Skip to content
Faisal Mahmud TusherDigital marketing, SEO & analytics
with Faisal Tusher

Digital Marketing › Marketing Strategy

Marketing Decisions Before You Have Data

How to make marketing decisions for a new business with no history: targets from the economics, one segment and channel, borrowed evidence and a first baseline.
Guidebeginner12 min readUpdated

What changes when there is no history

An established business plans from its own records: last year's sales by month, the cost per customer in each channel, the conversion rate of the main pages. A new business has none of these. It does not know which customers will buy most readily, which messages will work, which channel will pay back or how many visits it takes to win an order. Every number in its first plan is a forecast.

That changes the job of the plan. Its first purpose is not to hit a target but to produce the evidence that later targets will rest on. A plan written this way has three properties. Every number in it is labelled as an estimate or a measurement. Every estimate names where it came from and when it will be replaced. And the activity is narrow enough that the results can be read, because a plan that tries everything at once learns nothing from any of it.

Several existing guides answer one part of this situation. The budget allocation guide has a short section on allocating with no history, the segmentation guide covers segmenting with little data, and the objectives guide answers how to set a target with none. This guide puts the whole decision set in order, and the strategy section covers each choice in more depth once there is data to make it with. If the business is launching a product into a market it already knows, the go-to-market guide is the better starting point; if the accounts and tracking are not yet in place, the marketing setup tutorial comes first.

Targets from the economics, not from benchmarks

The most reliable number a new business has is its own margin. It knows, or can work out, what a sale is likely to bring in and what it costs to deliver: materials, staff time, delivery, payment fees and any other cost that rises with each sale. What is left is the Contribution Margin, and it sets the most the business can afford to pay to win a customer before marketing starts losing money: the break-even cost per acquisition (Break-Even CPA). The contribution margin and break-even guide walks through the calculation, the break-even calculator does it, and the CPA calculator checks the first results against it.

The break-even line is a ceiling, not a target. A new business usually sets its working target below it, for three reasons: its margin estimate is itself uncertain, attribution in the first months is imperfect, and some profit has to be left over. Where customers reliably come back, repeat purchases can lift the ceiling, but only when there is evidence that they will; a new business usually has none, so it plans on the first sale alone and treats repeat value as a later bonus.

Published benchmarks are tempting because they look like data. The guide to using industry benchmarks explains what they can and cannot do: a benchmark is a summary of someone else's sample under someone else's definitions. It can give a rough sense of scale for a metric that is new to you, and a figure far outside a matched range is a reason to check the tracking. It cannot tell you whether your figure is good, and the guide warns that a benchmark turned into a target usually does harm. Use benchmarks to sense-check, and set the targets from the economics.

NumberWhere it comes from firstWhat replaces it
What a customer can costContribution margin on a first saleThe same calculation on measured order values and costs
Cost per click or per thousand impressionsPlatform planning tools, before launchThe account's own prices once campaigns run
Conversion rate from visit to enquiry or orderA deliberately cautious estimateThe site's own rate, read as a range
Share of enquiries that become customersThe founder's experience of sellingThe measured share from the business's own records
Size of the reachable marketSearch demand and audience sizesImpression share and the point where more spend buys little more

Start narrow: one segment, one offer, one channel

With no evidence about which customers, messages or channels work, the instinct is to try many at once. It is the wrong instinct, because each slice of a small effort produces too few results to read, and the business ends the first months knowing no more than it started with. Narrowing is how a new business learns quickly.

One segment. Choose the group with the most urgent need, the clearest route to reach it and the best economics. The segmentation guide's advice for businesses with little data applies: treat each candidate group as a hypothesis, weight reachability heavily, and let the first customers correct the guess.

One offer. Lead with the product or service whose demand is clearest and whose margin can carry the cost of acquiring a customer (CAC). The rest of the range can stay visible; it simply does not receive the first marketing effort.

One channel, chosen where demand is visible. Step one of choosing channels is to locate existing demand: searches for the category and the problem, marketplaces and review sites, and existing contacts and referrals. Demand that already exists is the cheapest to serve and the quickest to read, so it is the natural first channel. If almost nobody searches for what the business sells, the job is to create demand, which is slower to show results; the budget allocation guide's advice is to treat any creation channel as a deliberately small test beside the capture channel rather than as a commitment. The guide to marketing on a small budget covers the same concentration from the money side.

