Told the agency the app was the wrong build. Told the client not to call us back either.
A digital agency's proposal for a custom mobile loyalty app to fix “low repeat purchase rate.”
A regional direct-to-consumer brand, around thirty staff, low six figures of monthly revenue, the founder still hands-on with the warehouse on Saturdays. The board had identified repeat-purchase rate as the single number that needed to move; the marketing director had circulated a memo about “owning the customer relationship”; and a digital agency the founder met at a trade conference had walked the leadership team through a proposal for a custom mobile loyalty app at a hundred and twenty thousand, with an additional twenty-five thousand a year in maintenance. The agency had previously delivered the brand's e-commerce redesign, which had gone well. The relationship was warm. The founder asked us in to second-opinion the build, mostly, we suspected, because the marketing director's brother had said something pointed about app projects at Thanksgiving.
The app is the wrong build. The repeat-purchase problem is upstream of any loyalty surface, it's a data and email problem, and you already own the tools.
We opened with the diagnosis, not the deck. Repeat-purchase rate in a brand of this size and product category sits roughly where the email programme sits, and the brand's email programme was running at a fraction of its capacity. The post-purchase flow consisted of one transactional receipt and a discount code sent twelve days later. There was no segmentation by category, no replenishment cadence keyed to product life, no win-back sequence at sixty or ninety days, no education series for the categories that needed it. The brand's email tool was already paid for and capable. The loyalty programme being proposed inside the app, points for purchases, redeemable for discounts, could be wired into the existing checkout in a week using the points module of the e-commerce platform the brand was already on. Building a native app to deliver a points programme that the website could deliver in a week was an expensive proxy for the actual gap, which was that nobody had been allowed the focused time to fix the email programme.
So they could see, by the numbers, that the app would not move the metric the board was tracking.
We pulled the brand's analytics for the previous twelve months and broke repeat-purchase rate down by acquisition channel, by product category, and by post-purchase email engagement. The pattern was clean. Customers who opened any post-purchase email at all had repeat-purchase rates roughly three times those who didn't. Customers who'd received a replenishment-cadenced email had rates roughly five times. The leverage was visible and concentrated. Then we modelled the realistic impact of the proposed app. Even with generous assumptions on download rate (twenty percent of repeat customers in year one, which is high for the category) and ambitious engagement assumptions (forty percent of installers using the points feature monthly, which is heroic), the projected lift on the headline metric was lower than the lift the email programme would produce with two months of focused work. The app would also take eight months to ship and require its own customer-acquisition spend to drive installs. We named the three things we'd do this quarter, in writing: hire a freelance lifecycle email specialist for one month (we listed three we trusted, by name); enable the points module on the existing checkout (a one-week wire-up by the e-commerce vendor); and instrument the email programme properly so the next decision could be made off real numbers rather than agency decks. Total cost: under fifteen thousand. Projected metric lift, at the eighteen-month horizon: more than the modelled app, with a confidence interval that was substantially tighter because we were modifying levers that already worked.
We told the agency the app was the wrong build and we told the client not to call us back either, they didn't need a consultancy at all, they needed a freelance specialist and a free Tuesday.
We sent a one-page recommendation to the founder, copied to the marketing director and the board chair. The first paragraph said don't build the app. The second paragraph named the three actions and the three lifecycle specialists. The third paragraph said, in writing, that the brand didn't need a consultancy on this problem, including us, on a follow-on engagement, and asked them to deal with whichever specialist they preferred directly. We also wrote a separate, shorter note to the agency we respected enough to send it to. We told them, candidly, that the app proposal was a misread; we'd prefer to keep the door open to refer in their direction on the work they were genuinely the best in town for (e-commerce platform work), and we hoped they'd hear the note as a courtesy. The reply we received was the kind that gets a relationship sent more work over time. The founder, when we declined the follow-on, asked twice if we were sure. We were. We took an hour's pay for the diagnosis, sent the freelance contact details, and stepped out of the conversation. The repeat-purchase rate moved seventeen percent over the next two quarters. No one needed an app.
A named person sat across the boardroom and said “don't hire the agency for this, don't hire us either, hire a freelancer for a month, and instrument what you already have.” In writing, on a single page, with the math on the second side. Not a faceless “we”, not a discovery phase to defer the conclusion, not a recommendation that hedged its way into a phased advisory engagement. The signature on a no costs the firm something visible and pays it back invisibly; this one paid back inside the year, in the kind of referral that doesn't get categorised on a CRM dashboard because it arrived in a private message from one founder to another.