence and high reversibility go first. A large saving with weak evidence and no way back is a bet, and it should be labeled that way in the board pack.

Pilots are how you cut costs without hurting growth

Structural reductions get made once and reversed slowly. Test them at a scale where a mistake costs a quarter of data. Reduce field marketing in three of twelve territories. Drop customer success coverage for one tier in one region.

Design the test before it starts. Name the treatment and control groups, set the baseline period, fix the measurement window, and write down the threshold that counts as success. A difference-in-differences comparison between treated and untreated territories filters out the seasonality and market movement that make a raw before-and-after read useless.

Every cut needs a tripwire

Each reduction should ship with a short list of metrics it is expected not to damage, and those metrics belong to the cut. For a commercial cut: pipeline creation, conversion rate, sales cycle length, CAC payback. For an operations cut: cycle time, service level, error rate, utilization. Anything touching the customer adds contribution margin and net revenue retention by cohort.

Set the threshold and the rollback trigger at the moment of the decision, while the case is still being argued and nobody is defending the cut yet. Schedule the review at 90 and 180 days and require it to explain the driver behind any movement. Savings almost always land. The question is what left with them.

Make it continuous

This analysis takes weeks the first time. Most teams run it once under margin pressure, then let the model go stale until the next round, by which point the cost base has moved.

Kudwa keeps costs connected to the drivers, customers and activities that generate them, so contribution margin, cost-to-serve and spend efficiency stay current between reviews. Finance can see where returns are weakening while there is still room to act.

Book a demo to see how this works on your own cost base.

‍