Evidence you can borrow and how far to trust it

A new business has no data of its own, but it is not without evidence. Four kinds are available before the first sale, and they deserve different amounts of trust.

EvidenceWhat it can tell youWhere it misleadsHow far to trust it
Conversations with people who have the problemHow they describe the need, what they use now, what made them switch beforePeople are polite about ideas and poor at predicting their own behaviourHigh on the problem and the language; low on whether they would buy
Search demandWhether people look for the category or the problem, and in what wordsVolume estimates are approximate; searches are not purchasesGood for choosing a first channel and the words to use
Competitors' presenceWhich positions, channels and offers rivals have chosenTheir choices fit their economics and history, not yoursUseful for seeing what is occupied; weak as a guide to what works
The founder's own sellingWhich customers said yes, which objections came up, how long a decision tookA handful of sales, often to people who knew the founderStrong on objections and process; weak on volume

Conversations come first because they reach the reasons behind behaviour. The customer interviews tutorial explains how to plan them around a decision and keep them on what people have actually done, rather than on what they say they would do, which is where interviews about a new product usually go wrong. Search demand comes next, because it shows whether the need is already being expressed in a form a channel can reach. Competitors' presence shows the ground that is taken. The founder's own experience of selling, often the richest evidence in a very new business, is the easiest to over-trust, because early customers are rarely typical.

Write each borrowed piece of evidence down with its source and the date, as an observation rather than a conclusion. That is the research step of the experimentation loop: an insight log that later hypotheses can be traced back to.

Reversible and irreversible early decisions

Caution is a limited resource in a new business, and spending it evenly is a mistake. The one-way and two-way doors framework sorts decisions by how hard they are to undo. Most early marketing decisions are two-way doors: ad copy, a landing page headline, which keywords to bid on, the time of day an email goes out. They can be made quickly, watched against a stated expectation and reversed on a set date if they miss.

A few are one-way doors and deserve research first: a long contract with a supplier or agency, a large production run of printed or physical material, a name or identity, a public price that customers will remember, a commitment to a channel that needs months of investment before it shows anything. The framework's most useful advice for a new business is to make these doors smaller where possible: a short trial instead of a long contract, a small print run, a price tested on a limited audience before it goes on every page.

Writing the classification down, in one line beside each decision, lets others challenge it before the work starts. It also keeps the founder from agonising over headlines while signing a long contract in an afternoon.

Reading the first results while numbers are small

The first weeks produce small numbers, and small numbers jump around. A conversion rate built on a handful of enquiries can halve or double from one week to the next with nothing real changing. The guide to reading small numbers sets out the habits that prevent false conclusions: read every rate beside the count beneath it, think in ranges rather than single figures, and widen the window until the count is large enough to read.

Three practical rules follow for a new business.

  • Run small tests with a stated end. Each test has a Hypothesis written in advance, a measure, and a date or a count at which it will be read. A test with no end point is just spending.
  • Set the reading rule before the result. Decide what result would mean continue, change or stop, and how many results are needed before deciding. The Sample Size may be modest, but it should be chosen before the data arrives.
  • Keep a learning log. One line per test or change: what was tried, when, what was expected and what happened. A new business changes many things at once, and without a log a good or bad month cannot be matched to its cause.

Asking every new customer how they heard about the business fills gaps that tracking misses, especially for word of mouth, and costs nothing.

When estimates become a baseline

Each estimate in the first plan should be replaced by a measured figure as soon as the measurement is solid enough, and not before. A figure is solid enough when the ordinary week-to-week movement in it has become smaller than the differences the business would act on; the small-numbers guide suggests setting that point from the business's own history once it has some, rather than from a figure found elsewhere.

The replacement happens one number at a time, usually in the order the counts grow. Prices paid for clicks or impressions become reliable first, because they arrive in large numbers. Conversion rates from visit to enquiry follow. The share of enquiries that become customers takes longer, and the value of repeat purchases longest of all. Keep a short record of each replacement: the estimate, the measured figure, the period and the count behind it.

When the main numbers in the plan, the cost per customer, the conversion rate, the share of enquiries that buy and the margin, are all measured rather than estimated, the business has a baseline. From that point the objectives guide applies in full: targets are set as a change from a baseline, measured the same way, with a deadline, and each KPI is judged against it. The budget allocation guide's marginal-return method becomes usable at the same time, because there are now points on the curve to allocate on.

A kayak-hire business's first season (illustrative)

Illustrative example — not a real result

Paddlewhim Kayak Hire replaces its estimates with measurements

Paddlewhim Kayak Hire, an invented business on a lake, opens for its first season with eight kayaks and no bookings history. A booking averages two people at £38 each, so £76. After staff time, insurance, wear and the booking system's fee, about £46 of each booking is contribution, so the break-even cost per booking is £46. Uncertain of its own figures, the owner sets a working target of £20.

She talks to 15 people at the lake shore and in a nearby outdoor shop. Most describe looking online for "something to do on the water" the day before or the morning of their visit, and several mention giving up on other hire firms because they could not see which times were free. Search planning tools show steady searches for kayak hire in the area, and two rivals appear in paid results. She chooses one segment, visiting couples and friends booking a day or less ahead, and one channel, paid search on hire terms in her area.

Her plan assumes 6% of search visitors book. Classing the fleet size as a one-way door, she rents two extra kayaks for the season rather than buying them, and treats every ad and page change as a two-way door.

NumberEstimateWeeks 1 to 6Weeks 7 to 12
Search spend£100 a week£540£600
ClicksNot estimated410470
Visit-to-booking rate6%4.6% (19 bookings)6.2% (29 bookings)
Cost per booking£20 target£28£21

After six weeks the cost per booking is above target but well under break-even. The learning log points to the booking page, which hid the available times behind two clicks. She moves the timetable to the top of the page in week seven. In weeks 7 to 12 the rate rises, though with 29 bookings behind it the plausible range runs from about 4% to 9%, so she reads it as an improvement rather than a precise new figure.

At the end of week 12 she writes the baseline: £21 per booking from search, a visit-to-booking rate of about 6% read as a range, and 35 further bookings from walk-ups and word of mouth, recorded by asking every customer how they heard. Next season's plan starts from these measurements instead of the estimates.

Common mistakes

  • Setting targets from published benchmarks. They describe someone else's business; the margin on your own sale sets what a customer can cost.
  • Treating break-even as the target. It is the ceiling; the working target sits below it until the margin itself is measured.
  • Trying every channel at once. Each slice buys too little to read, so nothing is learned from any.
  • Trusting polite interest. People are generous about ideas; what they have done before is better evidence than what they say they will do.
  • Equal caution for every decision. Spend it on the decisions that cannot be undone and move quickly on the rest.
  • Reading one week. Small numbers swing; read the count beside every rate and widen the window.
  • Forgetting to replace estimates. A plan still running on its first guesses months later has stopped learning.
  • No learning log. Without a record of what changed and when, results cannot be matched to causes.

Frequently asked questions

Can I use industry benchmarks to set my first marketing targets?

Not as targets. A published benchmark describes other businesses' samples under their own definitions, and the site's guide to using benchmarks warns that a benchmark turned into a target usually does harm, because the target carries none of the reasoning and all of the mismatch. Use benchmarks to get a rough sense of scale for a metric that is new to you, or to notice a figure so far out of range that the tracking deserves a check. Set the targets themselves from your own margin, prices and capacity.

How long before a new business has enough data to plan properly?

There is no fixed period, because it depends on counts rather than on time. A business that wins many customers a week reaches a usable baseline far sooner than one that wins a few a month. The test is whether the ordinary week-to-week movement in a figure has become smaller than the differences you would act on. Until then, read results over longer windows, keep the estimates labelled as estimates and replace them one at a time as each measured figure becomes solid.

Should a new business try several channels at once to see what works?

Usually not. Splitting a first budget across several channels buys each of them too little to read, so the business learns nothing from any. Start where demand is already visible, often search, and give that channel enough to produce a readable result. If the plan needs to learn about creating demand as well, run one small, bounded test in one creation channel beside it, with its own measure and end date, and add channels only when the first has a measured result.

Terms used on this page

Tools to try

Frameworks

Related